What AI can actually do for a 60-person company, in plain English

For a company of around sixty people, AI is useful for five practical jobs. It can read the tools you already use and join them into one live picture of the business, draft the repeat work like follow-ups and reports, brief you each morning on what matters, answer plain-English questions about your own company, and flag problems while they are still small. Every action it takes waits for a person's approval before anything goes out. I run my own companies this way, and this article explains each job in the plainest language I can manage.

Why is it so hard to get a plain answer about AI?

Most of what you hear about AI comes from two groups. Vendors who want to sell you something, and commentators who have never run a payroll. Both talk in abstractions, because abstractions are easier to sell and harder to check.

So a successful owner ends up in a strange position. You know AI is supposed to matter, you keep hearing it will change everything, and nobody will tell you what it would do for your business between nine and five on a Tuesday.

I sit on the other side of that. I run my companies on the same AI operating system my team at Imperium installs for clients, so I watch what it does every working day. What follows is that view written down, including the parts where AI is still weak, because those parts are where trust is earned.

What does AI do with the tools you already have?

The first job is unglamorous. A business this size usually runs on more software than the owner could list from memory. Accounting in one place, sales in another, projects in a third, and the truth about the company scattered across all of them plus everyone's inbox.

AI is very good at reading. It can sit across those tools, read what is already in them, and join everything into one live screen of the company, with nothing migrated and nobody retrained. Your team keeps working exactly as they do now, in the tools they already know.

A hospitality group we work with had the answer to almost every operational question already written down somewhere. The problem was the word somewhere. Once the system could read across their bookings, rotas, suppliers and inboxes, the answers stopped hiding.

Can it take the repeat work off your team?

Think about what your best people spend their afternoons on. Chasing invoices, writing follow-up emails, assembling the same weekly report, summarising calls, updating records after every meeting. None of it is the reason you hired them.

AI drafts this work. The follow-up is written and waiting, the report is assembled, the overdue invoice has a polite chaser ready to go. A person reads each draft, edits or approves it, and only then does anything get sent.

A recruitment firm we work with found the inbox was the bottleneck for the whole business. Once the routine replies were drafted automatically and a consultant simply approved them, attention went back where it belongs, to real conversations with candidates and clients. The work still carries a human signature. It just no longer starts from a blank page.

What does the morning brief look like?

Every morning, before the day begins, a short brief lands on your phone. Cash position, what moved yesterday, what needs a decision today, which client has gone quiet, which project is drifting. It reads like a note from a sharp chief of staff, and it takes about as long as your first coffee.

The value is in what it replaces. Most owners rebuild this picture by hand every day, by asking people, scanning inboxes and flicking between apps. The brief does the rebuilding for you, so your first hour starts with decisions instead of detective work.

Mine is the first thing I read each day, and it has quietly ended the oldest bad habit in ownership, which is running the company by whoever emailed loudest.

What happens when you can ask your business a question?

This is the part owners struggle to believe until they see it. You type a question in ordinary English. Which clients have not been invoiced this month. What did we agree with the contractor back in March. How is the new starter settling in against her plan. The system answers from your own data, in plain sentences, and shows you where each answer came from.

For a legal or accounting practice, this changes the texture of partner meetings. Instead of someone promising to pull the figures by next week, the question gets asked and answered in the room.

Nobody on your team has to learn new software for any of this. The system learns your business rather than the other way round.

How does it catch problems while they are still small?

Most business problems announce themselves quietly and early. A client starts replying more slowly, an invoice slips past due, a project's costs creep while everyone is heads-down. Owners usually find out weeks later, once the quiet signal has become an expensive fact.

Because the system reads continuously, it notices the quiet version. A property developer we work with wanted earlier eyes on supplier commitments, because by the time an overspend reached the monthly report it was already history. Now the flag arrives while the decision is still cheap.

One boundary never moves. The system raises its hand and a person decides. Nothing is actioned, sent, paid or promised without human approval, and we build it that way on purpose.

What is AI still bad at?

Plenty, and you should distrust anyone selling AI who skips this section.

It is poor at judgement calls. Whether to drop a supplier, how to handle a grieving employee, when to walk away from a deal that looks good on paper. It can lay out the facts around those decisions, and then a person has to make them.

It is poor at relationships. Your clients buy from your people, and a marketing agency's best account director does something no system can, which is to make a client feel understood.

And it cannot touch physical work or anything without a paper trail. If the knowledge lives only in your foreman's head or in handshakes on site, the system cannot read it. AI works on what gets written down. That is a real limit, and pretending otherwise is how owners get burned.

Will anyone on my team lose their job?

No. In my companies nobody has, and the system is not designed to replace a single role. What people lose is the busywork wrapped around the job they were hired for.

The bookkeeper stops rekeying figures between systems and gets her week back for the judgement work she trained for. Your operations manager no longer spends every morning chasing status updates. Salespeople write fewer recap emails and sit in front of more customers.

Staff tend to warm to it quickly, because the drafts land in their queue and make their day lighter. At this scale the fear has it backwards. Owner-led companies almost always have more useful work than hands available, and the system gives people their hours back so the useful work finally gets done.

Which part do teams actually use first?

Here is something I did not expect, and I have now watched it repeat in my own companies and in every installation since.

Everyone assumes the impressive parts will win the team over, the live screen of the whole company or the chat that answers anything. What gets adopted first, every time, is the approval queue. The plain list of drafted invoice chasers, follow-ups and updates waiting for a yes or a no. Checking it becomes a reflex within days, the way people check their messages.

I think I understand why. The queue is the one place where the system visibly asks permission, over and over, and every approval is small proof that a human is still in charge. Trust gets built in the boring room. Only after that do people wander over to the impressive features, and by then they already believe the system.

So what should you take from all this?

Treat AI at this scale as administrative infrastructure with good manners. It reads what your company already writes down, drafts what your team repeats, briefs you before the day starts, answers when you ask, raises its hand when something drifts, and waits for a person before anything happens.

You do not need to understand how it works, any more than you understand the inside of your accounting software while running a profitable company on top of it. You need to know which jobs it does, which decisions it must never be trusted with, and who is accountable for keeping it sharp. That last question matters most, which is why we install the system on site and then stay to run and improve it every month, rather than leaving you alone with a product.

If you would like to see a morning brief and one live screen of a real company, we would be glad to show you.