The Revenue AI Report

The Manual Tax

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.

The Manual Tax: a split editorial panel. The Talk column lists 97% of companies deploying AI agents and 30,000 customers live on Agentforce. The Tax column lists 95% of pilots with zero P&L impact, 89% pilot-to-production failure, 54% of workers bypassing AI tools, and 88% of heavy users reporting burnout.
Source: The Revenue AI Report. Cite: https://www.therevenueaireport.com/research/manual-tax

Share

Posting to Instagram or TikTok? Copy the link, it carries the title, summary and share image.

The short answer

What does the research show about The Manual Tax?

The Manual Tax is the administrative work a revenue team keeps doing after it buys AI. Reps spent 63% of their time on non-selling work in 2014 and about 60% in 2026, so the tools proliferated and the tax barely moved. The published record shows the same pattern at every stage: 78% of enterprises have an agent pilot running, 14% have scaled one, 95% of generative AI pilots show no measurable P&L impact, and roughly eight in ten workers avoided or bypassed the AI their employer bought. The programs that produce receipts name what stops being manual before the tool arrives, roll out in waves, and wire adoption into coaching and OKRs. That work is enablement's charter, which is why enablement is the change management owner in 2026.

Evidence

  • Reps spent 63% of their time on non-selling work in 2014 and about 60% in 2026 per Salesforce State of Sales. The tool count rose. The tax barely moved.
  • Workers lose the equivalent of 51 working days a year to technology friction, up 42% from 2025, per WalkMe's State of Digital Adoption 2026.
  • 33% of workers have not used AI at all and 54% bypassed their company's AI tools in the last 30 days. Combined, roughly eight in ten are avoiding or working around it.

Supporting pages

Dreamforce 2026 was the biggest production the category has staged. AIforce, Koa, Agentforce, Slackforce, and Claudeforce were on every wall, and Salesforce said 30,000 customers were live on Agentforce. In the hallways, the people carrying the AI mandate for their companies were tired. The sentence that travelled furthest afterwards was the one repeated behind closed doors: we can taste it, but we haven't captured it.

That sentiment is not one recap. It is the analyst, media, and worker-survey consensus in September 2026. MIT's NANDA initiative found 95% of generative AI pilots delivered no measurable P&L impact. Writer's 2026 adoption survey found 97% of companies deployed AI agents in the last twelve months and 29% see significant return. WalkMe's State of Digital Adoption, reported by Fortune found 54% of workers bypassed their company's AI tools in the last 30 days and did the work by hand.

The mechanism behind those numbers is not model quality. It is that AI was added on top of the work instead of taking work away. Harvard Business Review's 2026 study of 1,488 US workers recorded the cost of that addition directly: 88% of heavy AI users report increased burnout, and self-reported productivity falls once a worker is running more than three AI tools. Forbes named the sequencing error precisely, describing companies that push usage before trust, productivity before clarity, and automation before a thoughtful redesign of work.

This page measures the tax, shows the two rollout patterns that produce opposite outcomes, and sets out the operating loop and the 90-day plan an enablement leader can take into a QBR. Related reading sits in The Proof Gap and The Reversal Ledger.

What this page is

A measurement of the administrative work that stays on a revenue team after it buys AI, and a read on why the programs that produce receipts subtract work before they add tools.

The argument

Enterprise AI is stalling because the tool arrived without the conditions it needed to work. Data hygiene, workflow redesign, coaching cadence, adoption measurement, and behavior change are those conditions, and they are the enablement charter. That is why enablement becomes the change management owner in 2026.

How to read it

  • The bars on the pilot-to-P&L chart come from four different publishers and four different samples. Read the shape of the drop, not the arithmetic between any two bars.
  • The executive-worker gap is self-reported on both sides. Executives are answering about the organization and workers are answering about their own day, which is part of the finding rather than a flaw in it.
  • The Manual Tax ledger is an illustrative model built by this publication, not a measured average. Use it to size your own activity audit, then replace every number with your own.
  • Two case studies are not a rule. Klarna and the wellness-brand engineering org show that rollout mechanism separated the outcomes in those two cases.

The talk and the tax, measured in the same year.

Left column is what the market said in 2026. Right column is what the published research recorded in the same twelve months.

