AI that fits into your business.

1. i

   A working piece of AI inside your existing system, pointed at your own documents or your own workflow.

2. ii

   A test set built from your real examples, so you can measure how good the answers are and see when they get worse.

3. iii

   The model accounts and API keys set up in your name, so the running cost shows up on your bill.

4. iv

   A short written note on what it handles, what it gets wrong and what to do when it does.

Every business needs something slightly different, so we price the work after the call. These are the things that move it.

• How much material the AI has to read, and how messy it is.
• How the AI is used: a search box, a step in a workflow, or a reply drafted for a person to check.
• How good the answers need to be, which decides how much testing the work takes.
• What the model provider charges per use once it is live, which depends on your volume.
• Who looks after it once the first 30 days are over.

Estimated time

It depends on the project. We give you a timeline on the call.

1. i

   How much does an AI project cost?

   It depends on how much material the AI reads, how it is used and how good the answers need to be. We don’t publish a price list. On the free call we work out what you need, and you get a cost range before anything starts.

2. ii

   How long does it take?

   It depends on the project, so we give you a timeline on the call, once we have seen what the AI has to read and where it fits.

3. iii

   Can I change what the AI knows myself?

   Yes, where it makes sense. If the AI answers from your documents, adding or replacing a document is something you do yourself. Changing how it behaves is a new piece of work.

4. iv

   Who owns the system?

   You own your domain, your brand, your hosting accounts and your data. The code is licensed from CodeNXT Lab. If you want to buy the code outright, say so at the start, because it changes the quote.

5. v

   What happens after launch?

   We fix anything that breaks for 30 days after launch, free. After that, support is agreed per project. Model providers also change and retire their models, so we talk about who handles that on the call.

6. vi

   How will I know the answers are any good?

   We build a test set from your own examples and run it before launch. You can run it again whenever something changes, and the score tells you if the answers have slipped.

What we build it with

Python · PyTorch · Anthropic API · pgvector · LangGraph