Why Directors Running the Most AI Tools Have the Biggest Problem
By every conventional reading, the path from the first cohort to the second is the path AI adoption is supposed to follow.
Across the 2026 sample, the relationship between tool count and self-rated AI impact runs in the direction every vendor would want.
| AI tools used | n | Mean AI impact (1–5) | % rating positive (4–5) | % “no impact” (3) |
|---|---|---|---|---|
| 1 tool | 123 | 3.57 | 55% | 39% |
| 2 tools | 106 | 3.91 | 71% | 26% |
| 3 tools | 94 | 4.01 | 80% | 16% |
| 4 tools | 55 | 4.11 | 76% | 16% |
| 5+ tools | 31 | 4.30 | 88% | 8% |
The single-tool user looks ambivalent — barely above the neutral midpoint, with 39% reporting no impact at all. The five-tool user looks like a convert. The same trend holds by use-case depth (1-use directors at 3.59 mean; 5-use directors at 4.21), and directors doing harder synthesis work outperform the single-document cohort (75% positive vs. 62%). By every conventional reading, the path from the first cohort to the second is the path AI adoption is supposed to follow.
Now the same comparison against the negative-impact list:
| Reported negative impact of AI on board effectiveness | 1-tool users | 3+ tool users | Δ |
|---|---|---|---|
| Integration challenges with existing board systems | 8% | 24% | +17 |
| Inconsistent or unreliable outputs | 7% | 19% | +12 |
| Limited AI knowledge among board members | 17% | 31% | +14 |
| Increased resistance or skepticism among members | 8% | 15% | +7 |
The integration tax climbs three-fold. Inconsistent outputs nearly triple. The share of directors who say AI has revealed limited AI knowledge among their peers climbs from 17% to 31%. The same heavy-user cohort that posts an 88% positive AI impact rating also posts integration challenges at the highest level in the dataset.
The willingness-to-pay column is equally instructive. Mean investment intent in a secure AI solution runs 5.06 among single-tool users and 7.20 among 3+ tool users — a 2.14-point lift on a 10-point scale. A director self-rating their AI practice at 88% positive is, simultaneously, the director most aggressively shopping for something else.
This is not the profile of a sophisticated user who has solved the problem. It is the profile of a director who has built a workable practice and knows it is unsustainable.
Every paste is the director manually feeding context the AI doesn't already have.
A second category exists. Board management AI that operates inside the institutional record
Most Directors Think of AI the Way Most Consumers Think of AI
Beneath every pattern we found sits a category problem the survey wasn’t built to measure directly. Most directors picture AI the way consumers do: a personal assistant. You open a window, paste the material, ask a question, and take the answer. That model travels with you and feels natural, which is exactly what consumer vendors have spent years teaching people to expect.
The model has a ceiling no amount of tool-stacking can break. Every paste is the director manually feeding context the AI doesn’t already have. Every answer comes from a system that does not know what the board decided last quarter, who recused themselves last March, or which committee was named to follow up on the audit finding from November. The synthesis a fiduciary actually needs has to be assembled in the director’s head, every time, from the disparate threads of every consumer-AI conversation they have run that month.
A second category exists. Board management AI that operates inside the institutional record — that already knows the board’s history, inherits the room’s permission structure, and treats the synthesis problem as solved by default — is a categorically different product. It removes the requirement that the director be personally sophisticated about AI, because the sophistication lives in the platform. It removes the security exposure, because the AI never reaches outside the board’s existing envelope. It removes the integration tax, because there is no integration to perform.
Most directors have never experienced this category. The survey suggests they don’t know it exists to ask for. The literacy gap that 27% of respondents named as a negative impact of AI on their board is not, at root, a gap in how to prompt a chatbot. It is a gap in the picture directors hold of what AI in the boardroom can be.
What The Heaviest Users Are Doing With AI
A director running five or more AI tools has, by definition, at most one of them inside the board’s security envelope. The other four are consumer-grade, and they are doing the heaviest work in the practice: summarizing board books (54% of AI-using directors), reviewing governance and regulatory material (where 31% of directors who do it route it through DeepSeek), anticipating board questions before a meeting (67% in enforced-policy boards).
The heavy user’s value-generation engine is the same engine the main report named as the Shadow AI Boardroom. The directors getting the most out of AI are getting it by routing the most sensitive material a fiduciary handles through tools their boards have not approved, do not control, and in most cases do not know are in use. The 88% positive impact is real. The cost — paid in fiduciary risk that does not register on the impact scale — is also real.
The 22% of AI users who report no impact (n=104) average 1.88 tools and 2.67 use cases. They are not AI skeptics. They are the directors who tried the same approach the heavy users built — open a tool, paste a document, ask a question — and walked away. The heavy users stayed and compensated by adding more tools.
What the Heavy Users Figure Out First
The cohort buying the most tools and posting the highest investment intent is the cohort that has hit the personal-assistant ceiling hardest. They have built a practice that works, in the limited sense that it produces value. They have also built a practice that compromises security, accumulates integration debt, and depends on a level of AI sophistication their peers do not share. Their 7.20 mean investment intent — the highest of any cohort in the survey — is the data’s clearest signal that they are looking for something else.
The opportunity for board management software is to introduce the alternative category to a market that has not seen it. Not as a better chatbot. Not as a thirteenth tool added to the stack. As the system that retires the stack — for the heavy user already drowning in it, and for the 22% who walked away because the only model on offer wasn’t built for the work.
Boardroom takeaway. If your board’s most AI-active director is also your loudest voice for a different way to do this, the math above explains why. The model they are using has a ceiling. The model they need has not been on the table.