The talk

  • Companies that deployed AI agents in the last 12 months

    97%

    Writer 2026

  • Customers Salesforce says are live on Agentforce

    30,000

    Dreamforce 2026

  • Executives who say AI is a top strategic priority

    Near universal

    Writer 2026

  • Executives who say their AI strategy is more for show than guidance

    75%

    Writer 2026

The tax

  • Generative AI pilots with no measurable P&L impact

    95%

    MIT NANDA

  • Agent pilots that fail to reach production

    89%

    Deloitte 2026

  • Workers who bypassed company AI tools in the last 30 days

    54%

    WalkMe 2026

  • Heavy AI users reporting increased burnout

    88%

    HBR 2026

The difference between the two columns is work that never came off anyone's plate.

What this does not say

This does not show that AI produces no value anywhere. It shows the distance between announced deployment and measured return.

Publisher
Compiled by The Revenue AI Report from five published sources
Sample and method
MIT NANDA 2025, Deloitte 2026 Tech Trends, WalkMe State of Digital Adoption 2026, Harvard Business Review 2026, Writer AI Adoption Survey 2026
Field dates
Fetched September 2026

Five publishers, five different samples. The columns are not two measurements of one population.

Medium confidence

78% have a pilot. 14% have scaled one. 5% can show a P&L line.

Each bar is a separate published measurement of the same journey, not one cohort followed through five stages.

Share of enterprises at this stage

  • Enterprises with an AI agent pilot running78%

    Teradata survey, 2026

    pilots above this line, production and return below it
  • Enterprises with at least one agent scaled org-wide14%

    Teradata survey, 2026

  • Companies reporting significant ROI from AI29%

    Writer 2026, asked of all deployers

  • Generative AI initiatives with measurable P&L impact5%

    MIT NANDA, 2025

  • Companies McKinsey counts as scaling beyond pilot5%

    McKinsey via Softchoice

Top to bottom spread: 64 points

What this does not say

This is not a funnel of one cohort. No publisher tracked the same companies from pilot through to P&L.

Publisher
Teradata 2026, Writer 2026, MIT NANDA 2025, McKinsey via Softchoice
Sample and method
Four publishers, four independent samples, all fetched September 2026
Field dates
Not published by the source.

Different samples and different definitions of return. Read the shape, not the arithmetic between bars.

Medium confidence
78% have a pilot. 14% have scaled one. 5% can show a P&L line.

Each bar is a separate published measurement of the same journey, not one cohort followed through five stages.

78% have a pilot. 14% have scaled one. 5% can show a P&L line.
Share of enterprises at this stageValue (%)Note
Enterprises with an AI agent pilot running78Teradata survey, 2026
Enterprises with at least one agent scaled org-wide14Teradata survey, 2026
Companies reporting significant ROI from AI29Writer 2026, asked of all deployers
Generative AI initiatives with measurable P&L impact5MIT NANDA, 2025
Companies McKinsey counts as scaling beyond pilot5McKinsey via Softchoice

Source: Teradata 2026, Writer 2026, MIT NANDA 2025, McKinsey via Softchoice. Four publishers, four independent samples, all fetched September 2026 Confidence: Medium.

Caveat: Different samples and different definitions of return. Read the shape, not the arithmetic between bars.

What this does not say: This is not a funnel of one cohort. No publisher tracked the same companies from pilot through to P&L.

Executives and workers are describing different companies.

Each row is one statement put to both groups. The distance is the gap in percentage points.

Executives agreeWorkers agree
  • Workers embrace AI adoption-70 pts

    90% → 20%

  • Employees have adequate AI tools-67 pts

    88% → 21%

  • AI improves workplace efficiency-65 pts

    77% → 12%

  • AI can be trusted for complex, business-critical decisions-52 pts

    61% → 9%

What this does not say

A 67-point gap does not prove the tools are bad. It shows the two groups are not measuring the same thing.

Publisher
WalkMe, State of Digital Adoption 2026, reported by Fortune
Sample and method
Executive and worker samples reported side by side, April 9, 2026
Field dates
Not published by the source.

Self-reported on both sides. Executives answer about the organization, workers answer about their own day.

Medium confidence

Five in every hundred generative AI pilots produced a measurable P&L line.

Each square is one pilot in a hundred. Filled squares reached a measurable business result.

95 of every 100 pilots studied produced no measurable profit and loss impact.

What this does not say

This does not say 95% of AI projects were abandoned. It says their return was never demonstrated on the P&L.

