When AI Tools Become Free, What Do Small Teams Sell?
Free agent tools and big-vendor implementation squeeze build fees — here is what small teams can charge for instead.
Look at the past week of AI news and a pattern shows up. Google shipped a major update to Gemini Code Assist and opened its agent mode for free. Microsoft was reported to have stood up a large dedicated unit that works directly with enterprise customers to design and run their AI systems. In Europe, the AI Act moved into an enforcement phase where disclosure and transparency duties for chatbots and AI-generated content are actually being checked, with Anthropic preparing text watermarking and a detection API and OpenAI publishing training-data summaries and provenance signals.
Put together: tools keep getting cheaper, large vendors will now do the building for you, and the rules around running AI in production keep growing. For a small team that is half bad news and half good news. Selling "we will build it for you" loses value fast. Selling "we will keep it working every month" gains value.
What loses its price when tools become free?
What loses value is the act of building. A few prompts, a few screens, an API connection, a document-search chatbot — a free agent tool can now fake all of that in a day. What keeps its value is making the result behave the same way next month, and after the model underneath changes.
Offer a dental clinic "a chatbot that answers booking questions" and the owner will probably try it themselves with a free tool. Offer them "70% of booking questions handled without staff, every wrong answer fixed by the next business day, and a monthly error report" and they cannot do that alone. The first is a deliverable. The second is accountability. Only the first is under price pressure.
So what should a small team sell instead?
Turn the parts that need constant care into the product, instead of a one-off build fee. Even if big tech owns the models and the tools, two things stay with you: the customer's context, and who is on the hook when the answer is wrong.
- Operating promises: response time, deadline for fixing bad answers, a monthly quality report — written into the contract.
- Industry knowledge: the exact phrasing an accounting office uses to request documents, or how a clinic must explain non-covered fees. General models do not know these edge cases.
- Human review loops: design approval steps only where mistakes are expensive — refunds, pricing, medical, legal — and run that process for the client.
- Compliance artifacts: AI disclosure wording, how generated content is labeled, log retention rules. If you have EU or US customers, this is a reason to buy, not paperwork.
- Exit-friendly setup: keep source, prompts and data in the client's own accounts and sell the operations. Giving customers the freedom to leave usually keeps them longer.
Recent industry commentary points the same way: thin wrappers and hourly-labor pricing are being squeezed, while the openings sit in the messy B2B layers — compliance-heavy copilots, human-reviewed services, industry-specific workflows. Messy is exactly what a large vendor struggles to package as a standard product.
What can you change this week?
You can start by rewriting one proposal and one line of your price list. Before any big strategy shift, put your promise into a sentence.
- Split what you currently sell into "building" and "keeping it working." If most revenue sits in the first column, that is a warning sign.
- Write down three moments in your last project where the client was visibly nervous. Those are your paid operations products.
- Pick one measurable promise. Example: "we sample 100 auto-replies each month and fix them free of charge if accuracy falls below the agreed line."
- Add a model-change clause. A single line saying you run regression tests and adapt when a provider swaps models is enough.
- Draft your AI disclosure wording now. The moment you meet a European or US buyer, you will be asked for it.
Tools will keep getting cheaper. But what becomes free is the feature, not the outcome. The opening for small teams is still to write down, and charge for, the promises nobody else wants to make.