Inference Goes Exponential → Brute Force Implications

Our recent editions have covered the exponential curve for inference needs (h/t Siddharth Ramakrishnan of Scale Venture Partners) that sets up…

Dave Kellogg’s post on the "Brute-Force Era of AI". He rightly points out that while applying massive compute to architectures works (Sutton's Bitter Lesson), it doesn't inherently solve the efficiency problem.

We are officially entering Phase 2:

Why Does This Matter For Software?

While AI NRR might offset weak gross margins… those NRR scenarios rely on world-class, possibly unsustainable longer-term rates:

High gross margins have underpinned the business quality and investment case for software:

xkcd “Dependency” Source

ICONIQ (h/t Vivian Guo) recently released their “State of AI” report (47 slides, ungated).

Key call-outs: while AI product gross margins are projected to increase by ~1,400 bps (‘25: 45% → ‘26e: 53% → 2027e: 59%):

… the fabulously granular cost structure breakdown makes those margin increases look optimistic. “Talent” - a largely fixed cost - is only 28%-34% of the cost structure for the GA to Scaling stages. All other costs are highly variable, such as inference and cloud costs. Even for model training, there is little fixed-cost leverage, since it is an ongoing process.

Key message: There seem to be too few fixed costs to generate software economics.

Inference is the #1 non-talent cost at all stages (Beta, GA, Scaling). Ahead of infrastructure & cloud, data storage, and model training.

That said, the ICONIQ “State of AI” did include encouraging unit cost mitigants:

And as inference increases … in a volatile fashion, managing the cost of inference becomes even more critical.

As we researched this trend, we connected with Carmen Li, Dual CEO of Compute Exchange and Silicon Data, publisher of multiple AI market cost trend indices, including the Silicon Data LLM Token Price Index.

Share SaaSletter

Silicon Data - Token Index + CEO Commentary

The above LLM Token Price Index highlights:

• Increasing Costs: up 50%+ over past 6 months

• Volatility: like a ~34% decline in ~1 month seen in mid-January to February. With a similar rapid decline in June as companies take steps to mitigate the “tokenmaxxing” trend.

Key points from our research exchange with Carmen Li:

1. The Subscription Illusion vs. Commodity Reality “Revenue may look subscription-like on the surface, but underneath, the cost structure increasingly behaves more like a dynamic commodity market... Token expenditure volatility increasingly resembles the pricing behavior of energy, bandwidth, or other critical industrial inputs rather than conventional cloud software pricing.”

2. The Shift to Deployment Economics “One of the biggest misconceptions in AI right now is that model capability alone determines competitive advantage. In reality, deployment economics are becoming just as important. Two companies may offer similar AI functionality while operating under completely different infrastructure economics depending on how efficiently they route workloads, what compute environments they access, and how exposed they are to pricing volatility underneath the stack.”

3. Infrastructure Arbitrage & Regional Disparity “The same inference workload can carry materially different economics depending on where it runs, what hardware it runs on, and whether compute capacity was secured through long-term agreements or acquired dynamically in tighter market conditions.”

Morgan Stanley On Gross Margins + “Tokenomics”

Morgan Stanley’s very recent + excellent “The Moat & The Journey – Assessing Software Durability + Growth in the Age of AI” (170 pages; our selected excerpts here) covered both Gross Margins and “Tokenmaxxing” Mitigation:

MS View Recap: Some gross margin dilution offset at EBIT level; levers (like model routing or open-source) help mitigate AI COGS impact.

MS View Recap: “Token scrutiny is real, but not uniform.” + “Workflow-rich incumbents become the optimization layer: routing simple work to cheap models, reserving frontier for risk, capping low-value usage.”

With a useful - at least for our hyper-granular readers - sensitivity table wherein gross margin declines are/can be mitigated at the EBIT level:

Lazy AI: A Pet Theory

On our recent podcast with Tim Sanders of G2, he emphasized why AI engines are incentivized to provide high-quality answers (their own brand, trust, and competition).

However, a pet theory of mine relates to the AI engines’ strong incentives to be lazy. See these inference costs. Think through the costs of scaling web search and crawling. Especially into higher-inference knowledge sources like audio and video.

After all, what do I know if they missed the equivalent of page 16 of Google Search results?

AI engine outbound web traffic data, in absolute terms, suggests very few are clicking on the provided source links. If not even checking sources, are users exploring what was missed?

Meaning AI engines can (and should?) be lazy.

For our institutional investor audience, there are read-throughs to the respective moats and unit economics of Google, OpenAI, and Anthropic if this theory is right.

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

ICONIQ “State of AI” - AI Costs + AI Gross Margins

ICONIQ (h/t Vivian Guo + team) released their latest July 2026 edition of the “State of AI” report (47 slides, ungated).

Their granular AI cost structure chart evolution from Beta → General Availability (GA) → Scaling once again leads to this conclusion:

Key message: There seem to be too few fixed costs to generate software economics.

Looking strictly at customer-facing AI product costs, the 14% projected increase in AI Gross Margins (2025: 45% → 2027: 59%) seems questionable.

For our core software readers, the 2027e 60% gross margins for Application Layer Products seem encouraging.

However, the change in the unit economics survey suggests gross margin improvement is supportable: 66% of respondents improved their cost per query by 10%+, with 19% at 30%+.

Key unit economics drivers: #1 reducing inference costs, #2 implementing routing strategies, #3 revenue growth through cost leverage, #4 switching to lower-cost/open-source models. In our view, these are believable, “blocking and tackling” drivers.

ICONIQ “State of AI” - AI Revenue Mix

The revenue mix data for non-AI-native companies was striking:

• 2025A: 32% AI as % of total revenue

• 2026E: 42%

• 2027E: 53%

We’ve tracked public company disclosures of their AI metrics. Through early 2025, creating this summary was a heavy lift - actual, meaningful disclosure was limited.

AI disclosure transparency has improved dramatically. Granted, what is disclosed varies widely - like new deal attachment rates or stats on user actions/adoption… but no-to-limited clean AI financials.

Working with public company disclosures through June 2026 for 20 (at least partially) AI transparent companies, these are the only “clean” apples-to-apples AI mix ratios relative to ICONIQ:

For the “Is Software Dead?” debate, you could interpret ICONIQ AI's high mix as a positive sign that smaller private non-AI natives can innovate and quickly monetize AI within their software platforms. Or you could focus on public companies’ HSD/LDD AI penetration rates ~3.5 years post-ChatGPT launch as evidence that legacy software will be disrupted.

Other public disclosures - like deal attach rates - outline how/whether/when the ICONIQ-mix levels could be achieved. This is an “apples vs oranges” chart set at the same scale, so use this for ideation only:

Coming to an accurate view is company- and category-specific… requiring a range of modeling sensitivities, which inspired this:

Related Resource: Our Cloud Ratings Quantitative Strategy Analyst Nick Aiken created a dynamic model template - including the impact of cannibalization - to help answer:

• Does AI Revenue Offset The Loss Of Legacy Seats?

• How Fast Are We Actually Converting To An AI-First Business?

• Can A High-Performing AI Business Save Us From Flatlining?

All of which can help you see scenarios (by variables like AI mix, cannibalization rate, legacy + AI growth rates) where/when/whether AI leads to an overall re-acceleration (related chart = Goldman Sachs software sector-wide ARR growth):

While this AI as a % of software spend data set from Cledara, a software spend management platform (h/t Brad van Leeuwen) could help you choose assumptions on both cannibalization and relative growth:

Share SaaSletter

ICONIQ “State of AI” - More Excerpts

These excerpts should speak for themselves. And we will remind you to always go read the full ICONIQ x AI report:

Excellent data for validating AI ROI:

Data on agent risk and human intervention needs:

Model share amongst the ICONIQ sample (valuable read through given combination of growth rates and scale):

Finally, an update of ARR/employee:

May 2026 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through March 2026 - full slides below:

B2B AI Interest Index

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) continues its plateau pattern.

Bellwether Microsoft Copilot continues to stay at a higher plateau. Other “Blue Chips” (i.e., Hubspot, ServiceNow) remain range-bound.

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel-focused, forward-looking SaaS Demand Index) showed some moderation, but remained on a still-positive overall trendline.

