Fixed-scope offer · 3 to 4 weeks
The AI Prototype Sprint is a 3 to 4-week, fixed-scope AI proof of concept. Grapefruit agrees success metrics and an evaluation set with you, builds a working prototype on your own data in your cloud tenant, and measures accuracy and hallucinations. You get a go/no-go recommendation and a production estimate.
No obligation to continue. We reply within 1 business day, usually the same day.
The promise
A working AI prototype on your own data in 4 weeks, so you decide with evidence, not slides.
Who it is for
Week by week
Week 1
We define the use case, the users, what good looks like and the evaluation set of real examples the prototype must pass. Access to data and cloud is set up.
Week 2
A first working version runs on your data in your tenant. We review it with the people who will use it.
Week 3
We run the evaluation set, measure accuracy and hallucinations, add guardrails and decide where a human reviews output.
Week 4
Test report, security and data flow note, production estimate and a go/no-go recommendation presented to your decision makers.
Every document is yours to keep and to use with any partner.
What happens after
From validated prototype to a production AI system your team trusts and runs.
Your AI keeps working, stays compliant and keeps improving.
Proof
FAQ
Typical sprints build a customer or employee assistant over your documents, an agent that completes a workflow step, a brand-voice content model, or document extraction. The use case must be narrow enough to evaluate in 4 weeks. We confirm feasibility in the first scoping call.
We agree an evaluation set of real examples in week 1 and measure the prototype against it, including how often it invents answers. We then add guardrails, source citations and human review where errors are costly. The accuracy and hallucination report is a named deliverable, not a promise of zero errors.
The prototype runs in your cloud tenant or an EU region we agree on. Your data is not used to train public models, access is limited to the named team, and a data processing agreement covers the work. The security and data flow note documents every flow.
You do. Code, prompts, evaluation sets and documentation are handed over at the end of the sprint, so you can continue with us, with your internal team or with another partner.
No. The sprint ends in a go/no-go recommendation and there is no obligation to continue. If the answer is go, the production estimate is detailed enough for your procurement to compare offers.
The sprint has a fixed scope agreed before we start. What moves the fee is the number of data sources and integrations, how sensitive the data is, and whether the prototype must run on-premises. We confirm scope and fee after a 30-minute call.
A working AI prototype on your own data in 4 weeks, so you decide with evidence, not slides. Tell us a little about your context and we will confirm scope and timing.
We reply within 1 business day, usually the same day. Not ready to talk? Take the free AI assessment.