Publisher
MIT NANDA, State of AI in Business 2025, reported by Fortune
Sample and method
Study of enterprise generative AI pilots, published August 18, 2025
Field dates
Not published by the source.

Measurable P&L impact is a strict bar. Pilots with unmeasured or indirect value count as failures here.

Medium confidence

Managers see efficiency. Heavy users report burnout.

Two figures from the same 2026 study of 1,488 US workers.

77%

Managers who believe AI improves efficiency

88%

Heavy AI users reporting increased burnout

Same study, opposite direction

What this does not say

This does not show AI causes burnout. It shows the people using AI most are the people reporting the most strain.

Publisher
Harvard Business Review, 2026, reported by Help Net Security and TechCrunch
Sample and method
n=1,488 US workers. Heavy oversight associated with 14% more mental effort, 12% greater fatigue, 19% greater information overload
Field dates
Not published by the source.

Burnout is self-reported and the study is cross-sectional, so causation is not established.

Medium confidence

Six repeating failure modes in top-down AI mandates.

Each row is a pattern with a published figure or a first-party admission behind it.

  • Mandate without redesign

    Organizations still applying AI to existing workflows rather than rethinking how the work gets done. MTLC 2026.

    63%

    Medium confidence

  • Cost cutting as the only lens

    Klarna's CEO: cost unfortunately seems to have been a too predominant evaluation factor. What you end up having is lower quality.

    First-party admission

    High confidence

  • Usage as a KPI

    Forbes: when employees feel measured by how much AI they use rather than how intelligently they use it, incentives quickly become distorted.

    Mechanism

    Medium confidence

  • Layoff-first framing

    Gartner expects 30% of employees laid off because of AI to be rehired by 2029, often at higher cost.

    30% by 2029

    Medium confidence

  • Tool sprawl

    Self-reported productivity falls once a worker runs four or more AI tools, in a 2026 study of 1,488 workers.

    3 tools

    Medium confidence

  • No named owner in production

    Deloitte and Teradata point at the same absent seat: someone accountable when the agent makes a wrong call at 2am.

    Missing role

    Low confidence

What this does not say

The six are not mutually exclusive, and no publisher has measured which one costs the most.

Publisher
MTLC 2026, Klarna via Transformation Playbook, Forbes June 2026, Gartner via Inc., HBR 2026
Sample and method
Five publishers, compiled September 2026
Field dates
Not published by the source.

These are patterns observed across published cases, not a controlled ranking of failure causes.

Medium confidence

Two rollout patterns, opposite outcomes, same technology.

Top lane is the mandate-and-cut pattern. Bottom lane is the diagnose, embed, and measure pattern.

Mandate and cut: Klarna

  • Company-wide deploymentFeb 2024

    AI assistant reported handling 2.3 million chats in a month, described as the work of 700 agents

  • Headcount falls2024

    Staff reduced from roughly 5,000 to roughly 3,500

  • Quality complaints2024

    Complex disputes, distressed customers, and multi-step problems fell through the gaps

  • Public reversalMay 2025

    CEO tells Bloomberg the company went too far and starts rehiring humans

Diagnose, embed, measure: global wellness brand engineering org

  • BaselineWeek 0

    Near zero AI adoption. Three engineers embedded as an engineering engagement, not a training program

  • AuditWeeks 1 to 2

    Existing tooling, workflows, and team appetite diagnosed before any stack was selected

  • Wave oneWeeks 2 to 3

    Six-person core team, hands-on pairing

  • Wave twoWeeks 3 to 4

    Eleven-person dev team, adoption metrics wired into team OKRs

  • Full orgWeeks 5 to 6

    All 38 engineers. PR velocity up 40%, 13 outside contractors offboarded

What this does not say

Two cases do not establish a rule. They show that rollout mechanism, not model choice, separated the outcomes here.

Publisher
Forbes May 2025 and Transformation Playbook April 2026 for Klarna; Check + Pluris case study 2026 for the engineering org · Vendor research
Sample and method
Two published cases, different industries and different sizes
Field dates
Not published by the source.

The engineering case is a vendor-published case study. The outcome figures are the vendor's own.

Medium confidence

Where the Manual Tax sits, hour by hour, in an illustrative rep week.

Hours reclaimed per rep per week if each task is eliminated rather than assisted. This is a model, not a measurement.

