An RFP generator that learns from what past bidders asked, and lets you decide every change
Half the bidder questions, because every gap past bidders asked about is closed first.
See one case, screen by screen ↓RFP Creation is an RFP generator for sourcing teams: it drafts RFIs, RFPs and RFQs from a plain request, your templates and your past events. Eight agents gather the sources, draft with every paragraph tagged to where it came from, ask the right people about each gap and keep scope, price and terms in step. The sourcing manager accepts every change, reviewers approve with reason codes, and nothing goes to a supplier until a person releases it.
How to choose AI solutions for back-office and enterprise work →
Updated
Most RFPs start as a copy of the last one. The sourcing manager finds the previous RFP in the category, hunts for the current template and the latest approved clauses, and then chases the requester, Legal, information security and data protection by email for the facts the copy does not have: volumes by country, statutory filings, interfaces, service levels, where the data may sit. The price sheet and the evaluation criteria are rebuilt by hand, often from the last event too.
Whatever the RFP leaves vague comes back later. Bidders ask, and the deadline moves; a price sheet without volume bands gets bids that cannot be compared, and the event is re-tendered; a scope line nobody wrote down becomes a change order after award. At Kestrel & Vane, the 2023 payroll RFP left Mexico's IMSS and INFONAVIT filings vague. Three of five bidders asked, the answer was “per local law”, and after award the supplier charged €96.0k a year for them in change order CO-2024-03.
Estimated split for a strategic services RFP, by hand and with the solution.
Eight agents gather, draft, check and ask; the sourcing manager decides every change and releases.
Kestrel & Vane Group needs a new payroll provider for 6,230 people in Mexico, Brazil, Poland and Spain. Marta Lindqvist, the sourcing manager, works on it with eight agents in one conversation beside the RFP they are writing. Here is what happens, screen by screen, in the working solution.
Daniela writes her need in plain words: four countries, about 6,200 people, the incumbent's contract ends 31 Mar 2027, Brazil has had 3 late eSocial filings this year, budget around €1.4M a year. A minute later the intake agent proposes an RFP: €1.42M a year is above the €100k quote limit in the sourcing policy, and five years come to €7.10M, so the procurement director approves it. It spots REQ-2026-0416 from HR Brazil as the same need and suggests merging it.
Marta, 17:40: “Accepted. Draft it from template v7 and our last payroll events. Ask Daniela directly for anything missing; I want it out on Friday.”
The gatherer brings 8 sources: services RFP template v7, three past HR events with their bidder questions, the approved clause library 2026.2, the 36-point completeness checklist, the incumbent contract and the HR system's headcount. The drafting agent assembles v0.1: 9 sections, 64 paragraphs — 25 from the template, 16 from past events, 9 from the request and the HR system, 9 library clauses and 5 it wrote, marked for review. One click on a tag opens the passage it came from, here page 7 of the 2023 payroll RFP.
Drafting agent: “Where a fact was missing I did not guess: the completeness agent asked.”
The completeness agent tests the draft against the 36 checks and matches each gap to the questions past bidders sent. It asks Daniela four questions and Joost van Rijn, the data protection officer, one — whether Brazilian payroll data must stay in Brazil — each in plain words, with the reason and a Friday due date. Q1 asks whether Mexico's IMSS, INFONAVIT and state payroll tax filings are in scope.
Q1: “3 of 5 bidders asked this in RFP-2023-HR-007. Left out, it became change order CO-2024-03: €96.0k a year.”
The clarity agent compares the draft with 64 past bidder questions and flags 7 phrases and numbers. “The supplier shall process payroll in a timely manner” drew two questions in 2023; the rewrite says results are delivered by 12:00 local time, three working days before each pay date. Marta accepts four rewrites (v0.2). Three stay open: year-end extras, parallel runs, and Schedule 5 listing one Spanish entity where 2.1 lists two.
The clarity agent never changes meaning without approval: a rewrite is a proposal.
Daniela replies in the thread: “Yes: IMSS, INFONAVIT and the state payroll tax, for both Mexican entities. And the provider must hold REPSE, we were caught by that in 2021.” The completeness agent turns it into paragraph 2.2.4, cited to her answer, and shows what it closes: check S3, the three IMSS questions from 2023, and REPSE as pass or fail requirement M4. Marta has three choices: Add to scope, Edit first, or Not now with a reason.
