RFP evaluation for consultancies advising buyers
Vendor selection your client can defend
Reqify Tenders runs a client’s RFP from the delivery project: the RFP drafted from requirements the project already has, vendors’ response workbooks read back by requirement, every claim checked against the vendor’s own documents, and an award decided by people with blind scoring and a full record.
- Blind scoring on by default
- AI suggests, people decide
- Vendor files kept apart
- Append-only audit log
The hard part of an RFP evaluation is the record, not the spreadsheet
When a consultancy helps a client choose a platform or supplier, the evaluation usually lives in a scoring spreadsheet, a shared drive of vendor proposals and a lot of email. It works until someone asks how the winner was chosen: who scored what, whether anyone saw others’ scores first, why a mandatory failure was waived, and whether the vendor’s claims were ever checked against what it actually submitted.
Writing the RFP has its own trap. Requirements are retyped from the BRD and drift from the wording the client approved. Internal notes — budgets, open questions, what a stakeholder thinks — leak into vendor-facing text.
And reading vendor responses is slow: hundreds of rows of “Comply”, each one a claim somebody should check against the proposal it came with.
How Reqify Tenders runs the selection
A tender runs inside the client’s delivery project. Requirements are imported from the project’s BRD and FRD in their own words, numbered T-001 onwards, minus internal notes vendors must never read. Only the RFP’s background and scope are AI-written, from the project’s documents; a drafted paragraph that mentions money, internal risks, stakeholders’ views or anyone by name is held back as a question for your team. You download the RFP (Word), the response workbook (Excel) and each addendum — all in the buyer’s name — and send them yourself.
Vendors return the workbook and Reqify reads their answers back by T-number: the claim, the response text and the price. Assess with AI checks each claim against that vendor’s own documents — Supported, Partly supported, Unsupported claim, Contradicted or Not addressed — with quotes verified word for word, and suggests a score per criterion with its reasons, labelled as a suggestion.
People decide. Evaluators score every criterion themselves, blind by default; editors agree a consensus with the evaluators’ scores and the AI’s suggestion side by side; results are computed from the agreed scores only. An AI flag on a mandatory requirement waits for an editor to confirm or dismiss it with a reason, and awarding to anyone but the top-ranked vendor needs a written rationale. The decisions are written to Reqify’s append-only audit log, and the evaluation report is composed from that record.
Built for a defensible procurement
RFP from the BRD, wording kept
Requirements imported from the project’s BRD and FRD, numbered T-001 onwards, with internal notes stripped before vendors see them.
Claims checked against evidence
Each vendor claim checked against that vendor’s own documents, with verified quotes behind every verdict.
Blind scoring
Nobody sees others’ drafts; evaluators see submitted scores only after submitting their own. Early reveals are logged.
People decide the award
Results come from agreed consensus scores only. The AI never scores, ranks or fails a vendor on its own.
Vendor files kept apart
Vendors’ files never enter the project’s knowledge base, and are shown only to the tender’s editors and evaluators.
An evaluation report from the record
Composed from the recorded evaluation, with AI observations in their own section marked as not used for scoring.
From requirements to award
- Step 1
Set up the tender
Client, reference, one time zone for every date, rated and pass / fail criteria with weights, a price method and evaluators.
- Step 2
Draft and issue the RFP
Requirements imported from the BRD and FRD; issue checks that weights total 100%, the deadline is set and no SME ask is open.
- Step 3
Read vendor responses
Completed workbooks read back by T-number; vendors’ claims assessed against their own documents.
- Step 4
Score and agree
Evaluators score blind; editors agree a consensus; mandatory failures and waivers are decided with reasons.
- Step 5
Award and report
Award a ranked vendor and generate the evaluation report your client can take to its governance.
Who it is for
- Consultancies that help clients choose a platform, product or supplier
- Advisory teams running software selections before an implementation
- Programme leads who need the selection to stand up to governance and audit
- Firms that want the selected vendor’s requirements to flow straight into delivery
RFP evaluation FAQ
Does the AI score or rank the vendors?+
No. It checks each claim against the vendor’s own documents and suggests a score per criterion with its reasons. Evaluators score every criterion themselves, the panel agrees a consensus, and results are computed from the agreed scores only; the AI’s suggestions sit beside them for reference.
Does Reqify send the RFP to vendors?+
No. Reqify doesn’t email vendors or host a vendor portal. You download the RFP, the response workbook and each addendum — all in the buyer’s name — and send them yourself.
How is price scored?+
Choose a price method per tender: lowest price divided by the vendor’s price, linear between the lowest and highest, or not scored. Rated criteria weights must total 100% before the tender can be issued.
Can the decision be defended later?+
That is the point of the record. The evaluation starting, each evaluator’s submission, reopened scores, revealed blind scores, consensus changes, waivers, the shortlist and the award are written to Reqify’s append-only security audit log.
What does it cost?+
Setup, importing requirements, the workbooks, scoring, consensus and results use no documents. Each AI run uses 1 document — drafting the RFP, assessing one vendor, or writing the evaluation report — and a failed run is refunded. Professional includes 1 active tender, Team 5 per workspace and Enterprise unlimited.
Run the next selection from the project
Start from the BRD you already have, and finish with an award your client can explain.