Where does AI fit in a flexible space?
What to automate, and what to keep human.
By James Brouard, co-founder, HamletLast updated
The question operators ask me most at the moment is some version of this: where does AI actually fit in my space? Their team is usually asking a quieter version of the same thing. Where does it fit in my job? What should it do, and what shouldn't it?
My answer is a list. Every job in a flexible space sits on one side of a line. On one side are the jobs where a person adds nothing: work nobody was hired to do, and nobody misses once it's done for them. On the other side are the jobs where a person is the whole point. Software should take the first lot completely, and hand the second lot to a person with everything they need to do it well.
We've published our list on hamletco.space. This is the reasoning behind it, job by job.
What to automate
Raising the month's invoices. Nobody joins a coworking space to watch someone build invoices in a spreadsheet. The invoice run is rules applied to records: plans, credits, add-ons, cancellations. Software does it overnight, and it doesn't forget the member who cancelled on the 28th.
Collecting payment, and retrying the ones that fail. Most of the time a failed card is an expired card, or a bank having a moment. Retry it on a schedule, send the reminder, and only bring a person in when the retries run out.
Overdue reminders. The first and second reminder are the same email every time. Writing them by hand only makes them later.
Receipts and booking confirmations. If a member has to ask for a receipt, the system has already let them down once.
Taking a booking at 9pm. Someone wants the boardroom tomorrow morning and finds a free slot at nine at night. They should be able to book it and pay without waiting for anyone to wake up.
Sending the report before anyone asks. The owner wants revenue by site, how hard each room is working and who owes what. That report shouldn't depend on someone staying back on the last day of the month.
What to keep human
The member thinking of leaving. It's one of the most valuable conversations you'll have, and the easiest one to automate badly. Sometimes they're moving interstate. Sometimes they hate the chairs. Sometimes they'd stay for a smaller desk. A form can't find that out. A person can, and they should have the member's history and invoices in front of them when they do.
The late payer with a reason. Three months behind and gone quiet is a different situation from three weeks late with a sick kid at home. Software can tell you who's overdue. It can't tell you who needs a payment plan and who needs a firm phone call.
Welcoming a new member. Hotels worked this out a long time ago: the arrival sets the tone for the whole stay. Software can send the Wi-Fi details the night before. Someone still needs to say hello, show them the good desk and introduce them to whoever runs the Friday lunch.
The tour. Plenty of enquiries are won or lost on the tour. The follow-up should never depend on anyone remembering, but the tour itself is hospitality work.
Putting a double booking right. Mistakes happen. What people remember is how quickly someone fixed it, and whether there was a coffee in it for them.
Knowing the regulars. The person who's in every Tuesday, the company that just hired three people, the member who's quietly stopped coming in. Software can surface the pattern. Noticing it and doing something about it is a person's job.
Why draw the line in public
Two reasons.
The first is trust. Operators are being sold a lot of AI right now, and most of it is vague about what it actually does. A list is specific. If we say Hamlet takes the invoice run and leaves the cancellation conversation alone, you can check that against the product.
The second is that the line moves. AI is getting genuinely good at some of this work, and as it does, jobs will cross from one side to the other. Matching payments, drafting reminders and answering "where's my receipt" are the obvious next ones. When a job moves, we'll say which one and why.
Today the software side runs on automation. Hamlet's built-in AI features are on the way and will run on usage credits, so you only pay for what you use. For operators who want to experiment now, the Hamlet MCP server, in alpha, connects your own AI assistant to your workspace.
The human side stays where it is. A flexible space is a property business and a hospitality business behind one front door. The property side runs on accurate numbers. The hospitality side runs on people having the time to look after other people. Software that hands your team that time back is doing its job.
Draw your own line
If you run a space, try this with your team. Write down every task that happens in a normal week. Put each one on a side of the line. Then add up the hours sitting on the wrong side.
That's the number we're trying to get to zero.
Automate the admin, grow the relationship.
