AI in procurement: technical comparisons in hours instead of days
In hardware product development, technical decisions and procurement overlap: manufacturer data sheets, suppliers, prototypes, correspondence. Our agents turn the sources into comparisons and drafts, and a human decides and sends.
the problem
In a pilot project around a hardware product, technical and procurement work draws on many scattered sources: manufacturer specifications, public materials, emails, shared conversations and even AI brainstorming of more than a hundred pages. A small advisory team or an engineer takes long notes by hand, and it is easy to lose track of who is waiting for what and since when there has been no reply.
how we work
- The agent collects the material: manufacturer specifications and public sources, emails, shared conversations and documents.
- It compares the options: the level of integration, the suppliers and the route to a prototype.
- It delivers a summary, a decision table, an offer plan and a critical review, together with next steps. In Hungarian, as a branded PDF or an interactive page.
- It prepares letters, requests for quotation and attachments as drafts. Material for the manufacturer is written in English, with partner identifiers masked.
- For open threads it tracks who has to move and since when there has been no reply.
- It works on request, a few times a week. Several agents work in a shared group with divided tasks.
where a human decides
- Every outgoing document is a draft. It is sent only after human approval, an explicit yes.
- Before sending, the agent tells its teammate and checks that no draft already exists for the same message.
- The decision, meaning which route and which supplier we choose, is made by a human.
results
- A multi-source technical comparison is ready in hours instead of days.
- A hundred-plus-page AI brainstorm or a manufacturer data package becomes a structured, forwardable document within a few hours, with a recommendation for every problem.
- English business letters are produced faster and more consistently. There is less manual typing, note-taking and file hunting, and decisions are faster.
lessons
- When several assistants work on the same thread, two drafts can appear for the same message. So we clarify in advance who is responsible for what.
- We read AI brainstorming critically, because it tends to agree. We look separately for the missing costs, permits and the customer side.
- We keep confidential internal reports separate from the ones that go out.
- A check before sending and clear areas of responsibility keep the same message from going out twice.
want something similar?
If you want agents to speed up the technical and correspondence side of your procurement, talk to oiSystem.