Chapter 014 min read
The gap is about to open.
Waiting for AI to settle feels like the careful choice. I think it is the expensive one, and that the bill arrives all at once.
Faster desks, same process
The CTO of a company of about a hundred people told me they had reached 100% AI use in the first month they measured it. Everyone had a licence, and everyone used it.
Then I asked what had changed. He thought about it. Emails were written faster. Presentations too. He had bought 38 licences for one tool and could not tell me whether eight people used them or thirty.
I hear a version of this in almost every first conversation. AI arrived one person at a time. People got faster at their own desk, and the work between them stayed exactly as it was: the handovers, the checks, the copying from one system to the next.
Nobody decides to wait
Nobody decides to do nothing. They decide to look again after the summer. There is a reorganisation first, or a new system, or a privacy review that has to finish. Each reason is sensible. Together they are a few months a year.
Meanwhile the work goes on by hand. At a maintenance company with crews on more than a dozen sites, the foremen report the week’s hours by photographing a whiteboard in the canteen, or by building a spreadsheet that takes one of them up to an hour. Then someone in the office types every line into the accounting system. Every week.
Everyone at that company uses ChatGPT now and then. The whiteboard is still typed over by hand.
The gap opens all at once
Here is what the organisations that started have found. The first process is the hardest. You have to write down how the work runs, learn to test AI on real work, and agree who owns what. The second one goes faster, because that groundwork is there. So does the third.
Inside one software company we spoke to, one team had rebuilt the way it ships its work around AI, largely on its own. The other teams were still using it to rewrite emails. Same company, same tools, same year. Now picture that difference between you and a competitor.
That is why the gap does not grow in a straight line. For a while you see nothing. Then it shows everywhere at once: in response times, in prices, in who wins the tender.
Waiting never sends an invoice
Waiting feels safe because it costs nothing you can point at. The hours lost to work by hand sit inside salaries that are paid anyway. They are on nobody’s budget. Nothing breaks, so nothing is reported.
If I could change one thing about how organisations decide on AI, it would be this: put a number on waiting. Count the hours one process takes by hand each week, and write that number at the top of the page every time someone proposes to look again after the summer.
Rebuild one process, badly if you must
I would take one process rebuilt imperfectly in March over a perfect AI strategy in December. The first teaches you how your organisation works. The second teaches you how to write a strategy.
Pick the process everyone complains about. Write down how it runs today. Rebuild it so AI does the repetitive steps and a person keeps the decisions. Then do the next one. The organisations that will be hard to catch in two years are doing this now, quietly, one process at a time.
Questions
What teams ask when they compare.
What does an audit ask of us?
Very little. Your leadership sets the goal in one conversation, and each team talks to Stephen, our AI interviewer, by voice or chat whenever it suits them. Two weeks later you see which processes cost the most.
What if the audit shows AI is not the answer?
Then we say so. You keep the map of your operations, every fact with its source, and the ranking of your processes in hours, to use as you see fit.
Can we start small?
That is how every engagement starts: one audit, a fixed price agreed in writing, and one workflow rebuilt. You decide on more only once you have seen it work.