Delvops

Custom AI integration for your business

AI wired into your data and your processes, with a measurable scope — not a demo that impresses and changes nothing.

Typical timeline : 3 to 6 weeks

Built with

  • OpenAI
  • Anthropic
  • Google Gemini
  • PostgreSQL
  • Node.js

Pricing

Setup

from€3,500

Scoping the use case, choosing the model, building it, and measuring the quality in numbers.

Maintenance

€250/ month

Hosting, tracking answer quality, and keeping up as models change.

Free · No commitment · Reply within 24 hours

The problem

Everyone talks to you about AI; nobody tells you what it would concretely change in your working week.

Who it's for

  • Businesses processing a large volume of documents or forms
  • Teams losing several hours a week to one repetitive task
  • Projects where a demo impressed but never reached production

Scope

What's included, and what isn't

Stating the limits up front spares both sides an unpleasant surprise later.

Included in the setup

  • Scoping: the use case we keep, and above all the ones we set aside
  • Model choice argued on cost-to-quality, not on fashion
  • Retrieval over your documents (RAG): chunking, indexing, retrieval
  • A measured evaluation on real cases — you know what it is worth before deploying
  • Cost control: caps, caching, fallback to a cheaper model
  • Technical documentation and handover

Included in the maintenance

  • Hosting and monitoring of the pipeline
  • Monthly quality tracking against the reference case set
  • Adapting to new model versions without regressions
  • An alert if inference costs start drifting

Not included

  • Inference costs beyond the agreed allowance, passed through at cost
  • Building or cleaning a dataset that does not exist yet
  • Training a model from scratch (rarely justified, quoted separately if needed)
  • Any contractual guarantee of an accuracy rate

FAQ

Frequently asked questions

What is RAG, in practice?

Retrieval-augmented generation means finding the relevant passages in your own documents first, then letting the model write only from those. That is what lets AI answer about your business without inventing, and without retraining anything.

What does the AI cost to run?

It depends on volume and model. I produce a monthly estimate during scoping, set caps, and the maintenance fee includes a usage allowance. Beyond that, it is passed through at cost, with no margin.

How do I know whether the result is any good?

By measuring it. We build a set of real cases with their expected answers, and I publish a success rate before going to production. Without that number, an AI project rests on an impression.