AI Does Not Delete Work. It Moves It.

Every AI pitch I have sat through in the last two years has the same slide. A task that used to take forty minutes now takes forty seconds. Proposals, support replies, reports, code. The work just disappears.

It does not disappear. It moves. And where it moves to is the single most important thing to understand before you ask "where does AI fit in my business?"

After 35+ years in software and security, and as the working CTO of my own companies, I have watched this play out enough times to call it a rule: AI rarely deletes work. It relocates it, usually from doing the work to checking the work. Whether that relocation is a bargain or a trap depends on one number most companies never measure.

Where the hours actually go

When an AI tool drafts something, somebody still has to decide whether the draft is right. Someone reads the customer reply before it goes out. Someone confirms the numbers in the summary. Someone notices that the confident, well-formatted answer cited a policy you retired last spring.

That checking is real work. It lands on a real calendar. And it almost always lands on your most experienced people, because they are the only ones who can tell when the output is wrong. The hype counts the hours AI removes from the junior end of the process. Reality adds hours back at the senior end, where time is most expensive.

None of that makes AI a bad investment. It makes the sales slide incomplete. The honest question is not "how much faster does the draft appear?" It is "how much total time does the work take now, including the checking?"

The check-to-do ratio

Here is the filter I use as an AI advisor, and it predicts ROI better than any vendor benchmark I have seen. For any task you are thinking about handing to AI, compare two times: how long it takes a person to do the work, and how long it takes a person to check it.

Slow to do, fast to check: a real win. A product description takes thirty minutes to write and two minutes to read. A first pass at matching invoices to purchase orders takes an hour and a glance to confirm. A follow-up email after a sales call takes twenty minutes to compose and thirty seconds to approve. AI belongs here, and the savings are genuine because the checking is cheap.

Same time to check as to do: a quiet loss. A contract review where you have to re-read every clause anyway. A financial analysis someone has to rebuild from scratch before they will trust it. A technical answer that only an expert can verify, and only by working it out themselves. The AI produced the output in seconds and a senior person spent the afternoon redoing it. You did not save time. You bought a subscription and a new chore.

Most of the disappointing AI projects I get asked to look at are not failures of the technology. They are tasks with a terrible check-to-do ratio that nobody measured before the rollout.

You can engineer the ratio

The good news is that the ratio is not fixed. A lot of the work of a fractional CTO or AI consultant on these projects is reshaping the task so checking gets cheap. The tools to do that are rarely more AI. They are design decisions.

• Smaller pieces. Reviewing one paragraph is fast. Reviewing a forty-page report is not. Break the output into chunks a person can verify at a glance.
• Fixed formats. When every output has the same structure, the reviewer knows exactly where to look and what wrong looks like.
• Sources next to claims. If the AI has to show where each fact came from, checking becomes confirming a citation instead of redoing the research.
• A named reviewer with budgeted hours. If nobody owns the checking, one of two things happens: the review quietly stops and errors ship, or your best people become full-time editors of mediocre drafts. Put the reviewer's time in the plan, not in the footnotes.

Who absorbs the moved work?

The last question I ask before any AI project gets a green light is simple: whose calendar gets the checking? If the answer is "nobody in particular," the project is not ready. If the answer is "our three most senior people, for ten hours a week each," you need to price that in before you celebrate the savings.

I make these calls in my own businesses every week, and the pattern holds. The AI rollouts that pay off are the ones where the team counted all the work, including the work that moved. The ones that disappoint are the ones that counted only the work that vanished from the demo.

AI is a remarkable tool. It can take real weight off a business. It just does not make work evaporate, and the companies that win with it are the ones honest enough to follow the hours to wherever they land.

If you want a working CTO to look at which of your tasks are slow to do and fast to check, and which ones would only move the work onto your best people, I take a few clients at a time. Book your free 15-min AI Fit Call →