Share SaaSletter

Curated Content

“Ways to think about token pricing” from Benedict Evans

“2026 Midyear Review – 10 Questions Shaping AI (Part 1)” (30 slides, ungated) from Sapphire Ventures (h/t Kevin Burke)

“Work-Bench AI Snapshot H1 2026” (54 slides, ungated) from Work-Bench (h/t Kelley Mak + Proby Shandilya)

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

“State of AI” Reports - Best Of 1H2026

[NEW COVERAGE] “The state of the AI economy” - from Azeem Azhar + the Exponential View team. Released June 25, their 66 slides comprehensively cover AI while adding real insights + viewpoints.

[NEW] “AI eats the world” - from Benedict Evans (79 slides) with our 7 key excerpts here.

ICONIQ (h/t Vivian Guo and team) published their “State of AI: Bi-Annual Snapshot” report (44 slides, ungated). Their granular cost structure data led us to speculate:

Key message: There seem to be too few fixed costs to generate software economics.

Battery Ventures (h/t Dharmesh Thakker, Danel Dayan + co.) published their “State of AI” report (36 slides here). The report included important Battery frameworks for the impact of AI on business + financial models. For our newsletter audience, especially these pre → post AI GTM metrics:

• AE Quota: $1.0-$1.2m → $1.2m-$1.5m

• Conversion Rates: 20%-30% → 40%-50%

• Attainment: 70% → 80%

OpenAI released their own “The State of Enterprise AI” (24 pages). The strong growth across Healthcare, Manufacturing, and Finance could serve as a counterargument to AI as being overly dependent on the Technology sector adoption.

Mayfield released a report on “Enterprise Agentic AI” (n=266, h/t Gamiel Gran)

Beyond report formats, we will call out these AI articles:

• Jaya Gupta and Ashu Garg of Foundation Capital’s “AI’s trillion-dollar opportunity: Context graphs” may raise some solutions to the poor success rates covered in our “Is AI An Abject Disaster?”

• “The Brute-Force Era of AI (and What Comes After)” - Dave Kellogg

• “The Race to Own the Agentic Future” - Dave Yuan from Tidemark

• "How AI Search is Rewiring B2B Software Buying" - from G2 / Tim Sanders + co (n=1,076 B2B decision makers)

• Stratechery on “Agents Over Bubbles”

Beyond AI “State of _” Reports - Best Of 1H2026

Reports (a bit) less explicitly AI-oriented included:

[NEW COVERAGE] “Public Markets Update - May 2026” - 32 slides from Coatue

Redpoint (h/t Logan Bartlett, Adil Bhatia, and Lydia Day) released their annual “Market Update” (69 slides, our 13 key slide excerpts).

Avenir (h/t Jared Sleeper + Hannah Bao + co) published their excellent “Future of SaaS”- our excerpts here

a16z (h/t David George + co) released their “State of the Markets” - our excerpts here

Share SaaSletter

Benchmark Reports - Best Of 1H 2026

[NEW COVERAGE] Ray Rike + Benchmarkit released their flagship benchmark report (n=342). Our overall take: “fits with "maturing growth industry" theme, with signs of greater efficiency... but concerning retention trends for "legacy" vendors in an AI-world” - and slide excerpts here.

Related: Emergence Capital also released their “Beyond Benchmarks” report that builds upon their work with Benchmarkit

A staple of our reading and coverage: Fullcast x Pavilion 2026 GTM Benchmarks (n = $78 billion of sales opportunities), which warranted a newsletter edition AND a podcast with Guy Rubin of Fullcast:

FULL EPISODE: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

More GTM resources:

Scale Venture Partners (h/t Mark Gustaferro + Craig Rosenberg), in partnership with Ray Rike and Benchmarkit, released their GTM Benchmark interactive database.

We further analyzed their data to examine points of separation, meaning where the “best of the best” differentiate themselves from the median company. More in this edition, including a 7-page downloadable GTM benchmarking resource:

Share SaaSletter

We covered ICONIQ’s "2026 State of GTM" in this newsletter edition - TL,DR: largely positive trends:

Zylo (h/t Ben Pippenger + co) released their 2026 SaaS Management Index, covering $75b+ of SaaS spend (= massive read through on industry):

Our 1H 2026 Cloud Ratings Podcasts

Beyond our 2026 GTM benchmark episode with Guy Rubin, we were privileged to host:

“The AI Answer” - Tim Sanders - G2

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

“Where Vertical AI Wins” - Patrick McGovern - Bowery Capital

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

“Engineering Management in the Age of AI” - Karim Fanous

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Episode Topics:

• The 2026 GTM benchmark report and what $75B in pipeline reveals about sales efficiency today.

• Why sales efficiency dropped by 28% and what’s driving the decline.

• The growing gap between top-performing sellers and the rest of the sales team.

• Why AI-generated volume does not automatically create better revenue outcomes.

• How companies using AI for rigor, qualification, and process are increasing revenue per seller.

• Why territory balance matters and how overloaded sellers lose depth, relationships, and win rates.

• Moving beyond round-robin lead distribution toward smarter routing based on seller expertise.

• Why forecasting based only on deal stage is no longer enough.

• How CRM, email, calendar, call recordings, and relationship data can create a more accurate revenue picture.

• The role of Revenue Insights as a Service for private equity firms and portfolio companies.

• Why generic BDR/SDR pipeline often creates noise instead of quality opportunities.

• How AE-led outbound, partnerships, communities, and customer introductions are becoming more efficient growth channels.

• Why ICP definition should be dynamic, data-backed, and based on real buying signals.

• How leadership teams can use simple, visual revenue insights to align around better GTM decisions.

• Why relationships still drive revenue, even in an AI-enabled sales environment.

Resources:

https://www.fullcast.com/ebooks/2026-benchmarks-report-the-state-of-gtm-in-2026/

https://www.fullcast.com/content/guy-rubin-inside-the-go-to-market-benchmark-report/

https://www.joinpavilion.com/resource/2025-gtm-benchmarks-ebsta-pavilion 

About Cloud Radio by Cloud Ratings:

Made for "full stack" cloud + SaaS operators, Cloud Radio is a SaaS podcast covers all aspects of the business of software.

Our host - Matt Harney - is active on Twitter @saasletter and writes "SaaSletter", an investing-oriented SaaS newsletter on Substack.

https://cloudradio.cloudratings.com/

https://twitter.com/SaaSletter

https://www.saasletter.com/

About Cloud Ratings:

Cloud Ratings is a customer outcomes-focused, data-driven software research analyst firm. We exist to allow organizations to make more confident, lower-risk software purchasing decisions.

Our Cloud Ratings Category Reports combine user reviews with verified vendor data to impartially identify leading software products. Recent report examples include: Board Management Software, Account-Based Marketing Software, Cap Table + Equity Management Software, Lead Routing Software, and Sales Training + Onboarding Software.

Built upon investigative customer interviews, our True ROI Reports quantify and provide 3rd party validation of a software product’s business value.

https://cloudratings.com/

“The AI Answer” - Tim Sanders - G2

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

UPDATE: Can Superior NRR Offset Weak AI Gross Margins?

Amazing timing on this disclosure vs last Friday's newsletter:

The ServiceNow Investor Analyst Day from Monday included the above *AI* ACV Curve -> 4.5x implies 146% NRR.

Relative to *standard* SaaS LTVs, this implies AI LTV multiples that are:

• at 40% AI GM = 1.7x v SaaS

• at 50% AI GM = 2.2x

• at 60% AI GM = 2.6x

• at 70% AI GM = 3.0x

All of which is reflected in our modified “efficient frontier” for AI apps:

The key message: last week, we theorized about potential AI NRR curves, working from Tropic transaction data that is inherently short-term and less oriented toward enterprise orgs. With the caveat that this is an illustrative exhibit, this 146% implied AI NRR was put forward by a very sophisticated public company with a strong history of cohort unit economic disclosures.

“State of Martech in 2026”

Scott Brinker and Frans Riemersma released their flagship, ever-popular “State of Martech in 2026” (125 pages).

To add value to their excellent AI adoption impulse coverage by category, we summarized the findings:

Key messages:

• Preference For SaaS (44%) Over AI-Native (26%)

• Effectively, “no one knows” and “we are still early” → “Nascent” has the greatest representation across B2B + B2C combined.

