Everyone’s Adopted It, Few Are Ahead
78% of organisations now use AI somewhere in the business (McKinsey, 2025). Yet in the same research, more than 80% say it’s made no tangible difference to their bottom line. Read those two numbers together and the story is clear: the technology isn’t the hard part. Pointing it at the right work is.
You’ve probably seen this from the inside. Someone in finance uses it to summarise reports. A manager drafts with it. Marketing has a subscription nobody’s reviewed in a year. All of it genuinely useful, none of it joined up, and none of it large enough to move a number anyone reports on. That’s not a failure of the tools. It’s what happens when adoption is left to whoever’s curious enough to try, rather than aimed at the work that actually costs you.
Used Well, It Really Does Work
When Harvard researchers gave consultants AI for tasks it suits, they finished a quarter faster and their output was rated over 30% better (Harvard Business School, 2026). And the biggest gains went to the least experienced people. That’s the real promise: AI doesn’t replace your best people, it brings everyone else closer to them.
Picture what that means across a department. The newest person’s first draft arrives closer to the standard your best person sets. The report that used to come back three times comes back once. Quality stops depending so heavily on who happened to pick up the task. For most organisations the win isn’t a superhuman employee, it’s consistency, and consistency is the thing you’ve been trying to buy through process documents and training for years.
It Sounds Right Even When It’s Wrong
Same study, different task: one that needed judgement across conflicting information. With AI, people reached the wrong answer far more often, and their wrong answers were more persuasive than ever. AI writes a confident argument whether or not the analysis underneath holds up. Knowing that is your best protection. Keep a person on the decisions, and check the thinking, not the polish.
The researchers describe this as a jagged edge to AI’s ability. Two tasks can look equally difficult to an experienced professional while sitting on opposite sides of a line where AI either helps enormously or quietly makes things worse, and you can’t tell which is which by looking. That sounds unnerving, but it’s actually the most useful thing to know in this whole field. It turns AI from a gamble into something you map: try it on real work, find your own edges and put your review effort where the answer matters rather than spreading it thinly over everything.
The Difference Is the Setup, Not the Tool
Of 25 things McKinsey tested, one mattered most for genuine returns: redesigning how the work flows around AI rather than bolting it onto the old routine. Only 21% of organisations have done it (McKinsey, 2025). Handing out licences and hoping is the most common strategy in business today, and it’s why most AI spend produces nothing.
Redesign sounds like a project with a steering group. It usually isn’t. It’s deciding which step AI drafts and which step a person owns, agreeing what gets checked before approval and removing the manual stage that existed only because the old way needed it. Small decisions, made deliberately. Skip them and the technology sits alongside your process instead of inside it, which is exactly where value goes to die.
Start Right and the Rest Gets Easier
Strip away the noise and the picture is actually encouraging. AI is excellent at a growing set of tasks, unreliable at a known set of others, and the organisations seeing results simply matched the technology to the work. That’s it. No moonshot, no giant programme, no leap of faith.
Choose the right first task, set it up properly and keep people where the judgement lives. Done well from the beginning, it isn’t overwhelming at all. It’s one good decision, then another, and the pleasure of it is watching someone in your organisation get an afternoon back and spend it on the work that matters.
