The Revenue AI Report

Research

The evidence base behind every claim we publish.

19 themes, 71 charts, 157 primary sources. Adoption against outcome, measured rollback rates, named reversals with dates and dollar amounts, the variables that separate deployments that work, and the gap between what gets spent and what gets attributed.

13%
of sales teams report higher quota attainment from AI. 87% report their reps spend more time selling.Scale Venture Partners with Benchmarkit, n = 278 GTM leaders, published 3 November 2025. Field dates not published.
74%
of enterprises have rolled back or shut down a deployed AI customer communications agent.Sinch, n = 2,527 senior decision makers, 10 countries, fielded January to February 2026.
22%
of finance leaders can tie AI spend to a business outcome.CloudZero, n = 260 finance executives including 135 CFOs, published 24 June 2026.

Patterns worth a closer look

Five charts that carry their whole argument.

Showing chart 1 of 5: The Proof Gap

The Proof Gap·From the theme: The Proof Gap has a measured size

87% of sales teams got more selling time. 13% got higher quota attainment.

Activity outcomes sit above the line. Revenue outcomes sit below it.

Share of sales teams reporting the outcome

  • Increased time reps spend selling87%
  • Increased deal velocity40%
  • Increased pipeline per rep40%
    activity above this line, revenue below it
  • Increased win rate16%
  • Increased average ACV15%
  • Increased quota achievement13%

Top to bottom spread: 74 points

What this does not say

This does not show that AI reduced quota attainment. It shows two outcomes measured in the same sample that moved differently.

Publisher
Scale Venture Partners with Benchmarkit, State of GTM AI 2025
Sample and method
n=278 GTM leaders, published November 3, 2025
Field dates
Not published by the source.

Single sample, self-reported, no field dates published.

Medium confidence

Read the full analysis →

The library

19 themes, each with its own receipts.

Every method term, metric, and model concept used in these themes is defined in The AI and Revenue Dictionary, including what a sample size, a field date, and a confidence tag mean here.

Showing 19 of 19 themes.

  1. Reality check

    Job Loss, Fact or Fiction?

    71% of US adults expect AI to mean fewer jobs. 4% of service firms in the New York Fed district cut anyone because of AI. Total announced US cuts are down 41%.

    Seven in ten Americans now expect AI to shrink the job market. The payroll data, the residency match, the postings index and the Fed's own surveys describe a different economy. This piece measures the gap, lists the losses honestly, and gives each GTM seat one job.

    4 charts · 40 sources →

  2. Discourse pattern scan

    What the AI-in-Revenue Discourse Is Actually Saying, September 2026

    100,000 Coworker activations in 35 days. Outbound volume up about 6.4x while reply rate falls from 4.7% to 2.9%. GTM Engineer postings up 205% year over year.

    Six cross-vendor patterns beneath the launches, pricing changes, practitioner complaints and hiring data, and the question almost nobody is asking.

    2 charts · 19 sources →

  3. Reality Check

    The Manual Tax

    78% of enterprises have an agent pilot running. 14% have scaled one. 54% of workers bypassed the AI their employer bought last month.

    Enterprise AI is stalling because leaders bought tools without subtracting work. The receipts on the pilot-to-P&L gap, the executive-worker split, and why enablement becomes the change management owner.

    10 charts · 22 sources →

  4. Reality Check

    The Task Fallacy

    Anthropic's extreme scenario reaches 17.9% cognitive unemployment only with the reinstatement ratio set to zero. Eight named reversals say otherwise.

    Why Anthropic's extreme scenario assumes away the actual job. Eight named reversals, one enterprise return record, and one model with the reinstatement effect turned off.

    5 charts · 6 sources →

  5. Theme 01

    The Proof Gap has a measured size

    87% of sales teams report reps spending more time selling. 13% report higher quota attainment. Same sample, 74 points apart.

    Every GTM function adopted AI faster than it produced revenue. One dataset measures both sides in the same sample.

    3 charts · 4 sources →

  6. Theme 02

    Adoption is real, measurable, and slower than the discourse

    US Census firm-level AI use says 22.4%. A payments dataset says 55.73%. Both are right, and the survey wording changed mid-series.

    US government data has tracked firm-level AI use every two weeks for three years. It says 22.4%. A payments dataset says 55.73%. Both are right.

    2 charts · 5 sources →

  7. Theme 03

    Rollback is a measured pattern, not an anecdote

    74% of 2,527 senior decision makers had already rolled back or shut down a deployed AI communications agent. 98% were increasing investment anyway.

    74% of enterprises have pulled a deployed AI agent over a governance failure. The rate among self-described mature guardrails is higher, not lower.

    3 charts · 5 sources →

  8. Theme 04

    Named reversals, with dates and dollar amounts

    Twelve documented reversals in five years, each verified at a named source. Two of them put humans back.

    Twelve documented reversals in five years. Two of them put humans back. Public embarrassment gets the coverage, cost and volume get the reversal.