• B2B is very “Buy AND Build” oriented (37%, #1 rank) → so a very strong demand sign

• B2C is much more mixed between Buy and Build options, with “Nascent” the dominant view (51%, #1 rank)

Scott + Frans’ report includes classics like the total martech app count (a very rare plateau… albeit at 15,505!) and new data on what is falling out of the stack:

UPDATE: Inference Is Eating the World

Last week, we covered “Software Is Eating the World (But Actually This Time)” from Siddharth Ramakrishnan of Scale Venture Partners. With our key emphasis on how inference demands go exponential with complexity:

This post from Charlie Zvibleman of AlphaSense, especially the “ auto Think Longer” note, gives a real-world example of the vertical token consumption ramp (up ~230% in 5 months for AlphaSense based on eyeballing the graph)

Dave Kellogg just published a note - “The Brute-Force Era of AI (and What Comes After)” - closely related to this inference explosion.

We hope to cover Dave’s “Brute-Force Era” in the future → in the interim, here are excerpts to get you to read the full post:

Constraint as a Driver of Efficiency

It is not coincidental that much of this efficiency work is emerging from environments where compute is constrained. When you cannot simply add more GPUs, you are forced to extract more value from the ones you have.

That dynamic has a long history in engineering. Systems built under constraint tend to be more efficient, more elegant, and more disciplined. Anyone who has written low-level code on constrained hardware has seen this firsthand. When you only have kilobytes of memory, you learn to think differently about how software is structured.

and:

We have seen this pattern repeatedly in other domains. Early cloud infrastructure was heavily overprovisioned before cost optimization became a discipline. Early web applications prioritized functionality over performance before latency and efficiency became central concerns. Early software systems accumulated features before being refactored into more coherent architectures.

Share SaaSletter

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Episode Topics:

• AI’s role in reshaping the B2B software buying journey, from early research to shortlist creation.

• The rise of “one-shotting” and how buyers now compress days of vendor research into a single AI prompt.

• How G2 reviews, category rankings, and high-authority third-party sources influence AI recommendations.

• Why enterprise buyers are moving faster toward AI-assisted buying as teams get leaner and workloads increase.

• The new importance of “winning the answer” in ChatGPT, Gemini, and deep research tools.

• Mentions vs. citations vs. recommendations: which ones actually influence software buying decisions?

• Deep research is the new buying battleground, where the final recommended vendors matter more than general visibility.

• Why AI-driven traffic may convert better than traditional search traffic.

• The content shifts from polished corporate messaging to fresh reviews, proof points, Q&A content, and human-answer-shaped thought leadership.

• YouTube transcripts, podcasts, LinkedIn, reviews, and niche publications are emerging sources of AI visibility.

• The difference between machine learning, bots, and agents.

• The executive “30% rule” for AI literacy and why leaders should reframe business problems as prediction problems.

• Why companies need to close the trust gap between AI reliability and employee adoption.

• How AI changes hiring and promotion by increasing the value of judgment, curiosity, feedback, and “taste.”

• Inside G2’s own AI-first experiments, including generative summaries, dynamic FAQs, and learning how to make complex product pages easier for machines to understand.

Resources:

https://learn.g2.com/g2-2026-ai-search-insight-report 

https://learn.g2.com/g2-2025-ai-agent-insight-report 

https://www.tryprofound.com/blog/profound-g2-partnership 

https://university.tryprofound.com/

https://6sense.com/science-of-b2b/buyer-experience-report-2025/

https://promptwatch.com/ 

https://spotlightar.com/

https://firstpagesage.com/seo-blog/generative-engine-optimization-geo-explanation/ 

https://graphite.io/five-percent/ai-is-much-bigger-than-you-think

https://www.similarweb.com/blog/marketing/geo/build-prompts-from-keywords/ 

About Cloud Radio by Cloud Ratings:

Made for "full stack" cloud + SaaS operators, Cloud Radio is a SaaS podcast covers all aspects of the business of software.

Our host - Matt Harney - is active on Twitter @saasletter and writes "SaaSletter", an investing-oriented SaaS newsletter on Substack.

https://cloudradio.cloudratings.com/

https://twitter.com/SaaSletter

https://www.saasletter.com/

About Cloud Ratings:

Cloud Ratings is a customer outcomes-focused, data-driven software research analyst firm. We exist to allow organizations to make more confident, lower-risk software purchasing decisions.

Our Cloud Ratings Category Reports combine user reviews with verified vendor data to impartially identify leading software products. Recent report examples include: Board Management Software, Account-Based Marketing Software, Cap Table + Equity Management Software, Lead Routing Software, and Sales Training + Onboarding Software.

Built upon investigative customer interviews, our True ROI Reports quantify and provide 3rd party validation of a software product’s business value.

https://cloudratings.com/

Fullcast 2026 GTM Benchmarks

While last week we covered ICONIQ’s “2026 State of GTM“ positive trends in areas like win rates and sales cycles, the latest GTM benchmarks from Fullcast paint a more sobering picture.

We’ve covered the Fullcast (formerly Ebsta pre-acquisition by Fullcast) / Pavilion benchmarks each edition for years now due to a) granularity and b) sample size (this year = $78 billion of sales opportunities). Key highlights:

Sales efficiency and win rates continue to be poor, in absolute and directional terms:

AI is accelerating seller ramps:

Poor pipeline ICP mix very likely contributes to poor absolute win rate levels

Tactical call out - outbound BDR efforts 80% below average efficiency → in a noisy AI world, is this SDR cold outreach playbook dead + a brand net negative?

Fullcast Revenue Insights As A Service

We are helping Fullcast + Guy Rubin connect with software private equity investment firms to grow their Revenue Insights as a Service (RIaaS) offering:

• quarterly deep dives into your portco’s sales machine,

• using a mix of AI/software (linked into CRMs, email, and call recordings)

• with a professional services element that yields this type of report (attached)

Full Fullcast RIaaS overview deck here + book a demo

Can Superior NRR Offset Weak AI Gross Margins?

Inspired by this AI vs SaaS NRR data via Tropic (a software spend management platform covering ~$20 billion of spend) in Lazard’s “AI Impact on Technology M&A” deck (60 slides)…

The eye-popping AI NRR of 154% inspired a theoretical cross-check: Can Superior NRR Offset Weak AI Gross Margins?

The key assumptions and rationale: SaaS unit economics are well-established, with a reasonable baseline of roughly 105% net revenue retention (NRR) and 70% gross margin.

We applied a 7-year expected customer life (84% gross retention, so below core Mid-Market / Enterprise GRR levels… while recognizing 7 years is an eternity in terms of technical disruption) and a 10% discount rate.

Under these assumptions, a $10k ACV SaaS customer has an NPV of $39k.

This NRR vs Gross Margin sensitivity table helps frame the “efficient frontier” for AI apps:

Translating a bit, AI apps can have SaaS customer unit economics at these thresholds:

• At 50% AI gross margin, need NRR of ~118%

• At 40% AI gross margin, need NRR of ~130%

• At 30% AI gross margin, need NRR of ~141%

• At 30% AI gross margin, need NRR of ~150%

Of course, 130%-%150% NRRs over 7 years translate into exiting ACVs of $48k-$113k off of a $10k starting base! Meaning you would be underwriting towards ServiceNow’s legendary cohort growth (via latest ServiceNow Investor Presentation):

Inference Is Eating the World

Siddharth Ramakrishnan of Scale Venture Partners (h/t Rory O’Driscoll) published a thoughtful “Software Is Eating the World (But Actually This Time)” where he does a good job outlining the less visible complexity and increasing inference needs in AI workflows (our emphasis = bold):

Take a voice support agent handling something simple but real, like rescheduling a medical appointment. To the customer, it feels like one conversation. Under the hood, it is a small autonomous system running continuously. As the caller speaks, a speech recognition model transcribes audio in real time. An orchestration model then reasons over the transcript, pulls the patient record, checks scheduling constraints, looks up provider availability, decides what to ask next, and calls the relevant tools. Once it has enough information, it synthesizes the result into a response, and a text-to-speech model turns that back into natural audio. In parallel, other models may be monitoring sentiment, checking compliance, or deciding whether the call should be escalated.

The system is doing all the work itself: listening, retrieving, deciding, tool-calling, verifying, and responding in a loop. An 8 minute call might contain only ~3k tokens of raw transcript, but the orchestration layer can easily consume ~40k tokens once you account for repeated reasoning over the growing conversation, retrieved context, and tool outputs, on top of continuous ASR and TTS inference running for the duration of the call. “One AI phone call” is really a multi-model inference stack operating continuously.

