RFP software for government, built for the public record

A private company can quietly change course on a purchase. A public agency cannot: its solicitations are read by elected officials, auditors, local reporters, residents and, most closely of all, the vendors who did not win. The question is rarely only “did we pick a good vendor?” but “can we show that every vendor was treated the same?”

Pactwise Source answers that question by construction. Where it can, it removes the unfair move entirely instead of asking staff to remember not to make it; where an exception is legitimate, it requires a written reason and records it. The table pairs the pressures public buyers describe with what Source does about each.

A basic fairness rule in public purchasing is that no one sees a price before the deadline. In Source that rule is cryptographic, not procedural. Each pricing submission is envelope-encrypted with its own data key, wrapped by a key the database never holds. The buyer sees that a bid exists — never its contents — until the reveal.

A scheduled job opens every envelope at once after the deadline plus grace, re-checking the effective deadline at run time so an extension can never trigger an early reveal.

After publish a deadline can only move later, every change is historized with a reason, and every vendor is notified at once. Lateness is frozen at submission time.

If your rules call for a public bid opening, Source does not replace it. What it adds is proof that no one — not the buyer, not an administrator, not anyone holding a database copy — could have seen pricing before the time you announced. For context, the federal Uniform Guidance that governs purchases made with federal grant funds says sealed bids are opened at the time and place prescribed in the invitation for bids, and publicly for local governments (2 CFR 200.320(b)(1)). Read how a sealed bid works, and how it differs across an RFP, RFQ and RFI.

Protests tend to turn on the same few questions: who saw the pricing early, why one vendor got more time, whether everyone received the same information, and where the scoring rationale is. Source is built so each of those has a structural answer.

• One deadline for everyone. There is no per-vendor deadline anywhere in the schema. Every extension reaches every vendor at the same moment.
• The same answers for everyone. Vendors ask privately; answers publish to everyone, anonymized. Post-publish changes go out only as versioned addenda every vendor must acknowledge, and an addendum inside the notice window forces a deadline extension or a written override.
• No private side channel. Questions arrive privately; answers publish to everyone, anonymized. There is no side channel to favor anyone through.
• Consistent regret letters. Regret letters come from templates, so no one improvises a reason after the fact.

When a protest does arrive, there is one place to look. One export compiles the timeline, addenda and acknowledgments, published Q&A, submission metadata, locked scores, attestations and approvals into a single hashed document. See what a bid protest is and how the record is used to answer one.

Evaluators attest no conflict of interest before scoring, see only their own scores until lock, and outliers are flagged at lock time. Blind means each evaluator scores alone: until the evaluation locks, no one sees anyone else’s numbers, so the first score entered cannot anchor the rest of the committee.

AI never scores, ranks, or recommends an award.

That is a structural rule, not a setting. The publish gate will not let an RFP go out unless its criteria weights sum to exactly 100, and after publish the criteria change only through an addendum every vendor must acknowledge — so the evaluation rubric vendors read is the one evaluators use. More on blind evaluation.

If you spend federal grant funds, the Uniform Guidance asks recipients to maintain written standards of conduct covering conflicts of interest, and to identify every evaluation factor and its relative importance in a request for proposals (2 CFR 200.318(c)(1), 200.320(b)(2); states follow their own policies under 200.317). Source supports your process for applying those standards: the attestation, the published weights and the locked scores all sit on the record.

A records request about a solicitation usually means someone reconstructing months of email. In Source the record is built as the work happens. One write path into the audit log, no updates, no deletes, each event hash-chained to the one before.

No one — vendor or buyer — reads a file directly. Every read is logged and served through a signed link that expires in 60 seconds.

Uploads are stamped with the server clock and a server-computed checksum, scanned, then frozen. Late uploads are accepted only with a written reason and flagged as late. Uploaded files pass a default-deny structural inspection before a buyer can open them.

Recipients of federal awards must keep records sufficient to detail the history of each procurement, including the rationale for the method, the selection or rejection of contractors and the basis for the price (2 CFR 200.318(i)). Source supports your process for keeping that history; what is released or withheld under your public-records law remains a decision for your records officer and counsel. Read more about the procurement audit trail.

An award needs a locked evaluation, a written justification, every vendor question answered, every abnormal-bid flag acknowledged, and an approver who is not the author.

The two-person rule inside Source does not replace a council or board vote. It means the recommendation that reaches the agenda already rests on a locked evaluation, carries a written justification and a score snapshot, and has been approved by someone other than its author — and the approver cannot edit the justification they approved.

If fewer than two submissions arrive, the award gate also asks for a documented limited-competition justification before the award can be made.

Many agencies run solicitations with one or two people who also handle contracts, payables or grants. Source puts the checklist into the database so a busy office does not have to carry it. Publishing is a database check, not a checklist: contact details, Q&A cutoff spacing, minimum response window, criteria weights that sum to 100 and a pricing schedule are all enforced before an RFP can go out, and it warns when fewer than three vendors are lined up to invite. See how the publish gate works.

Prep Studio, “Draft with AI” in the buyer console, is the drafting help on the way. AI-assisted intake that drafts the scope, requirements matrix, weighted criteria and pricing schedule, reads its own draft as a hostile bidder would, and hands every artifact to you for approval. Built and verified against recorded model runs; live-model drafting switches on with the model key. It drafts; deterministic code validates; a person approves each artifact. Nothing reaches a vendor without that approval.

Vendors work on a separate host with its own sessions; a vendor identity can never become a buyer user. Single-use invitation links that only an explicit click consumes (corporate link scanners cannot burn them), plus step-up re-authentication before sensitive actions.

One subscription covers Pactwise Source and Pactwise Contracts.

• Compare the plans on the pricing page.
• See how the AI RFP Studio drafts, validates and seals a package on the product page.
• Compare a solicitation run by email and spreadsheet with one run in Source: manual RFP vs Pactwise Source.
• Buying for schools? Read RFP software for K-12 school districts.