In May 2026, Anthropic didn’t announce a new product but a new company, backed by Blackstone, Hellman & Friedman, Goldman Sachs, Apollo, and Sequoia, built specifically to deploy Claude inside mid-sized businesses that have meaningful operations, real complexity, and no in-house team capable of running a frontier AI deployment.

These companies — the ones PIF works with every day — such as community banks, regional health systems, multi-site manufacturers, are now the stated target market for some of the most institutionally-backed AI deployment capacity ever assembled.

The question for finance teams isn’t whether AI is coming. It’s whether they’ll be ready for what it costs, what it creates, and what it looks like in a data room when it’s already running.

Why “mid-market” changes everything

The enterprise AI story of the last three years has mostly played out at the top of the market. The Fortune 500 companies with dedicated AI teams, IT budgets measured in hundreds of millions, and systems integrators already embedded in their operations. Those companies had the internal resources to evaluate vendors, manage implementations, and absorb the cost of getting it wrong.

Mid-sized companies don’t. A regional health system with 800 employees and three clinical sites can’t build an AI deployment capability from scratch. A community bank with £2 billion in assets doesn’t have a team of applied AI engineers. The new services firm Anthropic has built with its institutional partners exists precisely because that gap is real and the commercial opportunity on the other side of it is large.

What this means in practice: AI deployment at mid-market scale is going to look like a managed services engagement, not a software purchase. Applied engineers working on-site. Custom systems built around specific workflows. Ongoing support and iteration that doesn’t end at go-live. The cost profile, the contract structure, and the balance sheet treatment are all materially different from anything most mid-market finance teams have processed before.

Most haven’t started thinking about it yet. The companies that have anticipated the next steps move through their next raise or their next audit significantly more cleanly than the ones that haven’t.

The close process wasn’t built for this

Month-end close at a mid-sized company is typically designed around a stable set of transaction types. Payroll, rent, software subscriptions, cost of goods if there's a product. The processes are documented, the categories are established, and the team knows which numbers to expect.

An AI services engagement introduces costs that don't fit neatly into any of those categories. A discovery phase that blends consulting and scoping. A build phase with milestone payments that may span quarters. Ongoing support fees that vary by consumption. And potentially, capitalised development costs sitting on the balance sheet with a depreciation schedule someone has to defend.

The close process that handled last year's software stack won't handle this without redesign, because it was designed for a different set of inputs: fixed monthly costs, clear vendor categories, and transactions that land in the same place every period.

AI services engagements don't work that way. The cost profile is variable, the phases don't map cleanly to calendar months, and the capitalisation question — whether build costs sit on the balance sheet or run through P&L — requires a judgement call that needs to be made before the engagement starts, not reconstructed afterwards. Most close processes have no category for any of it.

A close process that can handle an AI engagement needs a few things that most mid-market finance teams haven't built yet: a cost categorisation policy that distinguishes between discovery, build, and ongoing support phases; a consistent approach to period allocation for milestone-based contracts; and a clear capitalisation threshold documented before the first invoice arrives. 

The finance teams that figure this out early are the ones running a clean close six months in. The ones that don't are explaining variances to their board without clean supporting documentation.

The gap usually shows up first in a quarterly board pack. It shows up again, more visibly, in a data room.

‍

If you're not sure how your close process handles this, that's worth a conversation.

What investors will start asking

The investor community backing this new services company — Blackstone, Apollo, General Atlantic, Sequoia — will deploy capital into the companies that engage with it. That creates an interesting dynamic: the same institutional investors backing AI deployment are also the ones evaluating the finance functions of the companies being deployed into.

That alignment means the diligence questions are going to sharpen faster than most people expect: 

• What is your AI services spend and how is it categorised? 
• Do you have a material dependency on a single AI vendor or services provider? 
• What happens to your operations if that relationship changes? 
• Is there a system on your balance sheet that was built by a third party and, if so, how are you assessing its carrying value?

These questions aren’t standard yet but they will be. And the finance teams that will answer them cleanly are the ones that built the framework before the engagement started, not the ones trying to reconstruct the rationale two years later under diligence pressure.

The 18-month window

The services company Anthropic has built is operational now. The engagements it runs over the next 12 to 18 months will be the first wave of mid-market AI deployments at institutional scale. 

The companies in that first wave are making decisions right now (about contracts, about cost categorisation, about how to report AI spend to their boards) mostly without a framework for any of it.

Getting this right requires someone who understands how AI services engagements are structured, what they cost, and what investors will want to see on the other side. Most mid-market finance teams don't have that combination in-house. The AI expertise sits with the technology team. The investor readiness expertise sits with whoever prepared the last deck. Neither group is thinking about cost capitalisation policy before the first invoice arrives.

That's the gap an embedded finance partner fills. Someone inside the function who understands what the engagement is, what it will cost, and what an investor will ask about it twelve months from now, and who can build the framework before the spend starts.

PIF Advisory works with growth-stage companies as an embedded finance partner, bringing both the accounting rigour and the investor lens that AI-era finance functions need. If you're evaluating an AI services engagement and want the finance function ready before it starts, get in touch.

Most finance teams in this position start building the framework under diligence pressure, when it's already too late to clean it up properly. If you'd rather not be in that position, let's talk before the engagement starts.

‍