Trust went down as capability went up
Seven trust statements, asked twice ten months apart. All seven fell. Expected autonomy is far lower than the market assumes.
The short answer
What does the research show about Trust went down as capability went up?
Evidence
- What would improve trust, ranked: demonstrated accuracy and reliability 52%, explanation and transparency 45%, security and governance 42%, human oversight with the ability to overrule 36%, ability to measure AI impact 32%.
- Trust rises with deployment stage: above-average trust is 37% at exploratory, 44% at pilot, 47% at implementation. 60% do not fully trust agents to manage tasks autonomously.
- Deloitte, n=3,235 IT and business leaders across 24 countries: 21% of companies planning to deploy agentic AI report a mature agent governance model, so roughly 80% do not.
Supporting pages
- The Forecast Call After AI: What Breaks When Agents Touch Your Pipeline analysis
- Renewal and Expansion Planning When Agents Watch the Account analysis
- Dictionary plain-language definitions
Capgemini asked the same seven statements in June 2024 and again in April 2025. Every one of them fell, in the period when agent capability was improving fastest.
The autonomy expectation series is the corrective to most vendor roadmaps: 45% of organizations expect no agent involvement in the next 12 months, and only 2% expect fully autonomous end-to-end processes.
What this page is
The same seven trust statements asked ten months apart, plus what organizations actually expect agents to do inside their processes over the next year.
The argument
Trust fell during the period when capability rose fastest. Expected autonomy is far below what vendor roadmaps assume, and the barrier that is growing is compliance rather than imagination.
How to read it
- This is the same instrument asked twice, which makes the direction reliable even where the absolute levels are soft.
- Trust here is organizational trust reported by leaders, not end-user trust. They move differently.
- Declining trust alongside rising investment is not a contradiction. Organizations buy through uncertainty and constrain scope instead.
Seven measures of trust in AI agents. All seven fell in ten months.
- Trust a fully autonomous agent for enterprise use-16 pts
- Agents will drive higher workflow automation-14 pts
- Agents will improve customer service and satisfaction-14 pts
- Agents would help me focus on value-added activities-12 pts
- Trust an agent to analyze and synthesize data for me-16 pts
- Productivity potential outweighs the risks-5 pts
- Trust an agent to send a professional email for me-11 pts
What this does not say
Falling trust is not evidence that agents got worse. Two survey waves ten months apart cannot separate capability from expectation.
- Publisher
- Capgemini Research Institute, agentic AI report
- Sample and method
- Same seven statements, asked twice, ten months apart
- Field dates
- June 2024 and April 2025
45% expect no agent involvement in their processes over the next 12 months.
Expected process autonomy, complete six-level series.
Share of organizations, next 12 months
- Level 0, no agent involvement45%
- Level 1, AI-assisted23%
- Level 2, AI-augmented decision-making17%
- Level 3, semi-autonomous with human intervention9%
- Level 4, highly autonomous with strategic oversight4%
- Level 5, fully autonomous end to end2%
What this does not say
Expected autonomy is a plan, not a deployment. Nothing here says any of it shipped.
- Publisher
- Capgemini Research Institute, agentic AI report
- Sample and method
- Next 1 to 3 years: 30 / 25 / 21 / 14 / 7 / 4 across the same six levels
- Field dates
- Not published by the source.
Expected process autonomy, complete six-level series.
| Share of organizations, next 12 months | Value (%) | Note |
|---|---|---|
| Level 0, no agent involvement | 45 | |
| Level 1, AI-assisted | 23 | |
| Level 2, AI-augmented decision-making | 17 | |
| Level 3, semi-autonomous with human intervention | 9 | |
| Level 4, highly autonomous with strategic oversight | 4 | |
| Level 5, fully autonomous end to end | 2 |
Source: Capgemini Research Institute, agentic AI report. Next 1 to 3 years: 30 / 25 / 21 / 14 / 7 / 4 across the same six levels Confidence: Medium.
What this does not say: Expected autonomy is a plan, not a deployment. Nothing here says any of it shipped.
The barrier that grew is compliance. The barrier that shrank is finding a use case.
Deloitte Q1 and Q4 waves across 14 common countries.
