AI data RFP template

An AI data RFP is a request for proposal that describes the data you need and asks vendors to bid on price, quality and delivery. A good one gives every vendor the same specification and the same test, which makes bids comparable. Use the sections below as a template.

What should the RFP say about scope?

State the goal, the data type, the volume, the languages or domains and the timeline. Say how the data will be used, so vendors can price rights correctly.

How do you write the specification?

Describe the task, input format, output format and examples of acceptable and unacceptable work. Attach a rubric and ten to twenty worked examples. Vague specifications produce vague bids.

What acceptance tests should you include?

Define how you will accept a batch: sample size, who scores it, the pass threshold and the rework rule. Where possible, use a held-out set you control. Tie payment to acceptance.

What do you ask about rights and provenance?

Ask for ownership warranties, permitted uses, indemnity and provenance records. The licensing checklist lists the clauses.

What do you ask about security and people?

Ask how contributors are vetted, how fraud is detected, where work happens and how your data is protected. See the vendor evaluation checklist.

How should vendors present pricing?

Require a price per accepted unit, a fee schedule and volume tiers in a common table.

Item	Vendor to fill in
Unit price per accepted example
Review and management fees
Rework policy
Minimum volume
Rush surcharge

How will you evaluate bids?

State weights up front, for example quality 40%, price 25%, security 15%, speed 10% and references 10%. Shortlist two or three vendors for a pilot.

Can you run the RFP as an auction?

If specification and acceptance tests are tight, a reverse auction lets qualified vendors bid the price down. Without strict tests, the lowest bid tends to deliver the weakest data, so keep price as only one criterion.