TL;DR: Defence tenders are evaluated against a Schedule of Technical Requirements (SOTR) that can run to hundreds of line items, across bidders who each respond with their own document structure, language and evidence. Someone has to map every claim to every requirement before a single commercial bid is opened. Doing that by hand doesn't just take time — it's where genuine non-compliances get missed and genuinely compliant bids get flagged in error.
The document-matching problem at the center of every tender
Technical evaluation of a defence tender comes down to a matching exercise: every requirement in the SOTR has to be checked against every bidder's technical response, and a finding — compliant, non-compliant, or needs clarification — recorded with the evidence behind it. Multiply a few hundred requirements by a handful of bidders, each submitting a technical bid running to hundreds of pages in their own format, and the evaluation becomes a genuinely large cross-referencing task before a single commercial bid is even opened.
On top of the SOTR itself, evaluators are checking eligibility conditions — Integrity Pact compliance for tenders above the relevant threshold (its absence is grounds for outright rejection), registration and security-clearance status, MIL-SPEC or BIS Defence Standard compliance for the equipment on offer, and whether the financial bid follows the prescribed format. Each of these is its own document-verification exercise layered on top of the technical scrutiny.
Where Technical Non-Compliance (TNC) and Preliminary Non-Compliance (PNC) findings come from
The formal output of this process is usually captured as TNC (Technical Non-Compliance) and PNC (Preliminary Non-Compliance) findings — the record of exactly where a bid falls short of a requirement or an eligibility condition. Getting this right matters in both directions: a missed non-compliance can let an unsuitable bid through, while a wrongly-flagged non-compliance can knock out a genuinely compliant bidder and invite a legitimate grievance or appeal. Both failure modes come from the same root cause — a human evaluator, working manually across hundreds of pages, missing or misreading a match between a requirement and a response.
Why this is a scale problem, not a diligence problem
None of this is a criticism of the evaluators doing the work. The problem is structural: manual cross-referencing scales linearly with the number of requirements and the number of bidders, but the time available to complete an evaluation doesn't. A tender with 300 SOTR line items and five bidders is 1,500 individual checks before any judgment calls are made — and every one of those checks has to trace back to a specific page and clause if the finding is ever challenged.
What AI-assisted scrutiny should actually do
The useful application of AI here isn't to replace the evaluator's judgment — it's to make sure nothing gets missed and every finding is defensible. That means:
- Mapping every bid claim to the exact SOTR clause it responds to, so nothing is evaluated against the wrong requirement.
- Surfacing the verbatim evidence behind every compliance and non-compliance finding, down to the page and line.
- Flagging gaps — SOTR requirements with no corresponding response anywhere in the bid — which are easy to miss by hand and straightforward for software to catch exhaustively.
- Leaving the compliance call itself to the evaluator, who confirms or overrides every AI-suggested finding, with that decision logged as part of the audit trail.
Done this way, the evaluator spends their time on judgment calls rather than page-turning, and the resulting TNC/PNC assessment report is backed by a citation for every line — which matters as much for defending a decision under appeal as it does for reaching it faster.
Frequently asked questions
What is SOTR in defence procurement?
Schedule of Technical Requirements — the document listing every technical parameter a bid must meet. Evaluators map each bidder's response against every SOTR line item to determine technical compliance.
What are TNC and PNC in tender evaluation?
Technical Non-Compliance and Preliminary Non-Compliance — the findings evaluators produce during scrutiny, flagging where a bid fails a technical requirement or an eligibility condition. Unresolved non-compliances typically exclude a bid before commercial evaluation.
Can AI replace human judgment in tender evaluation?
It shouldn't try to. AI's useful role is mapping every claim to its SOTR clause and surfacing evidence — the evaluator still makes the compliance call, with every finding traceable to a citation.
Astiron Scrutiny
On-prem AI scrutiny of procurement documents — maps every bid response to every SOTR requirement, with a cited evidence trail behind every finding.