To help call out to the human eye what this means for token + inference needs, we translated the article charts from log to linear scale:

As always, go read the full post which includes deeper insights like:

Share SaaSletter

Avenir on Anthropic vs Microsoft

Jared Sleeper from Avenir published “How to Distribution-Mog Microsoft: Anthropic’s Path to Building the White Collar AI Platform” (16 slides). Our excerpts focus on their primary survey work, showing an interest in a) paying a premium for better AI and b) customers wanting M&A (as our research has shown, software buyers prefer platforms over point solutions, even when “best of breed”):

This 4-in-1 excerpt gives a sense of the strategic angles in the deck:

While focused on Microsoft, this thesis left a:

• stronger appreciation for the Google moat (specifically, Workspace + Gemini + Google Cloud, both in terms of compute + custom enterprise applications to the extent enterprises are also on platforms like BigQuery) in an AI world where “mogging” Microsoft has become a point of debate.

• The customer demand/interest in comms M&A is intriguing from a first principles perspective: Why would Anthropic owning Zoom or Box/Dropbox really change things? (Monday/Asana makes a bit more sense). What does this mean for feature gaps in AI <> SaaS integrations? Just a function of not wanting to do yet another integration (again, buyers do perform platforms)?

Curated Content

A status check on AI ramp at incumbents: Adobe disclosures from their Adobe Summit 2026 (n=1 here... but big!):

• $400m "AI-first ARR" = 1.5% of total ARR

• $1b "GenStudio ARR" = 3.8% of total ARR

• Generative AI credits growing +45% quarter over quarter = 377% annualized (yes, that high given exponential in QoQ at 45%!)

Lazard’s “AI Impact on Technology M&A” deck (60 slides) mentioned above is worth a rehighlight - strong aggregation of AI frameworks + studies.

This competitive framework from Goldman Sachs was interesting. TBD on the weighting of consumer familiarity - equal weighting would require a substantial increase in the “consumerization of IT” within enterprises, more than realized in past cycles… that said, AI is truly an epoch shift.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

“Where Vertical AI Wins” - Patrick McGovern - Bowery Capital

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

“Engineering Management in the Age of AI” - Karim Fanous

SPOTIFY | APPLE | OTHER PODCAST PLATFORMS| VIDEO

Redpoint Ventures - Market Update

Redpoint (h/t Logan Bartlett, Adil Bhatia, and Lydia Day) made a strong contribution to 2026’s “State of _” genre with their Market Update (69 slides, our 13 key slide excerpts). Selected highlights below:

G2 On AI x Software Buyer Search

"How AI Search is Rewiring B2B Software Buying" - from G2 / Tim Sanders + co (n=1,076 B2B decision makers)

#1 - Godard Abel was right on our earlier podcast - AI expands the shortlist! (direct clip linked). Anecdotally, in our own buying AI, both G2’s Chatbot and Gemini have identified vendors we otherwise would not have considered.

#2 - huge volume of insights in the report - like 71%(!) of the B2B software funnel starts with a chat prompt - especially at the CMO level. Huge implications for monitoring, content, attribution, pipeline visibility, and more!

ICONIQ 2026 GTM Benchmark Update

A largely positive "2026 State of GTM" from ICONIQ:

- Consistently improving win rates, an important sign relative to consistent win rate declines in 2025.

- Those challenging closing rates meant forecasts and pipeline coverage were significantly strained in 2025 (see slides from our Cloud Ratings’ “2025 Value Creation Summit”)

- Back to ICONIQ, positive trends in cost/opportunity and sales cycles.

Curated Content

Ciridae (an AI workflow architecture start-up, backed by A16Z and General Catalyst; h/t Felix Stocker & Jai Kondapalli) released an interesting AI transformation explorer, by sector and even by private equity firm. We made a quick recut to visualize the findings:

Sapphire Ventures (h/t Kevin Burke + co) March Market Memo included good coverage of AI agents…

along with this Stratechery “Agents Over Bubbles” take.

Apoorv Agrawal from Altimeter on “The Economics of Generative AI”

March 2026 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through March 2026 - full slides below:

B2B AI Interest Index

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) rebounded and reversed poor winter trends (while noting inherent seasonal lows in December).

Bellwether Microsoft Copilot continues to stay at a higher plateau. Other “Blue Chips” (i.e., Hubspot, ServiceNow) remain range bound.

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel focused, forward-looking SaaS Demand Index) performed strongly in Q1 2026.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Episode Topics:

• How Bowery approaches pre-seed in 2026 and what “first institutional capital” looks like today.

• The shift from “SaaS contracts” to outcome-driven and services-driven models.

• Pat’s vertical AI playbook.

• Case study on Enata and how voice-first AI gives field sellers time back while improving CRM quality and sales insights.

• Why more startups may compete directly for industry spend rather than selling tools to incumbents.

• What Pat wants in founders.

• The new definition of technical in the AI era.

• How to run customer pilots well, including shorter timelines, mid-point check-ins, proof-of-value gates, and prioritizing referenceable logos.

Resources:

https://x.com/pw_mcgovern?lang=en
https://www.linkedin.com/in/pwmcgovern
https://bowerycap.com/
https://capitalefficient.substack.com/

About Cloud Returns by Cloud Ratings:

Cloud Returns covers ALL types of software investing, whether seed, venture capital, growth equity, private equity, debt, and the public markets.

Our host - Matt Harney - is active on Twitter @saasletter and writes "SaaSletter", an investing-oriented SaaS newsletter on Substack.

https://cloudradio.cloudratings.com/

https://twitter.com/SaaSletter

https://www.saasletter.com/

About Cloud Ratings:

Cloud Ratings is a customer outcomes-focused, data-driven software research analyst firm. We exist to allow organizations to make more confident, lower-risk software purchasing decisions.

Our Cloud Ratings Category Reports combine user reviews with verified vendor data to impartially identify leading software products. Recent report examples include: Subscription Management Software, Session Replay Software, Sales Compensation Software, and Digital Sales Room Software

Built upon investigative customer interviews, our True ROI Reports quantify and provide 3rd party validation of a software product’s business value.

https://cloudratings.com/

Episode Topics:

• Why VPE is “people + product + processes”  and how breakdowns show up as business problems (growth, quality, churn, GTM).

• How “buggy processes” reveal themselves at the exec level through growth stalls, quality pain, churn, and GTM friction?

• Executive interview questions that reveal whether an engineering leader can connect engineering to business reality.

• The common scaling fallacy: doubling engineers does not double output.

• “AI made us faster… not necessarily better”: how AI boosts strong engineers, but used carelessly amplifies technical debt.

• Why human review won’t scale in an agent-driven world and why software will need machine verification to move at “GPU speed.”?

• Engineering shifts toward architecture, integration, trade-offs, experiments, and validation.

• The “10x engineer” debate in the AI era.

• AI security risks: non-deterministic systems + autonomous actions require controls beyond RBAC.

• Inside StrongDM’s approach on agents treated like coworkers with identities, visibility, and feedback that improves behavior.

Resources:
https://www.cummulative.io/ 

https://www.linkedin.com/in/karimfan
https://www.strongdm.com/
https://www.cummulative.io/p/ai-agents-are-here-security-isnt 

https://www.cummulative.io/p/preparing-for-a-world-where-humans 

https://www.cummulative.io/p/how-do-you-interview-a-vp-of-engineering 

About Cloud Radio by Cloud Ratings:

Made for "full stack" cloud + SaaS operators, Cloud Radio is a SaaS podcast covers all aspects of the business of software.

Our host - Matt Harney - is active on Twitter @saasletter and writes "SaaSletter", an investing-oriented SaaS newsletter on Substack.

https://cloudradio.cloudratings.com/

https://twitter.com/SaaSletter

https://www.saasletter.com/

About Cloud Ratings:

Cloud Ratings is a customer outcomes-focused, data-driven software research analyst firm. We exist to allow organizations to make more confident, lower-risk software purchasing decisions.

Our Cloud Ratings Category Reports combine user reviews with verified vendor data to impartially identify leading software products. Recent report examples include: Board Management Software, Account-Based Marketing Software, Cap Table + Equity Management Software, Lead Routing Software, and Sales Training + Onboarding Software.