Hours reclaimed per rep per week

  • CRM data entry5.5hrs

    6.0 hrs before, 0.5 after

  • Meeting prep and research4hrs

    5.0 hrs before, 1.0 after

  • Follow-up email drafting3.5hrs

    4.0 hrs before, 0.5 after

  • Post-call notes3.3hrs

    3.5 hrs before, 0.2 after

  • Pipeline hygiene2.2hrs

    2.5 hrs before, 0.3 after

  • Forecast prep1.6hrs

    2.0 hrs before, 0.4 after

Top to bottom spread: 20.1 hrs per rep per week

What this does not say

These are not measured hours from any survey. Individual results will vary by segment, motion, and CRM hygiene.

Publisher
The Revenue AI Report, illustrative model
Sample and method
Task list from Salesforce State of Sales 2026 (reps spend about 60% of time on non-selling tasks) and Accordion 2025 (conversational CRM reducing sales admin work 60% to 80%)
Field dates
Modelled September 2026

Illustrative. The hour splits are a model for sizing a baseline, not a measured average. Run your own activity audit before quoting any number.

Low confidence
Where the Manual Tax sits, hour by hour, in an illustrative rep week.

Hours reclaimed per rep per week if each task is eliminated rather than assisted. This is a model, not a measurement.

Where the Manual Tax sits, hour by hour, in an illustrative rep week.
Hours reclaimed per rep per weekValue (hrs)Note
CRM data entry5.56.0 hrs before, 0.5 after
Meeting prep and research45.0 hrs before, 1.0 after
Follow-up email drafting3.54.0 hrs before, 0.5 after
Post-call notes3.33.5 hrs before, 0.2 after
Pipeline hygiene2.22.5 hrs before, 0.3 after
Forecast prep1.62.0 hrs before, 0.4 after

Source: The Revenue AI Report, illustrative model. Task list from Salesforce State of Sales 2026 (reps spend about 60% of time on non-selling tasks) and Accordion 2025 (conversational CRM reducing sales admin work 60% to 80%) Fielded Modelled September 2026. Confidence: Low.

Caveat: Illustrative. The hour splits are a model for sizing a baseline, not a measured average. Run your own activity audit before quoting any number.

What this does not say: These are not measured hours from any survey. Individual results will vary by segment, motion, and CRM hygiene.

The enablement operating loop: five moves that turn AI capacity into behavior change.

Skip a node and the pilot stalls at the node before it.

  1. 01 · 01

    Diagnose

    AI now scores close to every call and demo rather than the two or three a manager had time to sample. Signal arrives at scale.

    https://braintrustgrowth.com/9-ways-ai-is-actually-changing-sales-enablement-in-2026/

  2. 02 · 02

    Redesign

    For every capability introduced, name what stops being manual work. If nothing stops, the deployment fails.

    https://www.mtlc.co/2026-ai-in-revenue-enablement-report-trends-insights/

  3. 03 · 03

    Coach

    Identifying the gap and closing the gap are two different jobs. Managers close them on a weekly cadence enablement designs and enforces.

    https://braintrustgrowth.com/9-ways-ai-is-actually-changing-sales-enablement-in-2026/

  4. 04 · 04

    Measure

    Hours reclaimed, cycle time, ramp time, win rate. Not licenses activated and not dashboard usage.

    https://www.forbes.com/sites/kathycaprino/2026/06/26/why-ai-adoption-is-failing-inside-many-companies/

  5. 05 · 05

    Prove

    Before-and-after evidence signed by RevOps unlocks the next wave. Without it the budget is cut in the next planning cycle.

    https://checkpluris.com/case-study/a-global-wellness-brand-lifted-pr-velocity-40-in-6-weeks

What this does not say

Running the loop does not guarantee a return. It removes the failure modes recorded in the cases above.

Publisher
Forrester February 2026, Braintrust September 2026, Check + Pluris 2026, CI&T June 2026
Sample and method
Loop compiled by The Revenue AI Report from four published sources
Field dates
Not published by the source.

This is a framework, not a measured process. No publisher has tested the five nodes as a controlled sequence.

Low confidence

The 90-day receipts-first playbook, phase by phase.

Four phases, each with a definition of done an enablement leader can be held to.

