Most enterprise AI spend is not a decision. It is an accumulation.
One team buys assistant seats. Then another. Then a third picks a different vendor because someone preferred it. Each purchase cleared on its own, and no single person owns the total.
At 1,000+ employees, that usually means three things at once: overlapping tools, frontier models billed at retail on separate agreements, and no view of who is using what.
The governance question tends to arrive later than the invoice.
At Bolt we put a gateway in front of models, keys, and tools so a company governs them in one place — SSO/SCIM, self-hosting, data residency, admin audit, and pooled consumption at cost + 5% instead of every team paying list price alone.
If you are the one person asked to explain AI spend across departments, I would be glad to compare notes on how you are handling it.