- Worries about complying with regulations+10 pts
28% → 38%
- Difficulty identifying use cases-10 pts
36% → 26%
- Difficulty managing risks+6 pts
26% → 32%
- Lack of executive commitment or funding+4 pts
15% → 19%
- Lack of technical talent and skills+3 pts
18% → 21%
- Lack of a governance model-2 pts
22% → 20%
- Trouble choosing the right technologies+1 pts
26% → 27%
What this does not say
A barrier rising in rank does not mean it got worse. Barriers are ranked against each other, so one falling pushes another up.
- Publisher
- Deloitte, State of Generative AI in the Enterprise
- Sample and method
- Q1 n=2,774, Q4 n=2,773
- Field dates
- October to December 2023, and July to September 2024
Also in the record
Figures that sit alongside these charts.
- What would improve trust, ranked: demonstrated accuracy and reliability 52%, explanation and transparency 45%, security and governance 42%, human oversight with the ability to overrule 36%, ability to measure AI impact 32%.
- Trust rises with deployment stage: above-average trust is 37% at exploratory, 44% at pilot, 47% at implementation. 60% do not fully trust agents to manage tasks autonomously.
- Deloitte, n=3,235 IT and business leaders across 24 countries: 21% of companies planning to deploy agentic AI report a mature agent governance model, so roughly 80% do not.
- Executive interest fell across the same waves: C-suite high or very high interest from 74% to 59%, board interest from 62% to 46%, while technical leaders held at 86%.
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.
Seven measures, one direction
Capgemini asked the same seven trust statements in June 2024 and again in April 2025. All seven fell. That happened in the window where agent capability improved fastest, which rules out the simple story that trust follows capability. Trust follows experience, and the experience of the first deployment wave was worse than the expectation that preceded it.
The autonomy series is the corrective to most roadmaps. Forty-five percent of organizations expect no agent involvement in their processes over the next twelve months, and only 2 percent expect fully autonomous end-to-end processes. Anything sold on full autonomy is selling to 2 percent of the market.
The barriers moved
The barrier that grew is compliance. The barrier that shrank is finding a use case. Read together, that is a market that has stopped struggling to imagine applications and started struggling to get them approved. The bottleneck migrated from product to governance in under a year.
Deloitte's number sets the ceiling on how fast that clears: among companies planning to deploy agentic AI, 21 percent report a mature agent governance model, so roughly 80 percent do not. Governance maturity is now the rate limiter on deployment volume.
What restores trust
The ranked list is unambiguous and unglamorous. Demonstrated accuracy and reliability at 52 percent, explanation and transparency at 45 percent, security and governance at 42 percent, human oversight with the ability to overrule at 36 percent, and the ability to measure AI impact at 32 percent. Four of the five are evidence and control. None of them is capability.
Trust also rises with deployment stage: above-average trust runs 37 percent at exploratory, 44 percent at pilot, and 47 percent at implementation. Contact with a working system helps. It does not get you to consensus, since 60 percent still do not fully trust agents to manage tasks autonomously.
The executive attention problem
Interest fell where the budget sits and held where the build sits. C-suite high or very high interest went from 74 to 59 percent and board interest from 62 to 46 percent, while technical leaders held at 86 percent. That divergence predicts a specific failure: technically successful programs that lose their sponsor.
The response is to report evidence in the format the ranked trust list asks for. Accuracy measured, oversight described, impact attributed. A capability demo does not address any of the top five.
What to do with it
The move, by seat.
- CRO
- Pitch scoped autonomy with human override. Forty-five percent expect no agent involvement at all in the next year, and 2 percent expect full autonomy.
- Vendor-facing teams
- Lead with accuracy evidence and oversight design. They are the top two trust restorers and neither is a feature.
- Program owners
- Rebrief the board quarterly with attributed impact. Board interest fell 16 points while technical interest held.
Questions this page answers
What the data says, in plain language.
- What does the research show about Trust went down as capability went up?
- Seven trust statements, asked twice ten months apart. All seven fell. Expected autonomy is far lower than the market assumes. Capgemini asked the same seven statements in June 2024 and again in April 2025. Every one of them fell, in the period when agent capability was improving fastest.
- What does the figure "Seven measures of trust in AI agents. All seven fell in ten months" show?
- Source: Capgemini Research Institute, agentic AI report. Same seven statements, asked twice, ten months apart Fielded June 2024 and April 2025. Confidence: Medium.
- What does the figure "45% expect no agent involvement in their processes over the next 12 months" show?
- Expected process autonomy, complete six-level series. Source: Capgemini Research Institute, agentic AI report. Next 1 to 3 years: 30 / 25 / 21 / 14 / 7 / 4 across the same six levels Confidence: Medium.
- What does the figure "The barrier that grew is compliance. The barrier that shrank is finding a use case" show?
- Deloitte Q1 and Q4 waves across 14 common countries. Source: Deloitte, State of Generative AI in the Enterprise. Q1 n=2,774, Q4 n=2,773 Fielded October to December 2023, and July to September 2024. Confidence: Medium.
- What else sits alongside these figures?
- What would improve trust, ranked: demonstrated accuracy and reliability 52%, explanation and transparency 45%, security and governance 42%, human oversight with the ability to overrule 36%, ability to measure AI impact 32%. Trust rises with deployment stage: above-average trust is 37% at exploratory, 44% at pilot, 47% at implementation. 60% do not fully trust agents to manage tasks autonomously. Deloitte, n=3,235 IT and business leaders across 24 countries: 21% of companies planning to deploy agentic AI report a mature agent governance model, so roughly 80% do not. Executive interest fell across the same waves: C-suite high or very high interest from 74% to 59%, board interest from 62% to 46%, while technical leaders held at 86%.
- Where does this data come from?
- Every figure is reproduced from a named publisher: Capgemini Research Institute, agentic AI report, Deloitte, 2026 State of AI in the Enterprise. Sample, field date, and confidence are shown on each chart. Sources marked as vendor research are labelled on the page.
Cite this page
Permanent URL and suggested citation.
https://www.therevenueaireport.com/research/trust
Kvarfordt, Jonathan. "Trust went down as capability went up." The Revenue AI Report, Research Library. https://www.therevenueaireport.com/research/trust
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.
Capgemini Research Institute, agentic AI report
Trust series asked June 2024 and April 2025
https://www.capgemini.com/wp-content/uploads/2025/07/Final-Web-Version-Report-AI-Agents.pdfDeloitte, 2026 State of AI in the Enterprise
n=3,235 IT and business leaders across 24 countries
https://www.deloitte.com/us/en/about/press-room/deloitte-survey-examines-ai-readiness-agentic-ai-success.html
Could not confirm
What we looked for and did not find.
Claims found during research and not charted
Nothing on this theme was dropped for sourcing. Every claim we found that met the standards on the Research hub is charted above, and anything that failed them would be listed here by name.
Read the analysis
Issues built on this theme.
Research on this site is the evidence layer. These essays take the numbers above and apply them to real decisions, so you can see how the data reads in practice.
- The Forecast Call After AI: What Breaks When Agents Touch Your Pipeline
AI now summarizes calls, updates stages, and nudges next steps. Great. Except your forecast call was built on the assumption that a human rep's judgment sat behind every field. Here is how to rebuild forecast integrity when the judgment is shared with a machine.
- Renewal and Expansion Planning When Agents Watch the Account
Risk detection can be automated. The renewal conversation cannot. Here is the split, the four signal classes worth acting on, and a 120-day renewal plan that says exactly which steps a machine owns.
- The Editor's Mind: How to Review AI Output Without Trusting or Dismissing It
Most AI disappointment is a prompt problem or a review problem. A working method for critical review, and why outlandish claims in both directions deserve the same scrutiny.
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). Trust went down as capability went up. The Revenue AI Report. Retrieved from https://www.therevenueaireport.com/research/trust
MLA
Kvarfordt, Jonathan. "Trust went down as capability went up." The Revenue AI Report, 31 Aug. 2026, www.therevenueaireport.com/research/trust.
BibTeX
@misc{kvarfordt2026trust,
author = {Kvarfordt, Jonathan},
title = {Trust went down as capability went up},
year = {2026},
publisher = {The Revenue AI Report},
url = {https://www.therevenueaireport.com/research/trust}
}Next theme
The same question, four different answersEnterprise AI is paying off for 5% of companies, or 74%, depending on the question asked. None of the four publishers is wrong.
Subscribe