Foundations
What is an AI model?
Vendors use the word loosely, and the looseness costs money. A model can mean the big rented system from a provider, a small one trained on your data, or the whole product. Which one you are paying for decides your running cost, your lock-in, and who owns the result.
In plain terms
An AI model is the trained artifact at the center of every AI feature, the part that learned from examples and now turns new inputs into answers, scores, or predictions.
An AI model is the trained part of an AI system. During training it studies examples and adjusts itself until its answers match the pattern in those examples. After training, the model is a finished thing you can run: give it a new case and it produces the answer it learned to give.
That is why two projects using 'the same AI' behave differently. One team rents a large general model and points it at their documents. Another trains a small model on their own records. The word model covers both, so the useful questions are always the same: whose model is it, what did it learn from, and what does it cost each time it runs.
Think of it this way: A model is like a trained employee's judgment, written down as numbers. Training is the apprenticeship. The model is what walks onto the floor afterwards and does the job on new cases.
A lender talks about 'our model' in two senses in one meeting: the provider's large model that reads loan documents, and their own small scoring model trained on past lending decisions. The contract prices them very differently.
A support lead asks why the assistant's answers changed overnight. Nothing in their product changed. The provider updated the model underneath it, and the new version phrases refusals differently. The fix is a version clause and a test set, agreed before the contract is signed.
Use the word deliberately in any AI conversation. Ask which model, whose it is, what it was trained on, and what happens to your feature if that model changes or is withdrawn. Do not let 'the model' stand in for the whole system. The model is one part. The documents it reads, the checks around it, and the screen your team uses decide whether it helps anyone.
You do not need to pick a model from this page. RaftLabs starts from the job: the documents, the decision, and what a wrong answer costs. You see a working version before a large build. The related work on our side is LLM integration.
This sits with the other foundations terms on the glossary. The words under every AI conversation, from the model itself to the data it learned from. Worth reading next: Artificial Intelligence (AI), Machine Learning (ML), and Large Language Model (LLM).
Frequently asked questions
Straight answers to what buyers ask before they scope. Still stuck? .
- No. The model is the trained brain. The tool or app is everything around it: the screen, the documents it can read, the rules it follows, and the log of what it did. You buy a tool. The model inside it is one line of the specification, and it should be named.
- Only if the contract says so, and it depends on the kind of model. A small model trained on your data can be yours. A large rented model never is: you own your documents, your prompts, and your test cases, and you rent the model by use. Know which parts are yours before you build on them.
Work with us
Tell us what's broken.
Tell us what's not working in your business. We'll find the real problem and tell you exactly what it would take to fix it.
- Scope and cost agreed before work starts. No surprises. No obligation.
- Working prototype within 3 weeks of kickoff.
- Pay by milestone. You see progress before each invoice.
- 60-day post-launch warranty. Bug fixes, UI tweaks, and deployment support. No retainer.
- All conversations are NDA-protected.