In defense of a better version of yesterday

Fortune’s CEO Daily led last week with a warning from the Fortune Leaders Forum in Macau: don’t fall into AI’s efficiency trap. The line that made the headline came from Malina Ngai, group CEO of AS Watson, who said the biggest risk is using AI to improve today’s processes, “so basically you get a better version of yesterday.” ServiceNow research set up the argument with a gap: about 60% of companies have AI agents, and only 5% have redesigned their workflows around them.

I want to defend the 55%.

At their scale, they’re right

Every speaker at that forum runs something enormous. AS Watson is a 185-year-old retailer with thousands of stores. Syngenta’s CIO talked about decomposing monolithic systems into modular pieces. Walmart China grew 20.7% last quarter. At that scale, the warning is sound: automating an existing process locks in assumptions that predate the technology, and the prize really is the redesign. When you operate ten thousand stores, a reimagined supply chain is worth billions and a faster version of the old one is a rounding error.

Under a hundred people, the advice inverts

Here’s what the headline version loses. That warning was built by and for the Global 500, and it will be read by a million companies a thousandth that size, most of whom have no workflows to redesign in the grand sense. They have an inbox, a task list somebody keeps in their head, and a founder who answers the same question four times a week.

For that company, a better version of yesterday is not a trap. It’s the whole game.

If customer email gets answered in two hours instead of two days, that’s yesterday’s process, faster, and no customer will ever complain that you failed to reimagine the workflow. A fifty-person company that routes invoices automatically hasn’t transformed anything. It got an afternoon back every week, forever, for an hour of setup. These wins are small, cheap, immediate, and they compound. The redesign, meanwhile, costs months of attention the founder doesn’t have and presumes they already know what the transformed version should look like. They don’t. Not yet.

Efficiency is how transformation earns its keep

The deeper problem with “skip efficiency, go straight to reinvention” is that it ignores where trust comes from.

Nobody hands software real autonomy on day one. Nor should they. Teams watch a system do small things reliably, then extend the leash. The drafts were good, so let it send the routine ones. The task extraction was right, so let it route them. The knowledge page updates were accurate, so let it update without asking. Each boring win buys the next delegation. Skip that phase and the reimagined workflow arrives both untrusted and unpracticed. It makes mistakes nobody was watching for, in front of people who never saw it succeed at anything small. A trusted system survives its errors, and a careful rollout earns trust eventually, but the combination of visible failure and zero credit is how transformation programs end up as demos.

We watch this sequence at our own customers. A team starts with yesterday’s process, faster: replies drafted, tasks pulled out of threads. A few months in, the same system has quietly become the place everyone checks to see what was decided and who owns what. That second thing is a transformation by any definition. Nobody planned it. It grew out of the efficiency wins, because the redesign only becomes visible from inside the work.

Even the forum’s own evidence points this way. AS Watson measures success with a customer love score, which improves, by their own account, as AI gives staff more time to spend with customers. Time given back is the efficiency dividend. The reinvention they’re proud of is running on it.

The other trap

So there are two traps, and the second one gets no headlines. Waiting for the transformed workflow means doing nothing while a competitor stacks boring wins every week. Analysis of what the reimagined process should be is cheap now; the trust and the accumulated context that let you actually run it are not, and both only come from shipping the modest version first.

Better than yesterday is how you get to different from yesterday. If your whole company fits on one floor, take the afternoon back. The reinvention will introduce itself. Until then, I believe in yesterday.