Built upon investigative customer interviews, our True ROI Reports quantify and provide 3rd party validation of a software product’s business value.

https://cloudratings.com/

ICONIQ (h/t Vivian Guo and team) recently released its “State of AI: Bi-Annual Snapshot” report (44 slides, ungated).

Key call-outs: while AI product gross margins are projected to increase by ~700 bps (‘25: 45% → ‘26e: 52%):

… the fabulously granular cost structure breakdown makes those margin increases look optimistic. “Talent” - a largely fixed cost - is only 26%-28% of the cost structure for the GA to Scaling stages. All other costs are highly variable, such as inference and cloud costs. Even for model training, there is little fixed cost leverage since model training is an ongoing process. Key message: There seem to be too few fixed costs to generate software economics.

Notably, larger organizations (>$500m revenue) are the leading deployers of customer-facing agentic AI.

The increasing AI deployment and returns to scale are consistent with our 2023 formula below (with emphasis on the volume variable) and Morgan Stanley findings that large enterprises are achieving the highest ROIs:

AI ROI = (Business Process Value x Volume) / AI Costs

Back to ICONIQ, flagging this data + commentary as a headwind: “Internal AI is increasingly funded through R&D budgets, with fewer companies relying on net new budget allocation.” This could have implications for both AI vendors (i.e., hitting budgetary guardrails + enterprise reality checks) and non-AI software vendors (i.e., AI truly competing in the same IT dollar pool rather than a separate “innovation” budget).

If you know, you know (“IYKYK”), any and all AI model market share charts attract a lot of “attention” amongst VCs:

And yet another data point for the “buy versus build” component of the “Software Is Dead” daily debate:

December 2025 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through December 2025 - full slides below:

B2B AI Interest Index

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) registered a notable decline. Note: holiday seasonality will be at play throughout this month’s report. Beyond seasonality, as AI-native vendor brands strengthen, generalized product searches might be losing share to brand-targeted search.

Bellwether Microsoft Copilot continues to stay at a higher plateau. Other “Blue Chips” (i.e., Hubspot, ServiceNow) had a weak December (down 4% YoY, so not just seasonality here).

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel focused, forward-looking SaaS Demand Index) registered a seasonal MoM dip, but interest in the Incumbent basket is up 119% YoY.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Scale Venture Partners (h/t Mark Gustaferro + Craig Rosenberg), in partnership with Ray Rike and Benchmarkit, recently debuted their GTM Benchmark interactive database.

We further analyzed their data to examine points of separation, meaning where the “best of the best” differentiate themselves from the median company.

Our downloadable 7-page document includes more on GTM motion (i.e., PLG vs Inside Sales vs Outside Sales vs Channel) and volatility:

DOWNLOAD HERE

A caveat: Some of these variances are structural (i.e., an enterprise ERP vendor will have a significantly longer sales cycle, regardless of execution; or a PLG vs Channel vs Sales-Led motion will have many different baseline metrics). Taking into account that dynamic, our key call-outs:

SALES REP TURNOVER: 773% Delta

This is the single most glaring “off the chart” stat in the entire dataset.

For the top 10% of companies, annual sales rep turnover is significantly lower than the mean, with a 773% difference between the best and the average.

The elite performers are playing a different game: Retention. You cannot build compounding momentum when you are constantly replacing your GTM engine parts. The “Best of the Best” have essentially solved the turnover problem, keeping institutional knowledge in-house.

CAC PAYBACK PERIOD: 192% Delta

The separation here is stark: the 90th percentile outperforms the average by 192%.

While the average company struggles with payback periods that drag on, top-tier firms are recovering their acquisition costs nearly 2x faster. Which allows them to reinvest in growth cycles significantly faster than their peers.

QUOTA + RAMP: >100% Delta

It’s all about having ramped productive reps hitting their number.

Exceeding Quota: The separation for “% of Reps Exceeding Quota” in the most recent period is 121%.

Ramp Speed: This is fueled by a 107% advantage in “Time for Rep to Fully Ramp”.

This suggests the “Best” aren’t just hiring better; they also have superior enablement, Finance → GTM hiring coordination, and territory-planning infrastructure.

Related Research Call Out:

Declining quota attainment and slower ramps have been relentless long-term trends in a “maturing growth industry” - see our 2024 “State of Software” (Full slides, 1st presented at Bowery Capital’s 2024 AGM) - perhaps making top decile outperformance all the more impressive.

INBOUND HANDRAISERS: 129% Delta

The “Best of the Best” generate 129% more pipeline from Inbound Hand Raisers than the average.

Put differently, the top decile has a “Brand Dividend” with enough market pull that prospects come to them.

Reminder - we have the Scale GTM Benchmarks available in a downloadable PDF here:

Scale GTM Benchmark - 7 Page Analysis

Curated Content + Upcoming Coverage

Dave Kellogg released his excellently comprehensive 2026 Predictions →

His callout of Jaya Gupta and Ashu Garg of Foundation Capital’s “AI’s trillion-dollar opportunity: Context graphs” may raise some solutions to the poor success rates covered in our “Is AI An Abject Disaster?”

Flagship reports have been in full bloom this January → we hope to cover these individually… but might as well give you the underlying links too:

Avenir (h/t Jared Sleeper + Hannah Bao + co) published their excellent “Future of SaaS”- our excerpts here

a16z (h/t David George + co) released their “State of the Markets” - our excerpts here

Mayfield released a report on “Enterprise Agentic AI” (n=266, h/t Gamiel Gran)

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Battery Ventures (h/t Dharmesh Thakker, Danel Dayan + co recently released their “State of AI” report (36 slides here).

Beyond strong core “State of _” type slides like this example:

… the report included important Battery frameworks for the impact of AI on business + financial models. For our newsletter audience, especially these pre → post AI GTM metrics:

• AE Quota: $1.0-$1.2m → $1.2m-$1.5m

• Conversion Rates: 20%-30% → 40%-50%

• Attainment: 70% → 80%

November 2025 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through November 2025 - full slides below:

B2B AI Interest Index Slides

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) registered a notable decline. Note: holiday seasonality might be at play throughout this month’s report.

Bellwether Microsoft Copilot continued its broad acceleration. Other “Blue Chips” (i.e., Hubspot, ServiceNow) registered a poor November.

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel focused, forward-looking SaaS Demand Index) registered a poor November, especially relative to a tough October comp. That said, the n=340 Incumbent trendline remains overall positive for 2H 2026.

Bonus Coverage: OpenAI “The State of Enterprise AI”

OpenAI released their own “The State of Enterprise AI” (24 pages). These excerpts should speak for themselves + pique your interest in the full report.

The strong growth across Healthcare, Manufacturing, and Finance could serve as a counterargument to AI as being overly dependent on Technology sector adoption.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Investor “State of __” Reports - Best Of 2025

[NEW] Meritech: Recently released their “2025 Market Update” (Slides | Video)

[NEW] Battery Ventures: “State of Enterprise Tech Spending December 2025” - covers 100 CXOs with $44 billion of tech spend = great coverage of “Enterprise Plus+” - h/t Evan Witte, Scott Goering, Nick Elsner)

[NEW] Menlo Ventures: “State of Generative AI In The Enterprise” - h/t Tim Tully, Joff Redfern, Deedy Das, Derek Xiao

Coatue: “Public Markets Update” (SLIDES | VIDEO) from October 2025 - with our excerpts + takes here

[NEW Curation] Scale Venture Partners: “State of GTM AI” - h/t Craig Rosenberg + co

Related Resource: Subscribe to Scale’s Quarterly Flash Reports - great real-time read of the market - h/t Dale Chang + co

[NEW Curation] Norwest: “B2B Sales & Marketing Benchmark Report” (102 slides, n=177) - h/t Renée Cohen

Redpoint - “2025 Market Update” - our takes + excerpts here - h/t Logan Bartlett, Adil Bhatia, Lydia Day

ICONIQ’s “State of Software 2025” (73 slides, h/t Christine Edmonds + co) - our full coverage here.

Bessemer Venture Partners - “State of AI 2025” (note the change from the historic Cloud → AI)

Felicis Ventures: “How to Size the Market for Vertical AI” - h/t Sundeep Peechu + Eric Flaningam

… plus my “State of Software + AI” presented at Bowery Capital’s Annual Meeting:

Benchmark Reports - Best Of 2025

Tidemark: 2025 Vertical + SMB SaaS Benchmark report (60 slides, n=200+). This is one of the most granular AND thesis-driven reports in the industry. Our full coverage here. h/t Dave Yuan, Andrew Walsh + co.

