CONTEXT(CLONE)
Sign in

Track data room uploads and Q&A answers, redo only what changed

For: Deal team associate running a live due diligence process over several weeks

Pattern: WatchtowerNeeds a connectorDesigned for 6 to 300 agents
Get early accessSign in to use

The pain today

The seller uploads new and replaced documents daily and answers questions in a separate log. Findings written last week silently go stale, and the report that reaches the investment committee cites superseded versions.

The ask

“I connected the data room and the Q&A log, and attached our current findings with the documents each one relies on. Watch for new, replaced and removed documents and new answers, tell me which findings they affect, redo only those, and show me what changed.”

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

  • Virtual data room with document versions
  • Seller Q&A log
  • Current findings register with source documents
  • Open document request list

The unit of work

One worker task per one new or changed document or answer.

Why a swarm fits

Each upload or answer is read on its own and matched to the few findings that cite it. Only those findings are reworked, so the register stays current without repeating the whole review.

Not for

A static data room reviewed once, or a process short enough that the team re-reads everything anyway.

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

    Is this upload a new version of, or a seller answer about, a document that a current finding cites?

    Sees only: The upload's index entry and the findings-to-sources map

    Why: Only findings that cite the changed document are reopened.

    • Yes: 0.40 or higherthenAccept
    • Unsure: 0.15 up to 0.40thenAccept
    • No: below 0.15thenSkip this unit

  2. Before workers, before a task runs

    Small worker or strong modelYes or no, with a probability

    Before a task runs

    Is the replacement identical in text to the prior version apart from scan quality, stamps or file name?

    Sees only: A text comparison of the two versions

    Why: Re-scans are most of the replacements and need no rework.

    • Yes: 0.85 or higherthenSkip this unit
    • Unsure: 0.50 up to 0.85thenAccept
    • No: below 0.50thenAccept

  3. After workers, the judge checks

    Evidence checkYes or no, with a probability

    After a worker answers

    Does the quoted difference between the two versions alter the term, party, date or amount the finding relies on?

    Sees only: The quoted difference and the finding's cited passage

    Why: A finding is rewritten only when its basis really moved.

    • Yes: 0.85 or higherthenAccept
    • Unsure: 0.50 up to 0.85thenEscalate to a strong model
    • No: below 0.50thenSkip this unit

  4. Reconciler, while merging

    Conflict checkA choice among options

    While reconciling

    How does the seller's answer compare with the document it refers to?

    Sees only: One seller answer and the referenced document passage

    Why: Answers that outrun their documents go back to the seller.

    • Answer matches the documentthenAccept
    • Answer says something the document does notthenMark unresolved
    • Referenced document is not in the roomthenMark unresolved

  5. Run control, between rounds

    Another round?Yes or no, with a probability

    Between rounds

    Are there uploads or answers since the last cycle that have not yet been matched to findings?

    Sees only: The unmatched items list for this cycle

    Why: A cycle ends when the change log is empty.

    • Yes: 0.50 or higherthenContinue
    • Unsure: 0.20 up to 0.50thenContinue
    • No: below 0.20thenStop

  6. Accountable person, before anything is settled

    Person decidesYes or no, with a probability

    Before anything is reported as settled

    Does the updated finding differ in substance from the text already given to the deal team or the committee?

    Sees only: The finding's old and new text

    Why: The deal team decides whether price, report or questions change.

    Accountable: The deal team decides whether a changed finding alters the report, the price or the questions put to the seller.

    • 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 keeps the map from findings to source documents and decides what each change can affect.

    Decisions here:1. Scope check

  2. Workers

    Small fast workers from an open-weight family read one changed document or answer and compare it with the version relied on.

    Designed for 6 to 300 agents, one worker task per one new or changed document or answer. Each worker receives only its own unit.

    Decisions here:2. Small worker or strong model

  3. Judge, from a different model family

    A decision model from a different family checks whether the change really alters the finding that cites it.

    Decisions here:3. Evidence check

  4. Reconciler

    A strong reasoning model updates affected findings, closes answered requests and records what was left untouched and why.

    Decisions here:4. Conflict check5. Another round?

  5. Accountable person

    The deal team decides whether a changed finding alters the report, the price or the questions put to the seller.

    Decisions here:6. Person decides

Checked before anything is accepted

  • Replaced documents are compared with the prior version, with differences quoted
  • Every updated finding cites the current document version
  • Q&A answers are checked against the documents they refer to
  • Removed documents that a finding relied on raise an alert

What comes back

  • Daily change log with affected findings
  • Updated findings, old and new text side by side
  • Document requests now answered or still open
  • Findings that cite removed or superseded documents

What to measure

  • Stale citations found in the final report
  • Time from upload to updated finding
  • Associate hours spent tracking the data room

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 Due diligence and M&A

Red-flag customer contracts for change of control and exit rights

For: M&A associate or corporate development manager on a buy-side review

The value of the target sits in its customer contracts, and the data room holds hundreds with order forms and amendments.

Pattern: Map, verify, reduceRuns today6 decisionsDesigned for 4 to 300 agents

Test every claim in the information memorandum against the data room

For: Private equity investment manager or corporate development lead before a binding offer

The seller's memorandum and management presentation make claims about customers, retention, pipeline, capacity and compliance.

Pattern: Adversarial reviewNeeds live models6 decisionsDesigned for 8 to 400 agents