Insights
Writing for executives who have bought AI and are trying to work out why the results have not followed. Read by around 10,000 people.
Eighty-eight percent of organisations use AI. About six percent see money from it. The piece argues the gap was never talent or curiosity: we handed everyone the tool and almost none of the conditions that let people use it well.
Read the piece →A CEO goes around the table. Most executives spend under an hour a week actually using AI, which means they are setting direction for something they have never operated. What changes their judgment is not another briefing, it is one morning of building something themselves.
Read the piece →When people suspect that writing down what they know is training their replacement, the sharing stops. Quietly, and without anyone announcing it. This is the argument the gap assessment was built on.
Read the piece →It feels like a breakthrough. Your idea gets validated, your plan sounds airtight, your confidence climbs. That feeling is the failure mode, and experienced executives are more exposed to it than junior staff, not less.
Read the piece →Companies are treating AI the way they treated mobile, cloud and the internet: as an initiative with a budget line and an end date, rather than as infrastructure. The result is paving the cowpath, optimising a process whose steps only exist because people used to have limits.
Read the piece →The AI era will not be won by the people who know the most. Deep expertise is exactly what the models absorbed first, which changes what you should be hiring for and what you should be training.
Read the piece →You don't learn to swim from the deck. This is the case against delegating AI to a task force and reviewing it quarterly, and the argument that distance from the tools is now distance from the truth.
Read the piece →The assumption that younger staff will naturally lead your AI adoption does not survive contact with the research. What HBR found about who is genuinely capable with these tools should change who you pick to lead it.
Read the piece →The stat that gave the firm its through-line. Nearly everyone has AI somewhere, almost nobody can point to earnings from it, and the difference is not the software. The tool isn't the moat. The willingness to rebuild around it is.
Read the piece →Customers did not want the human element. They wanted the outcome, faster and with less friction. The professionals who come through this are the ones using AI to remove steps rather than defend the ones they personally perform.
Read the piece →It doesn't feel like you're getting worse at thinking. It feels like you're getting faster. That gap between how it feels and what is happening is why capability has to be built deliberately rather than assumed.
Read the piece →The leaders who will define the next five years are not the ones with the most experience. They are the ones willing to set that experience down and be a student again, in front of their own teams.
Read the piece →AI made everything possible, and that turns out to be the problem. When the constraint on output disappears, the question stops being how much your teams can produce and starts being whether anyone chose what they should be producing.
Read the piece →Some firms are not just surviving the disruption. They are billing for it. A hard look at what happens to a business whose product is hours when the hours stop being the product.
Read the piece →Nothing under that heading yet.
The ideas we work from
These come up in every session we run, so it's worth being precise about them.
The shape of a failed AI rollout. A company invests, usage climbs for a few weeks, then collapses, and leadership concludes the technology doesn't work. What actually happened is a capability gap being mistaken for an adoption gap, so the company buys more tools instead of building the skill to use the ones it has.
The test for whether a piece of work was redesigned or merely decorated. Did somebody's job description change. Did you start from a blank sheet or from the current process. Did you budget for retraining people or only for licences. And is a number you have tracked for years now dropping. Most teams answer yes to one of the four.
The gap between people who know where the buttons are and the handful of superusers who get real leverage. It is where the value is won or lost, and it closes with six management skills rather than technical ones: briefing well, knowing what to verify, breaking work into pieces, treating a first draft as a starting point, building AI into a repeatable process, and knowing when to stop.
Every labour-heavy process carries two costs: the visible work, and the work behind the work. Scheduling, approvals, chasing, exception handling, reconciliation. That second layer rarely appears as a line in the accounts, and it is usually where AI has the most room to work, because agents can operate inside the systems where the chasing already happens.
The briefing pattern we teach. Situation, described in detail rather than summarised. Persona, named with real edge. Ask, which is the move most people miss: have the model interview you until it has what it needs, because people are poorer at volunteering context than they think. Result, defined precisely. Keep iterating, because a first draft is an opening bid.
The distance between what leadership believes about its own culture and what the team actually experiences. Seventy-six percent of executives think their people are enthusiastic about AI. Thirty-one percent of employees agree. That difference predicts whether a rollout survives contact with the people who have to use it.
New Insights go out to around 10,000 readers. No pitch, no sequence, and you can leave whenever you like.