AI and ML
How to Tell Custom AI Development From a Wrapper Around Someone Else’s API
A ready-made bot on someone else’s API and genuine custom development look identical in a demo. The difference only shows up in questions most people never ask.
In short
A chatbot or AI agent demo looks the same in a pitch whether custom development for a specific business process is behind it, or a thin wrapper around someone else’s off-the-shelf API with minimal configuration. The difference shows up in five technical questions: where the agent’s product knowledge comes from, what happens to data on the backend, who owns the output after the project ends, how answer accuracy gets measured, and what the client specifically keeps if they decide to switch vendors.
Why a demo proves nothing
A good chatbot assembled in a day on top of a ready-made builder and an external API can look indistinguishable in a demo from a system engineered over months around a company’s specific data and processes, both answer questions in natural language and give an impression of intelligence. What is worth evaluating is not what shows in the demo, but what sits under the hood and what the client keeps after the project ends.
Five questions that tell the two apart
| Question | A concerning answer |
|---|---|
| Where does the agent’s product knowledge come from? | Vague: "we trained the model" with no detail on the knowledge base, vector store or sources |
| What happens to data on the backend? | No clarity on where requests physically go or whether logs are stored elsewhere |
| Who owns the output after the project ends? | Prompts, knowledge base or configuration remain the vendor’s property |
| How is answer accuracy measured? | No testing process at all, just "looks good in the demo" |
| What does the client keep if they switch vendors? | Practically nothing — all the logic lives inside the vendor’s own infrastructure |
A wrapper is not always the wrong choice, but it should be named honestly
Two separate questions matter here: is a thin wrapper over an off-the-shelf API a bad technical choice, and is it being sold honestly. For a simple task with modest requirements, a wrapper can be a perfectly reasonable, fast and cheap solution, the problem is not using a ready-made API, it is that solution being presented as deep custom development and priced accordingly, at a much higher rate.
How to check the portfolio, not just what gets said in the meeting
Ask for a technical demo with real, unscripted questions, not a pre-agreed scenario but improvised ones, including edge cases and off-topic questions. A custom system genuinely grounded in a specific business’s context handles this noticeably better than a wrapper running a generic prompt, which on non-standard questions either hallucinates or falls back on vague generic disclaimers.
Frequently asked questions
How do you tell custom AI development from an API wrapper in a demo?
The demo alone will not show this, both can look equally impressive. The difference shows up in technical questions: where the agent’s knowledge comes from, what happens to the data, who owns the output after the project, how accuracy is measured, and what the client keeps if they switch vendors.
Is a wrapper over a ready-made API always a bad choice?
No, for simple tasks with modest requirements it can be a perfectly reasonable, cheap solution. The problem is not using a ready-made API itself, it is presenting it as deep custom development and pricing it accordingly at an inflated rate.
What should be requested from a vendor before signing a contract?
A technical demo with improvised, unscripted questions, including edge cases. A custom system handles this noticeably better than a wrapper running a generic prompt, which on non-standard questions either hallucinates or gives vague generic answers.
Why does ownership of the output after the project matter?
Because if the prompts, knowledge base and configuration remain the vendor’s property, the client ends up technically locked in indefinitely, switching vendors means rebuilding the system from zero, not migrating an existing one.
Need a hand with this?
We do this work, not just write about it. Describe the task and we will scope it and send a staged estimate.
Related services
- AI and ML developmentWe build AI that solves a defined task and pays for itself, not demos for the sake of demos. Classifiers, recommendations, text and document processing, LLM assistants, and model integration into an existing product.
- AI strategy consultingMost companies have already run at least one AI pilot. A minority have taken even one pilot to stable, ongoing use. The gap usually is not the model, it is whether the process had a measurable payoff and whether the real running cost was worked out before starting. We help pick the right entry point so budget does not disappear into a demo that stays a demo.
- IT consulting and product auditThe most expensive mistakes in software happen before the first line of code: a misread problem, a stack chosen out of a contractor’s habit, and a specification that does not exist. Consulting exists to settle all of that before the development meter starts running: what to build, out of what, at what cost and in which order. The result is a document, not an opinion on a call.
Read next
- How to Choose a Development Contractor: The Question List and the Red FlagsA practical guide to choosing a development vendor: who needs a freelancer and who needs an integrator, the 25 questions for the first call, how to verify portfolio cases and what belongs in the contract.
- AI Implementation Contracts: Accuracy KPIs, Error Liability, and Model OwnershipA contract copied from a standard software project does not answer the questions specific to AI: what counts as a model error, and who is liable if the agent gives a customer the wrong answer.
- Technical Due Diligence Before Investment or AcquisitionA codebase red flag rarely kills a deal outright — more often it changes the price, the escrow terms, or the post-close commitments. What gets reviewed, and why rushing it under deal pressure produces worse findings.