The first workflow to hand an AI: painful, frequent, low-risk, visible

The first workflow you hand to an AI matters more than the tool you pick, and almost nobody treats it that way. They start with whatever is most exciting, or most impressive on a slide, or whatever a vendor pushed hardest. Then it does not quite work, and the whole idea gets quietly filed under "not ready yet."

The first workflow is not really about the workflow. It is about belief. You are trying to convert sceptics into people who have felt the thing work with their own hands. Get that right and the second and third rollouts are easy, because you have credibility to spend. Get it wrong and every future rollout starts in a hole.

So the selection criteria are not "where is the biggest prize." They are "where can I most reliably create a win people can see." Four things.

Painful

Pick something the team actively hates doing. The tedious report, the thing everyone puts off, the task that eats a Friday afternoon and produces nothing anyone enjoyed making.

Pain matters because it sets up the emotional payoff. If the AI takes a genuinely dreaded task off someone's plate, they do not need convincing that this stuff is useful. They felt the relief. That relief is the thing that turns a sceptic into an advocate, and advocates are how adoption actually spreads through a team.

Automating something nobody minded doing produces a shrug. You want a task whose disappearance people will actually celebrate.

Frequent

It has to happen often. Weekly at least, ideally more.

Frequency matters for two reasons. First, a task that happens once a quarter gives you almost no chances to learn and improve the workflow before people forget it exists. Frequent tasks generate fast feedback loops, so you can tune the thing while attention is still on it. Second, the time saved compounds. Saving an hour on something that happens weekly is a real, felt, repeated win. Saving an hour on something annual is a rounding error nobody remembers.

Momentum needs repetition. You cannot build a habit, or a belief, on something that barely ever happens.

Low-risk

The first one has to be forgiving. If the AI gets it wrong, the cost of that mistake should be small and easy to catch.

This is the one people skip because it feels unambitious, and skipping it is how you blow up the whole programme on attempt one. If your first workflow is something where an error is expensive, public, or hard to reverse, then the first time the AI stumbles, and it will stumble, you have handed every sceptic exactly the ammunition they were waiting for.

Low-risk means there is a human check before anything consequential happens, and a mistake is a five-minute correction, not an incident. You are not trying to prove the AI is infallible. You are trying to prove it is useful while it is still learning, which means choosing ground where being wrong is cheap.

Visible

The win has to be seen. If the AI quietly saves someone an hour and nobody notices, you got the productivity but none of the momentum, and momentum is the actual goal of the first workflow.

Choose something whose output other people see, or whose improvement is obvious to the wider team. The report that lands in everyone's inbox looking sharper and arriving earlier. The handover that stops dropping things. Visibility is what turns one person's win into everyone's belief that this is worth their time too.

What this looks like in ops

The examples that hit all four sit right in the middle of operations work.

Weekly reporting is the obvious one. Painful, because assembling it by hand is a chore everyone resents. Frequent, by definition. Low-risk, because a human reviews it before it goes out. Visible, because the whole leadership team sees the result. That is close to a perfect first workflow.

Handovers are another. The moment work passes from one person or team to the next is where things get dropped, and drafting a clean handover from the existing record is exactly the kind of tedious, frequent, checkable, high-visibility task the AI is good at.

Meeting notes, the same shape. Nobody enjoys writing them, they happen constantly, a mistake is trivially fixable, and clean notes with clear actions are visible to everyone who was in the room.

The one mistake to avoid

Do not start with the moonshot. The temptation is to prove the technology with something ambitious, the workflow that would save the most money or impress the most people. That workflow is almost always infrequent, high-risk, or messy, which means it is the hardest to get working and the most damaging when it does not.

Start small, start safe, start visible. Bank a win people can feel. Then spend the credibility on something bigger. Momentum is the whole game early on, and the first workflow is where you either build it or burn it before you have any to spare.