Days 1 to 14: measure the Manual Tax

  • Activity auditWeek 1

    One-week audit on three to five reps using calendar data, CRM activity logs, and email metadata. Categorize every minute

  • Baseline publishedWeek 2

    Selling time as a share of total time, plus hours per week on CRM admin, meeting prep, forwarded threads, and forecast prep

  • Definition of doneDay 14

    A dollar figure named for the current Manual Tax, and the top three tasks reps would pay to eliminate

Days 15 to 45: design the subtraction

  • Capability mappingWeek 3

    For each of the top three tasks, name the capability that eliminates it. Not augments. Eliminates

  • Wave one selectedWeek 4

    Five to eight reps. No company-wide rollout

  • Accountability setWeek 6

    Adoption wired into team OKRs, weekly manager coaching cadence set, an operational owner named for every agent going into production

  • Definition of doneDay 45

    Wave one launched with named accountability

Days 46 to 75: run the wave, measure the receipts

  • Weekly measurementWeeks 7 to 10

    Selling time as a share of total, cycle time, ramp time, win rate on stage two and above, admin hours reclaimed per rep per week

  • CoachingWeekly

    At least one live scenario per rep per week, tied to a specific AI-flagged gap

  • Tool disciplineOngoing

    Kill any tool that pushes the count above three AI tools per rep

  • Definition of doneDay 75

    Four weeks of measured before-and-after data

Days 76 to 90: publish the CFO-grade evidence

  • Signed reportWeek 12

    Before-and-after table on the wave-one team, signed by the RevOps lead

  • ExtrapolationWeek 13

    Hours reclaimed multiplied by loaded cost per hour, extrapolated to full team size

  • Definition of doneDay 90

    Wave two approved on evidence rather than politics

What this does not say

Ninety days is a planning horizon, not a promise of return.

Publisher
The Revenue AI Report, synthesized plan
Sample and method
Built from the Check + Pluris 2026 rollout pattern, Softchoice October 2025, Forrester February 2026, and CI&T June 2026
Field dates
Published September 2026

This is an operating plan assembled from published cases, not a tested protocol with a control group.

Low confidence

Also in the record

Figures that sit alongside these charts.

  • Reps spent 63% of their time on non-selling work in 2014 and about 60% in 2026 per Salesforce State of Sales. The tool count rose. The tax barely moved.
  • Workers lose the equivalent of 51 working days a year to technology friction, up 42% from 2025, per WalkMe's State of Digital Adoption 2026.
  • 33% of workers have not used AI at all and 54% bypassed their company's AI tools in the last 30 days. Combined, roughly eight in ten are avoiding or working around it.
  • 93% of executives cite change management as the primary AI barrier, per Cornell ILR June 2026. It is also the line item most often cut.
  • Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027, per its April 2026 Hype Cycle.
  • CI&T reports 100% team adoption and AI maturity moving from 33 to about 65 out of 100 by embedding AI into existing engineering workflows rather than running a separate program.
  • Tim Sanders of G2 at Dreamforce 2026: the majority of agentic outcomes are not driven by frontier capabilities, they are driven by last year's AI.
  • Executives at AI companies told the BBC that AI means less work while their own staff described 70 to 90 hour sprint weeks.

The brief

What is going on here, and why it matters.

The charts above are the evidence. This is the read: what the data describes, the mechanism behind it, where the argument could be wrong, and what a revenue team does about it.

01

What is actually happening

Reps spent 63% of their time on non-selling work in 2014. Salesforce State of Sales puts the 2026 figure at about 60%. Twelve years of tooling moved the number three points. The spend went up, the tool count went up, and the administrative load stayed where it was.

The reason is sequencing. AI was layered on top of the existing process rather than used to remove steps from it. MTLC found 63% of organizations still applying AI to workflows they never rethought. When a capability is added without a matching subtraction, the worker carries both the old task and the new tool, which is what the burnout and bypass numbers on this page are recording.

02

Why workers are routing around the tools

54% of workers bypassed their company AI tools in the last 30 days and did the work by hand. 33% have not used AI at all. The trust split behind that behavior is wide: 61% of executives trust AI for complex business-critical decisions against 9% of workers, and 88% of executives say employees have adequate tools against 21% of workers.

Forbes named the mechanism that widens the split. Managers are put on the front line to monitor dashboards and push usage, so adoption becomes a performance signal rather than a business practice. Employees respond by using AI to satisfy the metric. The dashboard goes green and nothing downstream moves.

