Your client talked through the job offer before they ever got on a call with you. Just not with a person.

A survey released last week put a number on something most coaches have only sensed. 70% of respondents said they involve AI in significant personal decisions, including evaluating job offers, considering relocation or navigating disagreements with people close to them. 73% reported using it somewhere inside their personal relationships.

That is the client intake conversation, the career pivot conversation and the difficult-conversation rehearsal, all happening at 11pm for free.

But the interesting finding isn't the 70%. It's what separated the people who felt better off from the people who didn't.

The 28-point gap

Among respondents whose AI had prompted them to reconsider a question or belief, 76% said their thinking had become sharper since they started using AI. Among those who had never had that kind of interaction, the figure was 48%.

Same tool. Same daily use. Twenty-eight points of difference, and the variable was whether the thing pushed back.

The report's central claim is that how AI engages users — whether it challenges them or simply agrees — correlates strongly with how those users perceive their own cognitive growth.

Read that as a coach and the implication is uncomfortable in a good way. The part of your work clients valued most in that data is not the empathy. It's the interruption.

Who paid for this, and why it matters

Full disclosure on the data. The Vesela 2026 AI and Human Capability Report comes from Gray Sky AI, an Arlington-based startup, and is based on a September 2026 Pollfish survey of 800 U.S. adults who use AI chatbots daily for personal, non-professional purposes. The company's product is positioned as a "human-aligned" model designed to ask questions and promote independent reasoning rather than deliver quick answers.

So the company selling the challenging chatbot published research finding that challenging chatbots are good for you. Hold the finding loosely.

One analyst reviewing the report put the limitation plainly: the methodology — self-selected daily users, subjective self-reporting on cognitive improvement — limits the scientific weight of the conclusions, though the directional signal is worth taking seriously.

The directional signal is the useful part, because it shows up elsewhere. Peer-reviewed research published by the American Psychological Association followed 1,923 workers across North America through 10 simulated tasks involving planning, decision-making and reflective reasoning, and found that participants who modified, challenged or rejected AI suggestions reported greater confidence in their own reasoning than those who accepted output with minimal revision.

"The strongest outcomes were not linked to avoiding AI, but to staying cognitively engaged while using it," said study lead Sarah Baldeo, CEO of ID Quotient and a PhD candidate in NeuroAI at Middlesex University.

Two different samples, two different methods, same direction: friction is the active ingredient.

What to do with it on Monday

Most coaching marketing in 2026 still sells comfort. A safe space. A sounding board. Someone in your corner. Every one of those phrases describes something a free chatbot does at scale, instantly, without judgment, and with infinite patience at 11pm.

The thing it structurally struggles with is the question the client does not want asked. Models trained to be agreeable tend toward agreement, and the survey population is apparently splitting along exactly that line.

Practical version:

Stop writing "support" into your offer copy. Write what you actually do that costs the client something to hear.

Make the pushback explicit in the container. Clients who expect to be challenged interpret challenge as value, not rudeness.

Ask what the AI told them. If a client arrives already half-decided, the coaching work is no longer information. It's testing the reasoning they built with a machine that had no stake in being right.

The 48% group is the market opportunity. Those are people using AI every single day on real decisions, getting agreement, and reporting less of a sense that their thinking improved. They've outsourced the easy half of the job and found out the easy half wasn't the valuable one.