High Alpha: “2025 SaaS Benchmarks” (n=800+) with our takes here. h/t Kristian Andersen, Scott Dorsey, Mollie Kuramoto + co.

Benchmarkit + Emergence Capital: Cloud Ratings was honored to support Ray Rike + co’s benchmark report (n=563) as a partner, with our full coverage + update of our historical time series here:

Sapphire Ventures + KeyBanc: Here is the 2025 Benchmark Report (email required) from the longest-running series in the industry. Our coverage here.

Industry + Operator Reports - Best Of 2025

G2: Tim Sanders + team’s “A Leap of Trust: AI Agents Are Winning Hearts and Wallets” (57 slides; n = 1,000+ B2B software buyers). Key themes:

• More evidence of solid AI ROI

• Surprising signs of a “risk on” mentality from buyers despite meaningful incident rates and higher-than-expected human oversight. Those incident rates align with the poor AI success rates we covered in “Is AI An Abject Disaster?”

Related reading: G2’s “2025 Buyer Behavior Report” with our report takes here:

Fullcast: Guy Rubin + team’s “2025 H1 GTM Benchmarks” (covering $43 billion of pipeline) and their full year 2024 GTM benchmark report, which I deep dived with Guy Rubin on our “Cloud Radio” podcast (Audio | Video).

Zylo: Yet another favorite that was worthy of its own podcast episode: Ben Pippenger + team’s “2025 SaaS Management Index” covers $40+ billion of software spend across 40 million licenses, with uniquely strong visibility into the “Enterprise Plus” segment (average SaaS spend of $284 million per org!). Our report coverage focused on shorter contract durations as a symptom of AI-driven uncertainty.

[New] 6Sense: “2025 B2B Buyer Experience Report” - authoritative + scientific from Kerry Cunningham + co

Pavilion: Their excellent member surveys enabled our own research into n=227 forecast outcomes: “Pipeline & Deal Flow” is the most cited driver of misses, inspiring our “Marketing Takes The Blame” note:

~Indie Writers - Best Of 2025

Dave Kellogg always (deservedly) makes this list → some of his best work highlights of 2025 include “SLIDES - Selling Work vs Selling Software” + “How To Navigate the Pipeline Crisis” + “The Era of Haves and Have-Nots”

Along with ~anonymous writers deserving of shout-outs - + + …

And Patrick McGovern of Bowery Capital for taking on the Sisyphean task of creating a new VC Substack in the year of 2025 → help him by subscribing to the “” newsletter.

Recurring Market Updates - Best Of 2025

Recurring updates to subscribe to:

RepVue Cloud Sales Index - h/t Adam Little + co

ChartMogul Monthly Benchmarks - h/t Sophia Faustino + Nick Franklin

Scale Venture Partners’ Flash Report - h/t Dale Chang

Share

Our Own Best Content Of 2025

Our 2025 Cloud Ratings Podcasts

Godard Abel of G2: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

Joey Brookhart of AlphaRepo: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

Guy Rubin of Ebsta / Fullcast: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

Ben Pippenger of Zylo: AUDIO | VIDEO | All Other Platforms

Our Value Creation Summit

With the theme “Better Data = Better Forecasts + Returns,” we were honored to have Guy Rubin, Ryan Walsh, Sean Saint, Jaap Westrik, and Kai Mitsushio serve as expert speakers:

Event Replay Available On YouTube

With our written event recap here, my presentation slides here, and my session here:

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Since I’ve been regularly consuming examples of AI productivity gains - whether SaaStr or S&P500 public companies, more on both below - I’ve assumed any headline numbers from any AI benchmarking charts are similarly positive: “A 70.0 on a legal benchmark report surely means AI lawyers are 70% as good as humans, right? That seems pretty good.”

Using Scale AI’s SEAL LLM Leaderboards, we translated 3 different studies into approximate comparisons to their human equivalent by domain:

Unsurprisingly, coding scores well. With Finance and Legal OK(?) since we are early(?) in the AI cycle. The drop-off from All Questions (including Easy) to Hard Questions is meaningful across all 3.

A fascinating new study from the Center for AI Safety and Scale AI caught my eye: the Remote Labor Index, which benchmarks AI against the performance of experienced freelancers on UpWork across 240 projects.

The freelancer vs AI projects:

• Spanned from video, CAD, graphic design, game development, audio, architecture, and more.

• Mean Human Completion Time: 28.9 hours (Median: 11.5 hours)

• Average Project Value: $632.60 (Median: $200)

• Implying hourly rates of $17-$22/hr

The project success rates were surprisingly low: 0.8%-2.5%! Particularly in light of the Elo scores (Elo Wikipedia entry) as high as 509, which I, as a non-chess player, assumed implied a ~51% capability versus humans.

Relative to what’s “priced in” and the sheer amount of capital into AI, these success rates surely are an abject disaster?

Why are the headline success numbers so low? The Remote Labor Index measures Full Automation as: “The percentage of projects where an AI’s deliverable is judged at least as good as the human standard. This is the primary success metric, defined as whether a ‘reasonable client’ would accept the work.”

The critical failure driver: any 10+ hour project requires multiple steps. Therefore, any performance gap versus humans compounds. Especially if unsupervised.

Taking this (maybe blindingly obvious) insight back to the Scale leaderboards for other business AI domains, seemingly OK headline capabilities collapse once a multiple-step assumption is applied. With 5 and 10 step calculations applied here:

At 10 steps, AI Coding is 85% worse than human developers. While Finance AI and Legal AI yield 99.8% worse results once 10 steps are involved!

Caveat: This is a brute and mechanical calculation. Nor does it have an ounce of the sophistication of research benchmarks (the RLI paper involved 48 authors!).

What Does This Success Rate Math Mean?

My immediate reaction: these near-zero AI success rates are damning to the “Software Is Dead” / “End of Software” (Chris Paik 2024 Google Doc) take.

At a more philosophical level, the success results reminded me of the “AI gives you infinite interns” framing (see Benedict Evans’ November 2025 “AI Eats The World”).

Teasing out the analogy: can interns truly help with easy work (like moving logos around or taking meeting notes)? For sure. How often would an intern complete a 10-hour project (for an experienced professional) at an experienced professional’s quality? Very rarely. Taking into account any manager costs (like assigning or reviewing the project), those intern projects destroy value.

Compared to the latest Scale AI benchmarks + math, the intern framework still holds up.

As an aside, the *infinite* aspect is alluring and well captured by Benedict Evans’ 1800s steam-engine analogy. Anything *at all positive* multiplied by *infinity* is wildly attractive and worth trying to get right.

All of which aligns with Buck on Software’s 2024 “Winds of Change” and its key concept of unbounded task TAMs… so long as it is useful software.

Our macroeconomic work at Cloud Ratings on software developer value added (slides) sets a more encouraging precedent for AI: like software developers, AI interns will be better utilized over time, especially in terms of price/quality.

Share SaaSletter

Bad Benchmarks But Real World AI Successes: SaaStr + S&P500

Tracking the progress SaaStr (Amelia Lerutte + Jason Lemkin) has made internally with their implementation of AI agents has raised my expectations of what is possible in the real world… with hard human work to get those AI results.

SaaStr recently hosted a remarkably transparent workshop (SLIDES | VIDEO) on their AI agent learnings. Seriously, go watch it.

A key theme of “constant calibration + daily oversight” certainly aligns with the Scale benchmarks and the “infinite interns” concept.

The growing volume of AI success disclosures from S&P 500 components (via Morgan Stanley, 11/06/25) also serves as a counterpoint to any “AI = abject disaster” conclusions:

Of course, relative to what’s priced in for AI, perhaps 15% of the S&P500 disclosing AI gains is too low? Nor is it broad-based enough, with only mid single-digit disclosure rates from real-world segments like Materials and Health Care (versus a 39% disclosure rate in Tech broadly, 59% for S&P500 Software & Services)?

Of note, the mix of ROI drivers between AI and pre-AI software (per Cloud Ratings’ “Drivers of Software ROI” and “IT ROI” studies) is very consistent.

October 2025 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through October 2025 - full slides below:

B2B AI Interest Index Slides

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) declined sequentially versus a strong September, though above its summer lull.