03

What separates the two rollout patterns

Klarna deployed company-wide, framed the change as cost, and cut headcount from roughly 5,000 to roughly 3,500. Fourteen months later the CEO said the company had gone too far and started rehiring, positioning human support as a premium tier. The failure modes were specific: complex disputes needing judgment, distressed customers needing de-escalation, and multi-step problems falling through the gaps between interactions.

The engineering case ran the opposite sequence. Diagnose the existing tooling and team appetite first, embed engineers rather than schedule training, roll out in three waves over six weeks, and wire adoption into team OKRs so usage was a tracked commitment. The reported result was a 40% lift in PR velocity and 13 contractors offboarded. The technology was not the variable.

04

Why enablement owns this

Forrester's February 2026 read is that the barrier is readiness rather than functionality, and that enablement teams are historically underresourced while now being asked to manage data hygiene, orchestrate agents, and coordinate content, training, and performance insight across functions.

Braintrust names the fork. AI can score close to every call a rep runs and flag the skipped discovery question or the talk-to-listen spike, but identifying a gap and closing it are two different jobs. Organizations quietly waste the gain when they buy the tool that frees manager time and then fill that time with more reporting instead of more reps in the room with a manager watching.

05

Where this argument could be wrong

Most of the figures on this page are self-reported and cross-sectional, so none of them establish causation. It is possible that the pilots showing no P&L impact are simply early, and that the 2027 measurements look different once deployments mature.

The counter-case is the consistency. Adoption is high and return is low across four independent publishers using different samples and different definitions, and the worker-side data moves in the same direction as the executive-side data fails to. A measurement error would not usually be that tidy.

What to do with it

The move, by seat.

CRO
Refuse a purchase order without a named subtraction. Ask which hour comes off the rep week and who signs the before-and-after.
Enablement
Run a one-week activity audit on three to five reps, publish the baseline, and take the top three tasks reps would pay to eliminate into the next vendor conversation.
RevOps
Name an operational owner for every agent going into production, accountable when it makes a wrong call at 2am, and hold the count at three AI tools per rep.
CFO
Fund change management before the model. Measure hours reclaimed multiplied by loaded cost, not licenses activated or dashboard usage.

Questions this page answers

What the data says, in plain language.

What does the research show about The Manual Tax?
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. Dreamforce 2026 was the biggest production the category has staged. AIforce, Koa, Agentforce, Slackforce, and Claudeforce were on every wall, and Salesforce said 30,000 customers were live on Agentforce. In the hallways, the people carrying the AI mandate for their companies were tired. The sentence that travelled furthest afterwards was the one repeated behind closed doors: we can taste it, but we haven't captured it.
What does the figure "The talk and the tax, measured in the same year" show?
Left column is what the market said in 2026. Right column is what the published research recorded in the same twelve months. Source: Compiled by The Revenue AI Report from five published sources. MIT NANDA 2025, Deloitte 2026 Tech Trends, WalkMe State of Digital Adoption 2026, Harvard Business Review 2026, Writer AI Adoption Survey 2026 Fielded Fetched September 2026. Confidence: Medium. Caveat: Five publishers, five different samples. The columns are not two measurements of one population.
What does the figure "78% have a pilot. 14% have scaled one. 5% can show a P&L line" show?
Each bar is a separate published measurement of the same journey, not one cohort followed through five stages. Source: Teradata 2026, Writer 2026, MIT NANDA 2025, McKinsey via Softchoice. Four publishers, four independent samples, all fetched September 2026 Confidence: Medium. Caveat: Different samples and different definitions of return. Read the shape, not the arithmetic between bars.
What does the figure "Executives and workers are describing different companies" show?
Each row is one statement put to both groups. The distance is the gap in percentage points. Source: WalkMe, State of Digital Adoption 2026, reported by Fortune. Executive and worker samples reported side by side, April 9, 2026 Confidence: Medium. Caveat: Self-reported on both sides. Executives answer about the organization, workers answer about their own day.
What does the figure "Five in every hundred generative AI pilots produced a measurable P&L line" show?
Each square is one pilot in a hundred. Filled squares reached a measurable business result. Source: MIT NANDA, State of AI in Business 2025, reported by Fortune. Study of enterprise generative AI pilots, published August 18, 2025 Confidence: Medium. Caveat: Measurable P&L impact is a strict bar. Pilots with unmeasured or indirect value count as failures here.
What else sits alongside these figures?
Reps spent 63% of their time on non-selling work in 2014 and about 60% in 2026 per Salesforce State of Sales. The tool count rose. The tax barely moved. Workers lose the equivalent of 51 working days a year to technology friction, up 42% from 2025, per WalkMe's State of Digital Adoption 2026. 33% of workers have not used AI at all and 54% bypassed their company's AI tools in the last 30 days. Combined, roughly eight in ten are avoiding or working around it. 93% of executives cite change management as the primary AI barrier, per Cornell ILR June 2026. It is also the line item most often cut.
Where does this data come from?
Every figure is reproduced from a named publisher: Accordion, How conversational CRM reduces sales admin work by 60-80%, Axios, Dreamforce finds itself hosting AI debate, BBC, Tech leaders say AI means less work, their staff say they work up to 90 hours a week, Braintrust Growth, 9 ways AI is actually changing sales enablement in 2026, Check + Pluris, A global wellness brand lifted PR velocity 40% in 6 weeks, CI&T, 100% team adoption and 2x AI maturity for a leading automotive retail group. Sample, field date, and confidence are shown on each chart. Sources marked as vendor research are labelled on the page.
What could not be confirmed?
Chicago Booth research cited secondhand claims nearly 30% of employees actively sabotage their employer's AI strategy, rising to 44% among Gen Z. We could not reach the primary study, so it is not charted. The Teradata figures of 78% piloting and 14% scaled are widely quoted in September 2026 coverage. We could not reach a primary Teradata publication with the sample size and field dates, so the bars carry a caveat. Gartner's Agentic AI Hype Cycle prediction that more than 40% of agentic projects will be canceled by end of 2027 sits behind a paywall. It is listed as a takeaway from secondary reporting rather than charted. The per-task hour splits in the Manual Tax ledger are an illustrative model built by this publication. No publisher has released a measured hour-by-hour breakdown of rep admin time at task level.

