← Back to insights
Insight

The Minutes Go Somewhere

Time saved turns into a return when the work is redesigned around it.

Time saved turns into a return when the work is redesigned around it.

Reading the AI research for government buyers, part 2.

Every AI pitch I see aimed at government comes with a time savings slide, and to be fair, the slide is usually right about the minutes. Where I think agencies go wrong is assuming those minutes turn into a return on their own. In my experience the return comes from deciding, on purpose, where that time goes, and that is a management decision no vendor can make for you.

Recent research took a close look at why AI investment is booming while productivity is not. It was written for companies and investors, but I want to read it through the eyes of the finance, HR, and IT leaders I work with in public agencies, the people who have to explain a technology investment to a council, a board, and the public.

The cost of the tools is not the problem

By the research’s own math, what a typical organization spends on AI is a rounding error next to what an employee costs, which means AI only has to save that person about three productive minutes a week to pay for itself. That is a low bar, and I do not think the license is where the real cost sits. The real costs never show up on an invoice. They live in change management, in retraining, and in all the institutional machinery that surrounds the work. So the question I would ask inside an agency is not whether the tool pays for itself, but how you actually capture the value it creates.

Redesign the coordination, not just the production

Most of what AI speeds up today is the production of work: a first draft, a summary, a reconciliation. But if you look honestly at a working week, most of it is spent on coordination, the meetings, the alignment, and the sign offs. The research is direct about this. AI can hand back the minutes, but it cannot decide what happens to them.

Think about how that plays out in a public agency. A purchase request touches a department head, a budget analyst, and procurement. A new hire touches HR, payroll, IT, the hiring department, and benefits. If AI makes the first step faster but the request still waits in the same five inboxes, nothing meaningful has changed for the resident or the supplier on the other end. The return shows up when those routes are redesigned around the faster first step.

The organizations that won were the ones that reorganized

The finding I found most useful comes from an experiment with young, fast growing companies. Every firm received the same AI tools, the same training, and the same support, and some of them were also shown how other firms had reorganized their work around AI. Most of the gains went to a small group of leaders, and the firms that saw how others reorganized pulled furthest ahead.

I keep coming back to that, because the tools were identical. What separated the leaders was that they rebuilt how their work got done. Government is in exactly the same position. Every agency on the same platform gets the same features, so the features are not going to set you apart. Your process decisions are.

Simplify first, then automate

The research lays out an order of operations that I agree with completely. You question every process and requirement, delete what you can, simplify what is left, speed it up, and only then automate. It also adds a practical point that matters a great deal in our world, which is that automation pays off when AI steps are grouped so a person checks the work once, rather than after every single step. In government, where that review sits is a policy choice, and my view is that you should keep human review exactly where it adds judgment and stop scattering it everywhere out of habit.

Plan for the lag

The research recalls an old observation from economists, that you could see the computer age everywhere except in the productivity statistics. Productivity did pick up, but years later, once organizations had reorganized their work around the machines. I think the lesson for agencies is straightforward. The gain came from reorganizing the work, not from buying the technology, so I would plan for that lag and build the business case on the redesign rather than on savings you expect to see in year one.

What I would do before the RFP

This is work that belongs in Phase Zero, the work Avero does with agencies before an RFP is ever written, and none of it requires a vendor in the room.

I would start by mapping where the time actually goes. Pick three workflows that everyone recognizes, such as a purchase, a hire, and a month end close, and count the handoffs and the waiting. From there I would sort every approval along the way, keeping the sign offs that add judgment and challenging the ones that only add delay.

Then I would name an owner for the time. Someone has to decide what the freed hours become, whether that is faster service, a cleared backlog, or a position you choose not to refill, and I think leaders owe it to their staff to say that out loud rather than leave people guessing. I would measure outcomes the public can feel, like the days it takes to pay a supplier, fill a position, or close the books. And I would group the checks into one review, placed at the point where judgment actually matters.

The minutes always go somewhere. My advice is to decide where before you buy.

I also publish this essay on Substack.

Read Chapter 1 of Start at Zero free

Book a decision review

Next step

Tell us where you are.

A sentence is enough. You’ll hear back within one working day. If we can’t help, we’ll say so.