Manufacturing & Industrial Equipment
Answer every requirement in the customer's spec, with the evidence quoted
Built for bid and proposal engineers and project managers who answer tenders and RFQs. They work at manufacturers of valves, actuators, pumps, instruments, and rotating and pressure equipment. Upload the customer's specification package and your own datasheets. OdysseyGPT works through the requirements row by row. It pairs each one with the documents that can answer it and quotes the evidence. Then it drafts the compliance matrix and the deviations & clarifications list for your engineers to review.
Last reviewed:
Industry context
A tender, RFQ or enquiry for engineered equipment rarely arrives as one document. The customer's package bundles project specifications, a material requisition, an inspection and test plan (ITP), datasheets and the standards they call up. It is often hundreds of pages and several spreadsheets. The bid engineer or project manager has to answer every requirement against the company's own datasheets and documents. Then they write the deviations and clarifications that go back with the offer, before the EPC's technical bid evaluation (TBE). By hand, this is slow and depends on who does it. A requirement nobody answered at bid stage tends to come back after the order as a re-test, a non-conformance or a back-charge.
Upload the customer package and your reference documents. OdysseyGPT reads each document to work out what it is (specification, requisition, ITP, datasheet or standard) and how it is laid out. It decides from the content, not the file name, and asks you to confirm when it isn't sure. It then works through the customer's requirements row by row and keeps a ledger of every source row. Each row is reviewed as a requirement or set aside with a stated reason (a header or boilerplate, say). The review tells you about any row that isn't accounted for. Each requirement is compared only with documents that can answer it. Sometimes a requirement needs a document the package doesn't contain, such as an ITP that was never sent. Then it is flagged as a document to request. Every finding carries a quote from the source. The quote is checked against the source before the finding is presented as settled. Your engineers work through the open items, edit the wording, and export the compliance matrix and the deviations & clarifications list.
Key Takeaways
- Every source row accounted for: Every row of the customer's documents is reviewed as a requirement or set aside with a stated reason, such as a header or boilerplate. Any row that isn't accounted for is listed in the review.
- Quoted, checked evidence: Each finding quotes the text it relies on, with its document, page, sheet or cell. The quote is checked against that place in the source. If the quote can't be found there, the finding goes to review.
- Paired with the right documents: Requirements are compared only with documents that can answer them. A requirement that needs a document the package doesn't contain is flagged as a document to request.
- A list you can send: An editable compliance matrix and a deviations & clarifications list for the customer, with your internal to-dos kept on a separate sheet.
Questions buyers ask
Who is this built for?
It is built for bid, proposal and application engineers, and for project managers. They work at makers of engineered equipment: valves, actuators, pumps, and rotating and pressure equipment. These teams answer specification packages from EPC contractors and end users. At smaller valve and pump makers, project managers often answer the RFQs. It works on the supplier side of the enquiry: it checks the customer's requirements against your own datasheets and documents.
Which documents does it read?
From the customer: project and technical specifications, material requisitions, inspection and test plans, datasheets and the standards they reference. From you: your datasheets, standard specifications, test procedures and the other documents you answer from. It reads PDF, Word (.docx), Excel (.xlsx, .xls) and CSV files, including multi-sheet workbooks. Scanned pages are read from the page image where possible. A page it can't read is reported as unreadable, not skipped. Engineering drawings such as general arrangements and P&IDs aren't interpreted, so review those yourself.
How is the evidence checked?
Every finding quotes the text it relies on, with its document and its page, sheet or cell. Before the finding is presented as settled, that quote is checked against the cited place in the source. Sometimes the quote isn't there, or it is found somewhere other than where the finding says. Then it doesn't pass, and the finding goes to review instead.
What does the output look like?
A compliance matrix with one line per customer requirement. Each line shows its status, where it sits in the customer's documents, the quoted evidence from yours, the basis for the status and the reviewer's decision. Alongside it comes a deviations & clarifications list for the customer. It holds the exceptions you are taking, your questions, and the documents you need them to send. Internal to-dos, such as values your team still has to confirm, sit on a separate sheet and stay off the customer list. Both are spreadsheets you can edit before anything is sent.
What happens when a requirement can't be settled?
It is shown for review with the reason. For example, the evidence is ambiguous, the quote couldn't be verified, or the requirement needs a document the package doesn't contain. If any source rows weren't accounted for, the review lists them. OdysseyGPT doesn't send anything to your customer. Your engineers decide the open items and own what goes out.
Does it work for tenders as well as RFQs?
Yes. The customer may call it a tender (the usual word in the UK, Europe and the Gulf), an RFQ or an enquiry. Either way, the package is the same kind of thing: project specifications, a material requisition, an ITP, datasheets and the standards they call up. OdysseyGPT reviews it the same way and produces the same compliance matrix and deviations & clarifications list.
What status does each requirement get?
Complies, deviates, clarification, not addressed, document requested, not applicable, or out of scope. Clarification means the requirement is unclear and needs a question back to the customer. Not addressed means your documents don't answer it yet, so your team has to confirm. Document requested means the evidence has to come from a document the package doesn't contain. Every status comes with the quoted evidence or the stated reason behind it.
Can it check a valve datasheet against API 6D or another standard?
It checks each requirement against the documents in the review. Say a valve specification calls up API 6D, and the standard or your API 6D documentation is uploaded. Then requirements are compared against it with quoted evidence. If a requirement needs a document that isn't there, it is flagged as a document to request, not guessed.
How does this help at technical bid evaluation (TBE)?
During the TBE, the EPC's evaluator checks your offer against their specification. They raise technical queries (TQs) for anything unanswered or unclear. If you answer every requirement with quoted evidence, and state your deviations and clarifications up front, fewer points are left open to query. OdysseyGPT works on your side of the tender only. It doesn't see or score the EPC's evaluation.
How is our data handled?
Your documents are uploaded to your organization's workspace. They are processed to produce the reviews you ask for. Data is encrypted in transit with HTTPS (TLS). Our hosting and storage providers encrypt data at rest. Cloudflare R2 encrypts every stored file with AES-256. OdysseyGPT is hosted on Render in the United States (US-East, Virginia). You can delete documents and workspaces. Our privacy policy covers retention and your rights. OdysseyGPT is a managed cloud service. Customer VPC and on-premises deployments are available on the Enterprise plan, in your chosen configuration.