Cite this page

Permanent URL and suggested citation.

https://www.therevenueaireport.com/research/manual-tax

Kvarfordt, Jonathan. "The Manual Tax." The Revenue AI Report, Research Library. https://www.therevenueaireport.com/research/manual-tax

Figures on this page are reproduced from the publishers listed below. Cite the original publisher for the underlying data, and this page for the compilation and framing.

Sources

Every publisher used on this page.

If a metric, model term, or method on this page is unfamiliar, every one of them is defined in The AI and Revenue Dictionary. Sample size, field date, and confidence tags are explained there too.

Could not confirm

What we looked for and did not find.

Claims found during research and not charted

  • Chicago Booth research cited secondhand claims nearly 30% of employees actively sabotage their employer's AI strategy, rising to 44% among Gen Z. We could not reach the primary study, so it is not charted.
  • The Teradata figures of 78% piloting and 14% scaled are widely quoted in September 2026 coverage. We could not reach a primary Teradata publication with the sample size and field dates, so the bars carry a caveat.
  • Gartner's Agentic AI Hype Cycle prediction that more than 40% of agentic projects will be canceled by end of 2027 sits behind a paywall. It is listed as a takeaway from secondary reporting rather than charted.
  • The per-task hour splits in the Manual Tax ledger are an illustrative model built by this publication. No publisher has released a measured hour-by-hour breakdown of rep admin time at task level.

How to cite this research

Written by Jonathan Kvarfordt, Founder and Principal Analyst, The Revenue AI Report. Published under CC BY 4.0.

APA

Kvarfordt, J. (2026). The Manual Tax. The Revenue AI Report. Retrieved from https://www.therevenueaireport.com/research/manual-tax

MLA

Kvarfordt, Jonathan. "The Manual Tax." The Revenue AI Report, 31 Aug. 2026, www.therevenueaireport.com/research/manual-tax.

BibTeX

@misc{kvarfordt2026manualtax,
  author = {Kvarfordt, Jonathan},
  title = {The Manual Tax},
  year = {2026},
  publisher = {The Revenue AI Report},
  url = {https://www.therevenueaireport.com/research/manual-tax}
}

Share this research

Posting to Instagram or TikTok? Copy the link, it carries the title, summary and share image.

Next theme

The Task Fallacy

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.

Subscribe

Get the weekly issue built on this data.

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