    2 charts · 4 sources →

  9. Theme 05

    What separates the deployments that work

    Of 25 organizational attributes McKinsey tested, redesigning workflows had the largest effect on EBIT impact. At the time, 21% had redesigned any workflow.

    The largest gap between AI leaders and everyone else is not technology. It is having decided what to build.

    4 charts · 5 sources →

  10. Theme 06

    The money does not match the measurement

    22% of finance leaders can tie AI spend to a business outcome. Budgets are still going up.

    $1.76 trillion went to AI in 2025. $826 million of it went to AI data, which is the most-cited barrier in every survey in this library.

    2 charts · 4 sources →

  11. Theme 07

    Trust went down as capability went up

    Seven trust statements, asked twice ten months apart. All seven fell. Only 2% of organizations expect fully autonomous end-to-end processes.

    Seven trust statements, asked twice ten months apart. All seven fell. Expected autonomy is far lower than the market assumes.

    3 charts · 2 sources →

  12. Theme 08

    The same question, four different answers

    Enterprise AI is paying off for 5% of companies, or 74%, depending on which question was asked. None of the four publishers is wrong.

    Enterprise AI is paying off for 5% of companies, or 74%, depending on the question asked. None of the four publishers is wrong.

    1 chart · 4 sources →

  13. Theme 09

    Nobody in revenue publishes research with nothing to sell

    Sixty publishers assessed against four independence tests. Three pass all four. None of the three cover revenue.

    Sixty publishers who write about AI, GTM, or both, assessed against four tests. Three pass all four. None of the three cover revenue.

    2 charts · 3 sources →

  14. Theme 10

    The most-quoted AI numbers do not survive contact with their own methodology

    A 153-person conference survey became a Fortune 500 footnote pointing at example.com. Twelve failure modes now run against every number.

    How a 153-person conference survey became a Fortune 500 footnote pointing at example.com, and the twelve failure modes every number now gets run against.

    3 charts · 4 sources →

  15. Theme 11

    The fear surge is mostly a coverage surge

    Ten years of AI coverage measured directly: volume up 6.2x, tone down 64 percent, never negative. Fear coverage outran works coverage in exactly one year.

    Ten years of AI coverage measured directly. Volume rose 6.2x, tone fell 64 percent and never went negative, and fear coverage outran works coverage in exactly one year.

    7 charts · 5 sources →

  16. Theme 12

    Six of eight seats are losing definition, not headcount

    Six of the eight GTM seats are losing definition, not headcount, once the postings baseline runs first.

    The postings baseline that has to run before anything gets attributed to AI, then each GTM seat with its own series, its own break point, and its own confidence tag.

    6 charts · 6 sources →

  17. Theme 13

    One condition separates agents that work from agents that do not

    Vendors publish the first attempt. By the fourth attempt on the same task, agents without a cheap automatic check before an irreversible action collapse.

    Vendors publish the first attempt. The fourth attempt is the deployment. Where a cheap automatic check fires before an irreversible action, agents hold. Where it does not, they collapse.

    4 charts · 6 sources →

  18. Theme 14

    What the peer-reviewed literature says, ranked

    16 developers forecast a 24% speedup from AI. They delivered a 19% slowdown, and still believed they had been faster.

    Ten studies ranked on identification strength, sample size, outcome relevance, and peer-review status together. The strongest findings contradict the vendor narrative in ten specific places.

    4 charts · 7 sources →

  19. Theme 15

    Buyers push back on AI slop, and the receipts are in

    Human-written articles drew 5.44x the traffic of AI-written ones. Identical human content labeled AI lost engagement from 1.8M visitors.

    Label identical human writing as AI and engagement falls. Human-written articles drew 5.44 times the traffic. The backlash is measurable across writing, music, and images.

    4 charts · 6 sources →

Standards

How these charts are built.

These are constraints, not preferences. They exist so the Research section carries the same evidentiary standard as the newsletter.

  1. 01Every chart carries the publisher, the sample size, and the field dates. A chart without field dates will be wrong in six months and nobody will notice.
  2. 02Every chart carries its full source URL as visible text, so a screenshot of the chart still carries its receipt.
  3. 03Vendor-sponsored research is labeled Vendor research in the same size as the publisher name. The numbers are usable and the interest is disclosable.
  4. 04No figure comes from a press summary when the primary report is public. Where only the summary was reachable, the footer says so.
  5. 05Where sources contradict each other, we show the contradiction rather than picking a winner. Theme 8 exists for that reason.
  6. 06Every chart carries a confidence tag, so you can sort this section by evidence strength rather than by headline.

Subscribe

New research lands in the weekly issue first.

Every issue is built on this library: the receipts, the benchmarks, and the parts that did not work.

Arrives weekly by email. Free. Unsubscribe anytime. By subscribing you agree to our Privacy policy and Terms. We never sell or share the list.