At the start of this year, almost every business I spoke to had a plan for AI. A licence rollout, a pilot, a policy in draft, an intention to "look at Copilot properly this quarter". The middle of the year is when you find out which of those plans turned into something real.
This is not a report card, and it is definitely not an exercise in feeling behind. Think of it as the same mid-year tune-up you would give any other part of the business. Five questions, a straight look at each, and you will know exactly where you stand and what the highest-value next move is. In my experience it is almost never a new tool.
1. Who has actually been trained?
Not who has a licence. Not who attended the launch meeting. Who has sat down, hands on keyboard, and been taught how to use these tools on their real work? That list is usually much shorter than anyone expects, and it maps almost perfectly onto where the value is and is not showing up.
I have written before that a licence is not a skill, and six months into the year that gap is the single most common reason an AI investment is underperforming. The good news is it is also the fastest thing to fix. A focused, hands-on session moves a team further than another quarter of hoping people figure it out themselves.
The simple version: if you cannot name who has been trained, that is your answer, and your next step.
2. Does everyone know what is safe to put in?
One clear line, known by every person: what is fine to put into an AI tool, and what never leaves the building. If your team can recite it, you have done the work. If the honest answer is "we sent an email about it once", it is worth an hour of your attention this month.
The four habits I recommend for using AI safely have not changed, and none of them slow anyone down. This is the one I would re-check mid-year, because teams grow, tools change, and the line only works if the newest person knows it as well as the longest-serving one.
The simple version: safety is a habit everyone knows, not a document somewhere.
3. Are people reaching for the right tool?
By now most businesses are running more than one AI tool, and Copilot, Claude and ChatGPT genuinely do different jobs. The mid-year question is whether your people know which to reach for, or whether everyone defaults to whichever one they met first, for everything.
Watching a capable person wrestle the wrong tool through a task it is not built for is one of the quiet productivity leaks of 2026. It looks like AI is being used. The result says otherwise.
The simple version: the right tool for the job is an AI skill too, and it is teachable.
4. Can you name something that got faster?
Real AI value is visible. A document that used to take an afternoon now takes an hour. Meeting notes that write themselves. Customer replies that go out the same morning. If AI is paying its way in your business, you can point at specific work and say "that, right there, is quicker and better than it was in January".
If you cannot name one, that is not a failure, it is information. It almost always means the tools arrived but the working habits never did, which takes you straight back to question one. And when you can name one, say it out loud to your team. Wins that get named get copied, and the copying is where the compounding starts.
The simple version: value you cannot point at is value you probably do not have yet.
5. What is still manual that should not be?
The best mid-year exercise I know: ask each part of the business for the one task they do every week that feels like it belongs in 2020. The repetitive report, the copy-paste between systems, the summary someone writes by hand every Monday. Pick one, just one, and make it the second half's project.
This is where the second six months can comfortably outperform the first. The early value of AI comes from individuals working faster. The next layer comes from the business deliberately choosing the processes worth rebuilding, and that only happens when someone asks the question.
The simple version: pick one manual process before Christmas. Do it properly. Then pick the next one.
The honest summary
Notice what is not on this list: buying anything. Most businesses I talk to already own more AI capability than they use. The gap between the companies getting real value and the companies just paying subscriptions is not tools, it is skills and habits, and both of those are fixable in weeks, not years.
The tools stopped being the differentiator the day everyone got them. What your people can do with them is the whole game now.
If the check-in surfaced gaps, the fastest fix is the first question. Our AI training is hands-on, on your real work, across Copilot, Claude and ChatGPT, in plain English. One session moves the answers to all five questions.
