When the credits run out, your AI team stops.
Most AI employee suites sell you a workforce and then meter its heartbeat. You pay a monthly price, you get an allowance of credits, and every message, image, scrape and scheduled job draws that allowance down. When the balance hits zero, the helpers stop working until the month resets or you buy more. The staff you hired are still on the org chart. They are simply switched off.
We have spent the last month reading the pricing pages, the help centres and, more usefully, what paying customers say in public about month two. The pattern repeats often enough to be worth writing down, because it changes how you should judge every one of these products, ours included.
The arithmetic nobody can do in the shop
A typical suite includes a few hundred credits a month. Published burn rates run something like this. A chat message costs a fraction of a credit, though longer threads cost more because the memory is re-read each time. An image costs one or two. A scheduled research task with web access costs several, and a complex job that touches multiple tools and integrations can cost twenty or more. Stack that up and a few hundred credits buys somewhere between roughly a dozen and fifty automated runs before the workforce goes quiet.
Nobody standing in front of that pricing page can convert "250 credits" into "will this cover my Mondays?" That is not an accident of presentation. A unit the buyer cannot estimate is a unit the buyer cannot compare, negotiate or budget for.
What the meter does to your behaviour
The real cost is not the top-up. It is what the meter does to the way you use the thing you are paying for.
Owners on a credit plan learn to ration. They stop asking follow-up questions, because a long thread costs more than a short one. They skip the extra draft. They cancel the weekly report because it eats the budget they are keeping back for something more important. Every use becomes a small decision about money, and small decisions about money are exactly the friction automation was supposed to remove. You did not buy an assistant. You bought a taxi with the meter running while you think about what to say.
The second-order effect is worse. The expensive jobs are the ones that touch several systems, which are the very jobs where automation earns its keep, so the meter quietly pushes you towards the cheap and shallow tasks instead. The tool ends up doing work you could have done yourself, and not the work you actually needed help with.
When overage is cheaper than the subscription
There is a tell worth checking on any pricing page. Work out the cost of one credit inside the plan (monthly price divided by included credits), then compare it with the price of a top-up credit. On more than one product in this category, the top-up is cheaper than the plan. When your overage undercuts your subscription, the subscription is not priced for the work. It is priced for the annual prepay, and the included allowance is a sample rather than a supply.
How WorkMate OS prices it instead
We took a deliberately boring position: the crew's working day is included. Chats, triage, drafts, follow-ups, bookings, the Monday report, the scheduled missions that run while you are on a job. That is the product, and it is covered by the monthly price. No per-task counter runs behind your work, and nothing stops mid-month because you used your team more than the average.
Launch pricing is £19 a month for Mini Crew, £49 for Office Crew and £99 for Pro Crew (standard £29, £79 and £149), and beta customers keep the launch price for twelve months. What you get for it is a crew, a shared Brain and the modules your plan unlocks. Not an allowance.
One honest exception, stated up front rather than buried in a help article. Genuinely expensive media generation, premium video in particular, is bought separately as a top-up balance. It costs real money per second to produce, and pretending otherwise would mean either a much higher base price for everyone or a quiet cap discovered later. Everyday images are included on every tier. The rule we hold ourselves to is simple: the work you depend on is never the thing that runs out.
What to ask before you pay for any of them
Whichever product you end up with, these four questions separate a plan from a meter:
1. What happens when the allowance runs out? If the answer is "the
helpers stop," you are buying capacity, not staff.
2. What does one unit buy? If you cannot translate it into "how many
Mondays," neither can anyone else.
3. Is the top-up cheaper than the plan? If so, the plan is a deposit.
4. Which jobs cost the most? If it is the cross-system ones, the
pricing is pointed away from the value.
Late payments alone cost UK businesses almost £11 billion a year, with affected firms spending an average 86 staff-hours chasing money (DBT / Small Business Commissioner, July 2025). Chasing is patient, repetitive, unglamorous work a machine should own outright, and it is precisely the work you will ration first when every reminder costs a credit.