CONTEXT(CLONE)
Sign in

Prepare tender evaluation evidence with two independent readings

For: Procurement officer or evaluation panel chair in a public contracting authority

Pattern: Cross-examinationNeeds live modelsDesigned for 6 to 300 agents
Get early accessSign in to use

The pain today

Evaluators must score long submissions against published criteria and defend every score if challenged. Finding where each bidder addresses each criterion takes most of the time, and notes are uneven between evaluators.

The ask

“I attached the tender documents with the award criteria and all the submissions. For each bidder and each criterion, find the passages that address it, say what is offered and what is missing against the published requirement, and quote the text. Do this twice independently and show me where the two readings differ. Do not score.”

Plain words, as you would say it to a colleague. Edit it to fit your case before you send it.

What you attach or connect

  • Tender documents with award criteria and sub-criteria
  • Bidder submissions as text
  • Clarification questions and answers

The unit of work

One worker task per one bidder against one criterion.

Why a swarm fits

Bidders times criteria gives many small, independent lookups. Two readings from different families mirror the practice of independent evaluators and expose passages one reading overlooked.

Not for

Scoring, ranking or recommending an award, or procurements where rules forbid processing submissions with external tools.

The decision tree

6 typed decisions, each with an action for every answer

At fixed moments in a run, the engine puts one narrow question to a decision model. The decision model never writes text: it answers yes or no with a probability, picks from listed options, or gives a score, about a small slice of the material. The engine then does exactly what this tree says, which is what makes the run auditable. The thresholds are the template's design values, not measured results.

  1. Planner, while planning

    Scope checkYes or no, with a probability

    Before work starts on a unit

    Does this checklist requirement appear in the published award criteria or sub-criteria, in those words or a direct restatement?

    Sees only: One checklist requirement and the published criteria text

    Why: Rejects any expectation the tender documents did not state.

    • Yes: 0.60 or higherthenAccept
    • Unsure: 0.30 up to 0.60thenEscalate to a strong model
    • No: below 0.30thenSkip this unit

  2. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Does the quoted submission passage describe what the bidder offers for this published requirement, rather than general company background?

    Sees only: One requirement and one quoted submission passage with its location

    Why: Keeps evidence packs to passages an evaluator can rely on when giving reasons.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenReject and retry

  3. Reconciler, while merging

    Conflict checkA choice among options

    While reconciling

    Did the two independent readings find the same passages for this bidder and criterion?

    Sees only: Both readings' passage lists for one bidder and criterion

    Why: Passages one reading overlooked are recovered, which protects equal treatment.

    • Same passagesthenAccept
    • One reading found extra passagesthenEscalate to a strong model
    • Readings describe the offer differentlythenMark unresolved

  4. Run control, between rounds

    Retry or stopA choice among options

    After a rejection or low confidence

    Does the extra passage found by one reading address the requirement?

    Sees only: The extra passage and the requirement

    Why: Settles differences from the submission text and keeps arguable ones for the panel.

    • Addresses itthenAccept
    • Does not address itthenReject and retry
    • ArguablethenMark unresolved

  5. After workers, the judge checks

    Evidence checkA choice among options

    After a worker answers

    Does a clarification answer change or add to what the submission says on this requirement?

    Sees only: One requirement, the pack entry and the clarification log entries that mention it

    Why: Makes sure evaluators see clarifications next to the original text.

    • No clarification on this pointthenAccept
    • Clarification adds or changes contentthenMark unresolved

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Is the entry a requirement a submission does not address, a difference between the readings, or a point touched by a clarification?

    Sees only: One evidence pack entry

    Why: The panel scores and the contracting officer awards; the swarm neither scores nor ranks.

    Accountable: The evaluation panel scores every criterion; the contracting officer owns the award decision and its reasons.

    • Yes: 0.40 or higherthenAsk a person
    • Unsure: 0.15 up to 0.40thenAsk a person
    • No: below 0.15thenAccept

The fleet: who does what

Model tiers by role, not brands: you choose the models. Strong reasoning models plan and reconcile, small fast models do the wide work, and the judge is a decision model from a different family, so it does not share the workers' blind spots.

  1. Planner

    A strong reasoning model breaks the criteria into checkable requirements exactly as published, adding none.

    Decisions here:1. Scope check

  2. Workers

    Two sets of small workers from different families each locate and summarise the evidence for one bidder and criterion.

    Designed for 6 to 300 agents, one worker task per one bidder against one criterion. Each worker receives only its own unit.

  3. Judge, from a different model family

    A decision model from a third family checks each difference between readings against the quoted submission text.

    Decisions here:2. Evidence check5. Evidence check

  4. Reconciler

    A strong reasoning model assembles an evidence pack per criterion with identical structure for every bidder.

    Decisions here:3. Conflict check4. Retry or stop

  5. Accountable person

    The evaluation panel scores every criterion; the contracting officer owns the award decision and its reasons.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Every statement about a bid quotes the submission with its location
  • Only published criteria are applied, and any unstated expectation is rejected
  • All bidders are processed with the same prompts and the same order of checks

What comes back

  • Evidence pack per criterion and bidder with quotes
  • Requirements a submission does not address
  • Differences between the two readings
  • Record of what each agent was shown, for the audit file

What to measure

  • Evaluator time to locate evidence per criterion
  • Relevant passages evaluators found that both readings missed
  • Consistency of evidence packs across bidders
  • Cost per bidder-criterion pair

Names of measures only. No result is claimed for this template.

Templates open in the workspace chat with the ask filled in. Nothing runs until you send it.

Get early accessSign in to use

More in Public sector and policy

Code consultation responses by question and keep minority views

For: Policy analyst or consultation lead at a ministry, regulator or municipality

A public consultation brings thousands of free-text responses, many from campaigns, some from experts with one decisive point.

Pattern: Map, verify, reduceNeeds scale6 decisionsDesigned for 40 to 800 agents

Trace a changed legal definition through the statute book

For: Legislative drafter, government lawyer or policy adviser preparing an amending bill

Changing one defined term can ripple through acts, regulations and guidance that use or cross-refer to it.

Pattern: Hierarchical decompositionNeeds a connector6 decisionsDesigned for 30 to 600 agents