The real cost of re-briefing AI: the math for a 15 person team

Nobody budgets for pasting. It comes in 90 second pieces, spread across the day, and 90 seconds is below the threshold where anyone counts anything. That's exactly why it's worth counting once, properly, with the assumptions in the open so you can swap in your own numbers.

Here's the math for a 15 person team. Mine is deliberately conservative. Yours will probably come out worse.

The assumptions

Say 15 people, of whom 9 use AI meaningfully on a given workday. In the teams I work with, that share is higher and climbing, but let's stay careful.

Each of those 9 starts around 4 chats per day that need company context: a draft, a proposal section, a client reply, a summary in the house tone. Power users start far more. Again, conservative.

Each context-setting costs about 90 seconds. Finding the background document, copying, pasting, skimming to check it's the current version, topping it up with the details this task needs. If you've done it, you know 90 seconds is generous.

The pasting bill

9 people, 4 chats, 90 seconds. That's 54 minutes per day, team-wide. Call it 19 hours a month.

19 hours is half a working week, every month, spent typing the company's name, tone and rules into a machine that forgot them overnight. At a blended cost of 60 euros an hour, that's roughly 13,500 euros a year, for the pasting alone.

And the pasting is the small line on this invoice.

The bill nobody itemizes

When the context doesn't get pasted, and past three people it often doesn't, the cost moves downstream and changes shape.

The output comes back generic or wrong, so someone senior fixes it. Fixing a draft that missed the tone takes longer than the pasting would have, and it lands on the most expensive people, because they're the ones who can spot what's off. Ask whoever reviews client-facing work at your company how much of their week this is. Then watch their face.

Below that sits the drift: five people briefing from five memories, producing five slightly different companies, with clients quietly noticing. I wrote about that separately, because it deserves its own post. Here it's enough to say the review bottleneck and the drift are both children of the same missing thing, and the 19 hours is just the visible part.

Run your own numbers

Take your headcount, your daily AI users, your chats per day, your seconds per paste. Multiply. I've yet to meet a team that ran this honestly and landed under two working days a month, and I've met several that landed over ten.

Then compare it to what fixing it costs. Writing the company's facts down once is an afternoon, and I'd argue the afternoon pays for itself even if you never automate the delivery, because the writing forces agreements. Automating the delivery, which is the part we built Company Brain for, means the context loads into every chat on its own, and the 19 hours goes back to being work.

The senior reviewer gets a chunk of their week back too. In my experience that's the number that actually moves the decision.

The pasting was never the expensive part. It was just the countable one.