# Private AI for business: rent a model or own one?

**What you get:** when owning your AI pays off, what to own today, and 5 questions to ask any vendor.

**TLDR:** If your business does $10M to $30M a year, rent your AI today and keep doing that. What you should own now is the part that makes AI good at your work: your jobs written down, your examples and your tests. Get those right first, and you can move to your own model later without starting over.

## What Y Combinator says

Y Combinator, the startup program behind Airbnb and Stripe, put out a short video asking for companies that help businesses own their AI.

Their case: today almost every company sends its questions to a big AI lab and pays per answer. They say open models, ones anyone can download and run, now come close to the best ones on many tests. So more companies will want a model trained on their own data, running where they choose.

"Companies will keep renting frontier intelligence for their hardest problems, but we think many will also own specialized models that encode how their business works."

**Own Your Intelligence** by Jon Xu, Y Combinator: https://www.youtube.com/watch?v=DzblM1Mc-J8

By "frontier intelligence" they mean the biggest, newest models from the big AI labs. A model that "encodes how their business works" has learned that business's own steps, rules and judgment calls, the way a long-time employee has.

## Renting and owning, in plain words

- **Renting:** you use ChatGPT, Claude or Gemini through an account or plugged into your own software. You pay per use. The lab decides when the model changes.
- **Owning:** you take an open model, train it on your own work, and run it on servers you control. You decide when it changes. Your data stays with you.

## When owning starts to make sense

- **Your data can't leave the building.** Health, legal and finance work often has rules on where customer data may go. YC says this is where owning matters most.
- **One narrow job runs thousands of times a day.** Sorting claims, reading invoices, tagging support tickets. A small model trained on that one job can cost less and answer faster than a big rented one.
- **The model is the product.** If what you sell depends on the AI, you may not want a lab deciding when it changes.

If none of these fit, keep renting. The big models are still the best at hard, open-ended work, and YC expects companies to keep renting for that.

## The part you should own today

YC says the hard part is the work around the model: "Someone has to turn workflows into evals, expertise into training data, production failures into improvements."

Evals are tests that check whether the AI did the job right. Training data is the examples you teach a model with. Production failures are mistakes the AI makes while doing the work. So start with three things:

- Write each job down: the steps, the rules, and what good work looks like.
- Keep examples of great work and bad work.
- Turn every mistake into a test, so you catch it if it happens again.

That work pays off whether you rent or own. At Bluhook we run a large team of AI agents (AI software that does a whole job, with its own tools, rules and memory). They run on rented models. Each job's steps, rules and tests live in files we own. When a better model comes out, we switch and keep everything we learned.

If you train your own model one day, those same files become its training data and its tests.

## 5 questions to ask before you buy

If a vendor pitches you on owning your AI, ask:

1. Which one job will the model do, and how is that job done today?
2. How will we test it, and who writes the tests?
3. What happens when a better AI model comes out?
4. Who fixes it when it gets something wrong, and how fast?
5. What does it cost to run each month, next to what we pay to rent today?

If a vendor can't answer all 5, keep renting.

## Start with the jobs

**Take the free AI evaluation at https://bluhook.com/eval. It's about 20 short questions and takes 6 minutes. We email you the 3 jobs in your business an AI agent could take off your hands, and what each one is worth.**