Public sector tender software, and the rules that govern it

The audit trail is the product

In commercial bidding an audit trail is good discipline. In public procurement it is the thing that defends the award. Who saw what, when, on what basis a score was given, and whether every bidder was treated the same — these have to be recorded as the work happens, because they cannot be reconstructed afterwards.

That shapes the design. State changes are recorded rather than overwritten, scores carry their justification, and nothing important happens without an actor and a timestamp.

Published criteria constrain everything downstream

The evaluation criteria and their weightings are published before responses arrive, and they cannot move afterwards. Software that lets an evaluator invent a new consideration late in the process is not a convenience, it is a liability.

The practical consequence is that the scoring model should be configured up front, locked, and visible to the people using it.

Frameworks, thresholds and the shape of the process

Whether a competition is above or below threshold, run through a framework, or part of a dynamic purchasing system changes which steps are mandatory and which are optional. A system that models one route and bends for the others produces non-compliant processes quietly.

Rules differ by jurisdiction and change over time. Any serious build starts by writing down which regime applies and getting that confirmed by whoever is accountable for compliance, rather than assuming a general model.

What good looks like

• Criteria and weightings configured before responses are opened
• Scores recorded with their justification, not just a number
• Every state change attributed and timestamped
• Equal-treatment controls on what evaluators can see and when
• Process route modelled explicitly, not assumed
• Records retained in a form that survives a challenge

Questions

Does this handle our jurisdiction's procurement rules?

Only if they are written into the build. Procurement regimes differ by country and change, so we start by establishing which rules apply and who signs off that interpretation. We do not assume a general model fits, and we are not your legal advisers on it.

Can AI score public tender responses?

It can help evaluators by extracting, summarising and checking completeness. It should not produce the score. A public award has to be defensible by a named person who can explain the reasoning, and "the model said so" is not a defence.

The rest of this subject

• Tender management software: the platform we built and run
• AI Tender & Bid Management Software Development
• Bid management software for teams that actually bid
• Tender response software, from documents to a submitted answer
• Tender tracking software, and finding the ones worth bidding
• E-procurement software, and where custom development fits
• Tender management software for pharmaceutical companies