AI Made You Faster. Is Your Business Still Waiting on You?

One of the owners I coach recently used ChatGPT to rebuild a complex accounting spreadsheet covering 28 employees. The work took roughly five to six hours, including interruptions and checking every employee’s figures. She found two minor errors and corrected them.

A comparable piece of work had previously taken three days and had been billed at around $10,000. That isn’t a like-for-like calculation of savings, but it gives you a sense of the commercial significance.

The immediate win was completing the work much faster. Then she noticed something else: she now had a template she could use again. She could see other clients who might need the service, a role for staff in checking the work, and a reason to rethink hourly billing.

Those possibilities weren’t yet a proven service operating without her. But the conversation had moved beyond how quickly she had finished the spreadsheet.

It had become a question about what the business could now carry.

A faster owner can still be the bottleneck

If you’re using AI in your business, you may already have experienced a version of this. Something that used to take an afternoon now takes an hour. You can prepare the proposal, analyse the information or get a workable first draft without doing every part yourself.

That is useful. Less evening work, a clearer head or simply getting home earlier can be enough reason to keep using it.

But suppose you still have to initiate every task, explain the context, recognise what’s wrong, check the result and tell someone what happens next. The work has accelerated. The dependency around it may be much the same.

You can get through more while remaining essential to every step.

I see a related distinction in delegation. An owner hands a task to someone else, then finds themselves answering the questions, remembering the deadline and chasing completion. The activity has moved. Much of the responsibility for making it happen has stayed with the owner.

In those coaching conversations, I bring the attention back to capability, expectations and ownership. Can the person carry the work? Do they know what good looks like? Can they make the appropriate decisions and follow it through?

AI gives us another place to ask those questions.

What remains after the win?

The spreadsheet matters because something useful remained after the task was completed.

There was a reusable asset. There was also a possible delivery process: use the template, perform the checks and provide the service. Making that process reliable would still require work. Who maintains the template? Who knows when it is appropriate? What needs expert review?

That is where I think the distinction becomes useful.

Productivity helps you complete work with less effort. Business leverage changes how much the business can carry without a proportional increase in your effort.

They can overlap. A productivity gain may improve margins immediately. A faster process may already allow you to serve more clients. You don’t have to build a system around every saved hour. The question is whether a recurring gain remains dependent on you recreating it each time.

If it does, there may be an opportunity to retain the knowledge, clarify the process or develop someone else’s capability. The useful part of the win becomes available beyond the moment in which you produced it.

What the research helps us see

A field experiment reported in an NBER working paper involved 7,137 knowledge workers across 66 firms. Among treated employees who used the integrated generative AI tool, time spent on email fell by about two hours a week, and work outside normal hours also reduced. Researchers detected no significant change in the quantity or composition of tasks.

That supports a distinction between saving time within work and changing the broader pattern of work.

It does not show that AI cannot create organisational leverage. The intervention provided AI at the individual level; it wasn’t a test of deliberately redesigning an entire business. It also wasn’t measuring founder dependency.

The OECD’s survey of more than 5,000 SMEs across seven countries adds another layer. Among SMEs using generative AI, 65% reported improved employee performance, while 35% said it helped them scale. About one-third reported reduced staff or owner workload.

Some businesses were reporting broader benefits. The point is that performance, workload and scaling are different outcomes.

These were self-reported findings. They don’t tell us which operating changes produced the benefits, or prove that capturing knowledge and transferring ownership reduces founder dependency.

That remains a proposition to test in the business, rather than a sequence the research has settled.

When the work starts moving

In a separate part of the owner’s business, completed onboarding now triggers the creation and routing of a new job. A team leader takes the delegation step. The owner can see what is happening without having to create and allocate every job herself.

That is a more concrete change in what the business can carry.

It doesn’t mean everything is resolved. Older work and duplicate tracking still create friction. Nor did the spreadsheet template cause this workflow to happen; they are separate examples.

But together they make the distinction visible. One shows an accelerated task leaving behind an asset. The other shows a process moving work forward through clear steps and human ownership.

The owner still has a role. Her involvement becomes more deliberate.

Look at one recurring task

Choose something AI has made significantly faster in your business. Follow it from the point it begins to the point it is genuinely finished.

Where do you still have to remember, explain, check or chase?

Then look for one change worth making. Perhaps the instructions need to be retained. Perhaps someone needs a clear standard and the authority to act. Perhaps the next step needs an owner, or exceptions need a defined route back to you.

Keep the checking that protects quality. Keep your expertise where it adds necessary value. And count the effort of maintaining the process when you judge whether the change is worthwhile.

Try it across subsequent runs. Notice whether the work moves reliably with fewer recurring interventions from you.

The next time AI gives you a substantial productivity win, pause before filling the space with more work.

Ask what the business could retain from it—and what you would need to change for that capability to be available again without you rebuilding it.

2026/08/02