Marta adds 2.2.4: completeness 32 of 36, version v0.3. She writes “Added. Make sure bidders price it.” The pricing builder has already found the gap — 2.2.4 is a scope line with no price line, which is how the 2023 change order happened — and proposes P3.4: IMSS, INFONAVIT and state payroll tax filings for Mexico, priced per entity per month. She accepts it.
Pricing builder: “Left like that, bidders price it differently or not at all; that is how CO-2024-03 happened.”
The price schedule has 16 lines in 6 groups, plus P3.4: per employee per month by country and pay frequency, off-cycle runs, year-end extras, one-off implementation and interfaces, changes and exit. From the 2025 Poland re-tender, prices sit in two volume bands: up to 110% of today's 6,230 employees (6,853) and 110% to 125% (7,788). The internal estimate of €1.42M a year stays out of what bidders see.
The pricing builder never puts a target price in the bidder workbook and never reuses a past bidder's prices.
Nine clauses from library 2026.2 are placed and locked, from data processing under GDPR Article 28 to liability of at least €5M and payment at 60 days. Daniela asked that the supplier pay statutory penalties in full, without a cap; that conflicts with the liability clause, so the clause agent sends it to Owen Achterberg with the library alternate CL-LIA-05. The buttons stay locked for everyone but Legal.
Clause agent: “I cannot accept a deviation; Legal decides.”
Six people approve their part: Daniela the scope, Joost data protection, Ilse Brandt information security, Owen the legal terms, Rahul pricing and evaluation, and Catherine Dufresne, procurement director, because €7.10M is above €5M. Ilse approved section 4 and the security questionnaire at 09:36 with AP-01, “Approved as drafted”. The release gate lists everything still holding the RFP, by name.
Only the sourcing manager schedules the release, and it stays held until every approval lands.
Seven suppliers are invited, the incumbent included, each checked against the supplier master: one needs light onboarding first, one has no REPSE found for its Mexico partner and is tested at bid, and one covering only Poland is not proposed. Before release, a checklist confirms the approvals, the locked terms, a 12-day question window and a 21-working-day bid window. Release now stays disabled until every check passes; bids are due Mon 16 Nov, 12:00 CET.
“Nothing reaches a supplier until you release.”
The same payroll RFP, seen by four of the people who carry it — what it took before, and what it takes now.
Eight agents gather, draft, check and ask; the sourcing manager decides every change and releases.
Reads each plain request and proposes an RFI, RFP, RFQ or a call-off from an existing framework under the sourcing policy thresholds; spots duplicate requests and missing basics such as budget or need-by date. Before release it checks each invited supplier against the supplier master.
Finds the template, past events in the category with their bidder questions, change orders and lessons, the incumbent contract and scorecard, and the headcount from the HR system.
Assembles the RFP section by section from the template, past events and the request, and tags every paragraph with its source: template, past event, request, library or drafted.
Tests the draft against the 36-point checklist, asks the person who can answer each gap with the reason, reminds them on the agreed response time and proposes their answer as a cited paragraph.
Finds the phrases past bidders questioned and the numbers that disagree between sections, and proposes a rewrite with the past questions as evidence.
Places the approved clauses for the category and risk, locks them in the draft and routes any requested deviation to Legal with the library alternate.
Builds the price schedule from the scope, keeps every scope line priced, sets volume bands tied to today's headcount and prepares the workbook bidders fill.
After release, groups duplicate bidder questions, drafts answers from the RFP and the requester, rewrites any question that names a bidder, and prepares one addendum for all bidders.
Accept, approve with reason codes, release. The agents propose; a named person decides.
Ask about any paragraph, question or approval in plain words, or give an instruction. Every answer points to its source, and every change waits for you.
Why did the agents put IMSS filings in the scope?
Because it cost us before. In the 2023 payroll RFP, 3 of 5 bidders asked whether IMSS and INFONAVIT filings were in scope and were told “per local law”. After award the supplier charged €96.0k a year for them in change order CO-2024-03. Daniela Ortiz confirmed this morning that they are in scope, and that the provider must hold REPSE.
What would bidders still ask about this draft?