Bellwether Microsoft Copilot continued its acceleration. Other “Blue Chips” (i.e., Hubspot, ServiceNow) remain range-bound at best and small in absolute terms.

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel focused, forward-looking SaaS Demand Index) have seen a notable 2025 breakout relative to a very range-bound 2024. With October accelerating:

Podcast With Godard Abel Of G2

FULL EPISODE: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

Our recent AI-oriented podcast with Godard Abel (CEO of G2) complements much of this newsletter edition.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

KeyBanc + Sapphire 2025 SaaS Benchmarks

I’ve been reading this series dating back to the Pacific Crest Securities (before their joining KeyBanc) days. We’ve kept a compilation of the historical KeyBanc SaaS metrics back to 2009 through 2024 to create a long-run time series - our full slides here and write-up version here:

Now onto the new 2025 KeyBanc + Sapphire report (email required).

My excerpts + takes will be a bit negative despite some positive trendlines. The summary metrics recap is very green for both 2025e and 2026e. Relative to 2024, 2026e reflects:

• ARR Growth: Reaccelerating from 15% to 25%

• Account Executive Sales Productivity: Up 60%

• Rule of 40: Up 21% (percent points basis)

This all sounds great… but the forecasting track record in software has been poor recently. Enough so we hosted a Value Creation Summit (Event Recap | Videos | Slides) with a key focus on forecasting:

In fact, this very KeyBanc/Sapphire report shows its own history of forecast misses:

In particular, the optimism of these key metrics/drivers stood out:

Even factoring in AI as a sales productivity enabler, this per AE improvement looks aggressive:

As does quota attainment: up 10% (percent point basis; note RepVue Q3’25 Cloud Sales Index shows quota attainment at 43%):

Drivers like this flow through to declining CAC ratios and improving Rule of 40 scores.

There is a lot more in the full report - like AI coverage → so go read the full report.

High Alpha 2025 Benchmarks

High Alpha (h/t Kristian Andersen, Scott Dorsey, Mollie Kuramoto + co) released its 2025 SaaS Benchmarks (64 slides, email required).

Notably, the sample size exceeded 800, making this one of the largest in the software benchmark universe.

This metric recap - including great CACs and strong Rule of 40s even at low ARR - suggests a respondent mix of higher quality than comparable surveys.

This slide captures the wide growth gap between AI winners and the rest (i.e., for $1-$5m ARR, “AI differentiation drives 70% faster growth”).

This employee count chart captures the degree of “getting fit” from the 2021/2022 market peak, particularly in the $20m-$50m and >$50m ARR segments.

This slide makes a compelling case for variable pricing models:

This In-Office vs. Remote comparison will surely attract attention: In-Office organizations grow faster (+11%) while also achieving a higher Rule of 40 (+3 points).

… with a similarly novel Founder Stress Levels vs Growth chart that is worth downloading the report to see.

This slide visualizes the holy grail of low CACs and high retention well:

Bonus: Accel Globalscape 2025

More of a “State of …” than benchmarks, Accel released its Globalscape 2025 report (64 slides). Our call-outs:

Anonymized gross margin data for AI application leaders:

At a simple average gross margin of 25% for AI leaders, this is radically below public SaaS benchmarks (76% per Accel) and below the 60% - 65% AI gross margins shown in Bessemer Venture Partners reports. To be fair, BVP also reports “Supernovas” (a group likely to be very similar to Accel’s top 7) at 25% gross margin.

… all of which gives me the chance to insert my XKCD-inspired importance of gross margins graphic (with more context here):

Back to Accel, they also laid out their version of “AI Bubble Math” here - TL,DR: AI CapEx needs to increase global GDP growth by 150 bps.

Other Recent Benchmark Coverage = Tidemark:

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Tidemark (h/t Dave Yuan, Andrew Walsh + co) recently released their 2025 Vertical + SMB SaaS Benchmark report (60 slides, n=200+).

This is one of the most granular AND thesis-driven reports in the industry, so Cloud Ratings was honored to serve as a report partner.

Tidemark’s Vertical SaaS Knowledge Project (VSKP) Framework frames the report:

The report classifies companies into primary product areas (i.e, Back Office vs FinTech) with our call-outs of particularly key metrics:

Of particular note:

• FinTech Gross Retention of 96% (vs ~90% for other product types): From a *compounding growth* perspective, having only to backfill 4% logo churn is a far, far easier headwind than replacing ~10%+ of logos typically seen in SMB + many vSaaS categories.

• Low CACs: All categories exhibited CACs below $5,000. Regardless of revenue per account, this is excellent in absolute terms.

We modeled1 the Tidemark data into estimated LTV / CAC ratios, including LTV on an 8-year customer life basis, to normalize for extreme gross retention rate scenarios (i.e., FinTech 96% GRR implies 25-year customer lives!). Even when normalized, the LTV/CAC ratios implied from the Tidemark sample are VERY strong (average = 23.5x).

Of note, publicly traded vSaaS also exhibit relatively high LTV/CACs - 9x-20x (for VEEV, TTAN, CCCS) vs 5x-6x for the overall public universe.

Multi-Product Investment Thesis Support

The Tidemark report covered multi-product data in depth - key excerpts here with much more detail in the full report.

Multi-product outperforms on the key growth + net revenue retention metrics (despite substantially lower revenue/account and slightly lower gross retention)

This TAM expansion slide speaks for itself:

RELATED RESEARCH: In part inspired by Tidemark/Lone Pine’s work on multi-product outperformance, we examined “Platforms vs Pure Plays” in partnership with G2 in late 2023:

Key messages:

• Platforms generated substantially equivalent customer satisfaction versus pure play specialist offerings

• G2 Buyer Behavior data shows buyers consistently prefer working with one vendor

… with more in our 32-slide, downloadable “Software Platforms vs Pure Plays” report.

Share SaaSletter

AI x vSaaS/SMB Benchmarks

Tidemark also covered AI penetration in vSaaS/SMB. Key excerpts:

Attach rates are growing explosively (median: 2024: 22% → 2025: 59%):

… with AI-enabled vendors posting moderately higher growth + retention rates:

There is a *lot* more in the full Tidemark report. As always, do your own work + go read the full report.

Vertical Software Investing Podcast With Dave Yuan (2024)

VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

We’re re-highlighting our *2024* vertical software investing podcast with Dave Yuan, Founder of Tidemark.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Having first connected thanks to their impressive benchmark data ($40 billion+ of customer pipeline under management) and hosting Guy Rubin on our podcast 2x, we are helping Ebsta (recently acquired by Fullcast) connect with software private equity investment firms to launch their Revenue Insights as a Service (RIaaS) offering:

• quarterly deep dives into your portco's sales machine,

• using a mix of AI/software (linked into CRMs, email, and call recordings)

• with a professional services element that yields this type of report (attached)

Full Ebsta RIaaS overview deck here + book a demo

LEARN MORE

Fullcast 2025 GTM Benchmarks - ICP Mix

Fullcast and Ebsta recently released their 2025 GTM Benchmarks, covering $43 billion of pipeline across 440,000 opportunities.

My #1 callout: only 23% of pipeline is high-fit ICP (Ideal Customer Profile). ~45% is medium-fit ICP, and ~30% is low-fit ICP.

This mix has improved somewhat: in 2023, ~40% of pipeline was low-fit ICP, with this trendline fitting the narrative of GTM “getting fit” after the 2022 peak.

What does the ICP fit data really mean?

Poor ICP Mix = Opportunity: The *Marketing* aspect of improved ICP targeting + eliminating GTM waste is obvious. Taking low-fit ICP from 30% to 10%, should theoretically yield 20% in areas like ad spend savings. This is easier said than done: for example, there are limited ways to gate low-fit ad clickers on Google Ads.

The less obvious (at least to non-operators) issue from a poor ICP mix is the *ongoing drag* of bad-fit deals through the sales cycle. The Fullcast benchmarks show low-fit ICP deals have 87% lower Sales Efficiency (= (Win Rate * ACV) / Sales Cycle).

Let alone their post-sales underperformance, even *if* a seller can win a low-fit ICP buyer:

or…

Poor ICP Mix = Symptom Of Crowded, Maturing Growth Industry:

Or are these poor ICP leads simply a function of overly crowded categories, causing GTM leadership - under pressure to hit often unrealistic forecasts - *needing* to target marginal-fit buyers?

