Trends · McKinsey report briefing
State of AI Trust in 2026 — Shift to the Agentic Era
Original title: State of AI trust in 2026: Shifting to the agentic era
McKinsey & Company · Tech Forward / Tech & AI · 25 March 2026
This page is a briefing and analysis of the McKinsey article. It is not original Core On research. Figures are only those stated in the briefing.
Slide 1
What this report is
- Analysis of a McKinsey Tech Forward article.
- Lead: trust maturity advanced; gaps remain in strategy, governance, and risk management.
- Survey: 2026 AI Trust Maturity Survey. December 2025–January 2026. About 500 organizations.
- Respondents own or specialize in AI governance, risk, or investment.
Frame
Five responsible-AI dimensions
AI Trust Maturity Model. Four-level scale: from building basic practices to implementing comprehensive, proactive programs.
| Dimension | What the briefing states |
|---|---|
| Strategy | Alignment and oversight lag the speed of technology spread. In the low 3+ group. |
| Risk management | A leading axis. Mitigation still lags awareness. |
| Data and technology | Ahead of governance and agentic controls in every region. |
| Governance | With strategy and agentic, about 30% at 3+. |
| Agentic AI governance and controls | New. First measured this year. In the low 3+ group. |
Reframe
Not only wrong words. Wrong actions.
In the agentic era, stopping a bad sentence is not enough. The system must also handle a bad act.
| Older problem | Current problem |
|---|---|
| Wrong words | Wrong work / actions |
| Responsible AI as a side issue | Responsible AI as a foundation |
Two outcomes of trust:
- Adoption in core workflows, and value
- Management of an expanding risk surface
Theme A · Current state
Insights 1–3
Maturity rose. Strategy, governance, and agentic controls lag. Industry and region differ. Investment is associated with maturity and value. Causation is not claimed.
Insight 1
Maturity improved. Controls lagged.
Responsible-AI maturity improved. Strategy, governance, and agentic controls trail.
Source: McKinsey Exhibit 2. Chart not reproduced.
Insight 2
Industry and region differ
- Asia–Pacific leads.
- TMT and financial services lead.
- In every region, governance and agentic controls lag data and technology.
Source: McKinsey Exhibits 3 and 4.
Insight 3
Investment is associated. Not proven causal.
Organizations that invested $25 million or more in responsible AI show higher maturity. They are more likely to report EBIT impact above 5%.
The article’s wording: not a tax, an enabler.
Do not read $25 million as a magic number. Do not treat the link as causation.
Source: McKinsey Exhibit 5.
Theme B · Risk
Insights 4–7
The top barrier to scale is security and risk. Mitigation lags awareness. Incident reporting is low. Satisfaction with response is lower.
Insight 4
Security and risk block agentic scale
Security and risk were named the largest barrier to scaling agentic AI. Nearly two-thirds. Ahead of regulation and technology.
Source: McKinsey Exhibit 6.
Insight 6
Mitigation lags awareness
Active mitigation lags perceived risk in almost every category. The gap is sharpest for intellectual property and personal data.
Source: McKinsey Exhibit 8.
Insight 7
Few incidents reported. Weak response.
Source: McKinsey Exhibit 9.
Theme C · Response
Insights 8–10
Knowledge and training lead the execution barriers. Explicit ownership tracks higher maturity. Those who treat trust as a business enabler report outcomes more often than harms.
Insight 8
Skills are the leading execution barrier
Knowledge and training gaps lead responsible-AI execution barriers. Nearly 60%. About 50% a year earlier.
Source: McKinsey Exhibit 10.
Insight 9
Explicit ownership, higher maturity
Source: McKinsey Exhibit 11.
Insight 10
Trust as an enabler
Some organizations treat AI trust as a business enabler more than as compliance. They report business performance, efficiency, and customer trust more often than negative outcomes.
Source: McKinsey Exhibit 12.
Gaps
What this implies for agentic AI
- Strategy, governance, and risk gaps are open at once.
- Failure mode moves from speech to action.
- The scale bottleneck is confidence in safe deployment.
- Dual track: value and risk.
- Explicit ownership moves with maturity.
- Skills are a bottleneck.
- Response capacity is weak.
- This is not a program you buy and install.
Critique
Separate survey facts from prescriptions
In the survey: self-reported maturity, barrier ranks, an association between spend bands and outcomes, incident-report rates.
Easy to misread as fact: $25 million as a threshold. Spend causing value. A forecast that one industry must win.
Cautions:
- Self-report bias.
- Possible selection bias. Governance owners answered.
- Exhibits are copyrighted. This page uses captions and stated figures only.
Appendix
Exhibit 1–12 map
Only mappings stated in the briefing. No invented numbers.
| Exhibit | Fact stated in the briefing |
|---|---|
| 1 | No figure or caption in this briefing. |
| 2 | Average 2.0→2.3. About 30% / one-third at 3+ in strategy, governance, agentic. |
| 3 | Asia–Pacific leads. TMT and financial services lead. |
| 4 | Governance and agentic lag data and technology in every region. |
| 5 | $25M+ RAI spend, higher maturity, more likely EBIT impact above 5%. Association. |
| 6 | Security and risk as top barrier. Nearly two-thirds. |
| 7 | Inaccuracy 74%. Cybersecurity 72%. |
| 8 | Mitigation < awareness. Sharpest for IP and personal data. |
| 9 | Incident reports about 8%. Nearly 60% of those with an incident unsatisfied with response. |
| 10 | Knowledge/training barrier nearly 60%. About 50% prior year. |
| 11 | Explicit ownership 2.6 vs none 1.8. |
| 12 | Enabler view reports performance, efficiency, customer trust more often. |
Citation
Source
Alex Singla et al. State of AI trust in 2026: Shifting to the agentic era. McKinsey & Company. 25 March 2026.
This page is a briefing. McKinsey exhibits are not redrawn here.