From the 64 questions bidders sent on 3 past events, the agents expect 9 on this draft. Still open: who sends the time-and-attendance file, asked 4 times in 2024; how many parallel runs there are, and whether they are charged; and whether PTU is part of year-end processing.
Make the bid window 25 working days
Policy needs at least 20 working days; today the bid window is 21 (16 Oct to 16 Nov). 25 working days moves bids due to Fri 20 Nov, 12:00 CET and award to Thu 10 Dec. It changes paragraph 9.1 and the portal dates. Shall I make the change?
Is this RFP ready to release on Friday?
Not yet. 1 of 6 approvals is in; the procurement director must approve because five years come to €7.10M, above €5M. Also open: questions to Daniela and Joost, the remaining clarity rewrites and deviation 8.5 for Legal. Release stays held until every approval lands; nothing reaches a supplier before Marta schedules it.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds, windows and agent rules every RFP follows are settings. Each change is saved as a new version and logged.
| Setting | Default | Choose from |
|---|---|---|
| Quotes instead of an RFP below | €100k | Any amount above zero, in € thousands |
| Procurement director approves above | €5M | Any amount above zero, in € thousands |
| Question window, at least | 10 days | 10 days or more (sourcing policy GSP-4 3.6) |
| Bid window, at least | 20 working days | Any number of working days above zero |
| Indexation cap | 3% a year | Any percentage above zero |
| Mark a paragraph for review below this confidence | 90% | Any percentage |
| Reuse past events from | 2020 | 2018 · 2020 · 2022 |
| Requester answers a question within | 2 working days | 1 working day · 2 working days · 5 working days |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (APQC open standards benchmarking — median 60 days for a sourcing event, from the internal request to a signed contract; The Hackett Group 2025 Key Issues Study — procurement workload projected to grow about 10% against budgets about 1%, a 9% capacity gap; GAO B-231986 on ambiguous solicitations, which must be corrected and reissued). People, companies, suppliers, events and amounts named on this page are fictional — characters and sample data in the working solution, at Kestrel & Vane Group.
RFP Creation is an RFP generator for sourcing teams: it drafts RFIs, RFPs and RFQs from a plain request, your templates and your past events. Eight agents gather the sources, draft with every paragraph tagged to where it came from, ask the right people about each gap and keep scope, price and terms in step. The sourcing manager accepts every change, reviewers approve with reason codes, and nothing goes to a supplier until a person releases it.
It starts from what you already have: your current template, your approved clauses and the past events in the category, with the questions their bidders asked and the change orders that followed. The completeness and clarity agents use those past questions as this RFP's checklist, so the gaps that drew questions last time are closed before release.
The intake agent proposes one under your sourcing policy thresholds: three quotes below the quote limit, an RFP above it, an RFI when there is no budget or specification yet, or no event at all when an existing framework covers the need. It also spots duplicate requests and names what is missing. The sourcing manager decides.
No. Legal and security text comes only from the approved clause library and is locked in the draft. When a requester asks for something different, the clause agent sends it to Legal with the library alternate; it never edits clause text or accepts a deviation.
Every scope line is mapped to a price line and to the criterion it is evaluated under. When a new scope line is added, the pricing builder finds it has no price line and proposes one, for the sourcing manager to accept. Prices sit in volume bands tied to today's headcount, so bidders price the same units.
Yes. The Q&A agent groups duplicate questions, drafts answers from the RFP and the requester, flags the ones that need a colleague and rewrites any question that names a bidder. A person approves the answers and publishes one addendum to all bidders; the agent never publishes.
Yes. Agents propose, cite and ask; people decide. The sourcing manager accepts, edits or declines every change with a reason, six reviewers approve their parts with reason codes, the procurement director approves events above the policy limit, and the release stays held until every approval lands.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your templates, completeness checklist, clause library, sourcing policy and a sample of past events with their bidder questions — and it goes live once every quality gate has passed. We will walk you through it on one of your own requests first.
Hidden costs priced before award, and every score backed by the page it came from.
1-min demoSpend AnalyticsEvery dollar classified, and every saving pinned to the invoice lines behind it.
1-min demoSupplier Negotiation PreparationPrice increases held to what costs support, every number in the brief cited.
1-min demoWe'll run RFP Creation on one of your own requests, with your templates, clauses and past events.