VIDEO | SLIDES | NEWSLETTER RECAP

Share SaaSletter

Fullcast 2025 GTM Benchmarks - More Highlights

The Fullcast benchmarks reflect continuing market challenges:

Interestingly, the “sales velocity delta: (i.e., top rep outperformance gap) has stabilized… though it remains a “check engine light” in terms of sales sustainability (fewer AEs driving revenue = more risk):

There is a lot more in the full Fullcast 2025 GTM Benchmark report, like these excerpts below - as always, go read the full source.

Guy Rubin’s RIaaS Context + Demo

Fitting this newsletter, here is Guy Rubin’s talk at our July Value Creation Summit that covers the Revenue Insights as a Service offering.

RepVue Q3 2025 Cloud Sales Index

RepVue (h/t Ryan Walsh + Adam Little) released their Q3 2025 Cloud Sales Index.

Their data was more favorable than the Fullcast benchmarks, with quota and leadflow stable to slightly improving:

For our many investor subscribers, I would encourage you to visit the expandable category drill-downs:

… with these flags for *public* companies:

Favorable: Amazon Ads AND AWS ($AMZN), Twilio ($TWLO), Monday ($MNDY), VMware ($AVGO), ServiceTitan ($TTAN)

Negatives: Many of the top performers are private (like Databricks, Apollo.io) → with possible read-throughs to their publicly traded counterparts.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

* Disclosure: Cloud Ratings has a commercial partnership agreement with Fullcast / Ebsta

Coatue recently released their Public Markets Update (SLIDES | VIDEO) - our excerpts + takes below:

Much of the deck covered the “AI Bubble?” debate, framed against 30 bubbles dating back 400 years:

Recognizing that Return On Invested Capital (ROIC) is a crucial metric for the sustainability of any capex cycle, I found Coatue’s Microsoft Azure ROIC history slide powerful:

“Renting GPUs is a bad business” has been a common bear argument. “Commodity” infrastructure can yield 2%-5% ROICs (examples include data center REIT Digital Realty and colocation provider Equinix, as shown below).

Supported by the AI TAM scale and integrated value-add offerings (i.e., Google’s Vertex AI), my view is that this generation of AI Hyperscalers achieves more Cloud-like ROICs than Data Center REIT returns.

Or I can quote Satya Nadella from a recent TBPN appearance:

“The best day in hyperscaler business was when Amazon announced its operating margins. That’s when everybody knew hyperscalers are unbelievable businesses.

It’s a commodity, but at scale, nothing is a commodity.” - Satya Nadella, Microsoft CEO on TBPN 10/28/25

Related callout: Coatue’s ‘30e-35e forecast reflects a 20+% ROIC:

More From Coatue On An AI Bubble?

My call out from this slide: in the “AI Bubble Math” debates (like the Substack from David Cahn from Sequoia - see here and here; even more “AI Bubble Math” coverage here + here), I’ve found internal use capex to be consistently overlooked.

The Coatue slide does a good job capturing internal use AI capex with “Indirect: AI adoption in advertising, commingled with existing core revenues.” To tease internal use further here, try to picture what fully developing + *iterating* AI features for these scaled platforms would cost?

• Amazon: Adding GenAI to their core retail business (scale = $588 *billion* 2025e) and Advertising offering (scale = $75 *billion*). Let alone any undisclosed AI projects (strong competitive reasons to hide).

• Google + Facebook: At consumer scale (i.e., multiple apps at 1 *billion+* user bases), you need a lot of GPUs + datacenters to add always-on GenAI! Let alone whatever “secret moonshots” either is working on.

• Microsoft: A similar dynamic has to exist with adding GenAI to their global installed base. That said, Morgan Stanley (09/26 report with model) *only* estimates 2025e capex directly for M365 Copilot at $630 million + “Internal AI Training” at $1 billion → this example hints my internal use focus might be closer to a rounding error in a Bubble/No Bubble calculation.

Back to more meaningful AI Bubble topics:

The ChatGPT adoption curve is running way ahead of the PC and Internet cycles…

… but when using US Census Bureau data, *business AI* adoption is on an Internet / PC historical trendline. Relative to what is “priced in” for equity valuations, is that good enough?

These Coatue slides should speak for themselves:

As always, do your own work and go read the full Coatue report.

September 2025 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through September 2025 - full slides below:

B2B AI INTEREST INDEX SLIDES

B2B AI interest trends overall reversed recent weakness. Of course, beware of seasonality exiting summer + winter months.

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) reversed a summer lull.

Bellwether Microsoft Copilot accelerated in September. Other “Blue Chips” (i.e., Hubspot, ServiceNow) remain range-bound at best and small in absolute terms.

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel focused, forward-looking SaaS Demand Index) have seen a notable 2025 breakout relative to a very range-bound 2024. With September acclerating:

Only in slides: AI Native B2B apps (n=50) have shifted from strength to decline.

Software Investing Podcast With Joey Brookhart

FULL EPISODE: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

Given the public market theme thanks to Coatue, I wanted to highlight our recent podcast with Joey Brookhart, Co-Founder of AlphaRepo and a longtime long-short software investor, as Managing Partner and Portfolio Manager of Sandbrook Capital.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm:

Tim Sanders + team published G2’s “A Leap of Trust: AI Agents Are Winning Hearts and Wallets” (57 slides; n = 1,000+ B2B software buyers).

My report takeaways:

• More evidence of solid AI ROI

• Surprising signs of a “risk on” mentality from buyers despite meaningful incident rates and higher-than-expected human oversight

Excerpts for the AI ROI support:

This slide inspired my “risk on” take - 83% of buyers are satisfied despite this mix of incident rates:

• Major Incident: 28%

• Moderate: 40%

• Minor: 32%

I appreciated the paraphrasing of (beyond his ‘Co-Intelligence’ book, he is the author of my strongly recommended AI digest on Substack and a Professor at Wharton):

Today’s agents are the worst agents we’ll ever work with. They get better every single day.

I interpreted the 63% more-than-expected need for human involvement as another indication of a “risk on” and “let it rip” despite challenges buyer posture:

August 2025 B2B AI Interest Index from Cloud Ratings

We’ve updated our Cloud Ratings B2B AI Interest Index through August 2025 - full slides below:

B2B AI Interest Index Slides

B2B AI interest trends continue to show some signs of *moderate weakness*.

With this moderate weakness aligning with AI/ML Prioritization in Morgan Stanley’s CIO Survey series, which peaked in Q1 2025:

Thematic Category Interest (n = 47 sub-categories tracked - i.e. “manufacturing AI” or “supply chain AI”) declined for the 5th straight month.

Bellwether Microsoft Copilot ticked up in August. Other “Blue Chips” (i.e., Hubspot, ServiceNow) remain range-bound at best and small in absolute terms.

SaaS Incumbents (n=340: the same 340 vendors tracked in our top-of-the-funnel focused, forward-looking SaaS Demand Index) have seen a notable 2025 breakout relative to a very range-bound 2024. That said, interest is ~flat past quarter:

Only in slides: AI Native B2B apps (n=50) have shifted from strength to plateauing.

Curated AI Content

One of my *very* favorite AI reports: Nathan Benaich and Airstreet Capital published their 313(!) slide “State of AI Report” for 2025

Nathan + team do an excellent job digesting the broad universe of AI research - like this people-pleasing, yes-men finding:

Share SaaSletter

AI x Search Joint R&D Opportunity

For this audience, this slide is a critical topic with significant impacts on a) GTM and b) forecasting publicly traded SaaS going forward that we will be covering in depth:

We are open to *joint R&D* on this AI x Search topic with both institutional investors (as part of our quantamental Alto Model) and software vendors.

Please contact me if interested in an AI x Search R&D collaboration.

Contact For R&D Collaboration

Podcast With Godard Abel Of G2

FULL EPISODE: VIDEO | AUDIO: Apple Podcasts | Spotify | All Platforms

Our podcast with Godard Abel of G2 complements much of this newsletter edition.

About Cloud Ratings

In mid-2024, we announced a research partnership with G2 - more here:

with this slide showing how our G2-enhanced Quadrants (like our recent Sales Compensation Software) release, this business of software newsletter you are reading, our podcasts, and our True ROI practice area all fit within our modern analyst firm: