ProcessLegal Services

AI conflict checks for law firms: from intake request to clearance decision

AI helps conflict checks most at the two points where they usually stall: turning a loosely worded intake request into structured parties, and matching those parties against years of client, adverse-party and related-entity records despite name variants and corporate families. It should draft and match; conflicts analysts and counsel should decide. This page walks the process stage by stage and shows which records each stage leaves behind.

Reviewed 7 min read

On this page
  1. Where new-matter intake and conflict searches lose days
  2. The intake-to-clearance path, with the human decision points
  3. Six stages, what AI does in each and who signs off
  4. What the rules require at each decision point, in England and Wales
  5. A hypothetical corporate firm opens an acquisition matter
  6. Failure modes in AI-assisted conflict searches and the control for each
  7. Questions and answers
  8. Sources

Where new-matter intake and conflict searches lose days

A partner emails the intake team: a new client wants help buying a business, the seller is a family-owned group, there may be a lender. The conflicts analyst now has to work out who the parties actually are, find every former name and subsidiary, and search them against current clients, former clients, prospective clients who shared confidential information, adverse parties and related parties on past matters. Each step involves retyping and judgement.

Three things slow it down. Intake requests arrive as free text with parties missing or ambiguous. Party names vary across the firm's own records: abbreviations, former names, trading names, individuals recorded under different spellings. And corporate families are invisible to a plain text search, so a hit on a parent company is missed when the matter names only a subsidiary. AI is useful against all three, provided the decisions stay with people.

The intake-to-clearance path, with the human decision points

party listentity setcandidate hitsreviewed hitscleared matternew party01Structured intake02Entity resolution03Records search04Hit summaries05Counsel decision06Ongoing re-checks
  1. Structured intake

    Parties, roles, matter type and related entities captured in fields, with gaps flagged back to the requesting lawyer.

  2. Entity resolution

    Each party matched to known entities, former names, parents, subsidiaries and individuals, with confidence and evidence.

  3. Records search

    Resolved entities searched against client, former client, prospective client, adverse and related-party records.

  4. Hit summaries

    Each potential conflict summarised with the source matter, the role played and the dates, for an analyst to review.

  5. Counsel decision

    Conflicts counsel clears, declines or clears subject to consent and safeguards; the reasoning is recorded.

  6. Ongoing re-checks

    New parties, lateral hires and changes in a matter trigger a fresh search against the same record.

Conceptual process for AI-assisted conflict checking. AI drafts and matches in the first four stages; people decide at review and clearance. It is not a description of any specific firm's system.

Six stages, what AI does in each and who signs off

  1. Capture the request in structured form

    An intake form or an assistant reading the requesting lawyer's email extracts the client, counterparties, other parties with an interest, matter type and jurisdiction. Missing roles are asked for before the search starts, not after a partial clearance.

    Output
    Structured intake record
    Owner
    Requesting lawyer
  2. Resolve every party to an entity

    Match each name against the firm's entity records and, where licensed, corporate registry or ownership data. Propose parents, subsidiaries, former names and connected individuals with the evidence for each link, and let the analyst accept or reject them.

    Output
    Resolved entity set with evidence
    Owner
    Conflicts analyst
  3. Search all record types, not only clients

    Run the resolved set against current and former clients, prospective clients who disclosed confidential information, adverse parties, co-defendants, lenders and other related parties on past matters, plus the firm's own business interests.

    Output
    Candidate hit list
    Owner
    Conflicts system
  4. Draft a summary for every candidate hit

    For each hit, state the earlier matter, the role the entity played, the responsible partner, dates and whether the firm may hold confidential information. Link the source records so the analyst checks the facts rather than the summary.

    Output
    Reviewed hit report
    Owner
    Conflicts analyst
  5. Decide, and record consent and safeguards

    Conflicts counsel decides whether the firm can act, declines, or clears subject to written consent and safeguards such as an information barrier. The system records the reasoning, the consents received and who is walled off.

    Output
    Clearance memo and barrier record
    Owner
    Conflicts counsel or risk partner
  6. Re-check when the facts change

    Add parties as they emerge during the matter and re-run the search. Check incoming lateral lawyers' prior matters against current work before they start, and refresh barrier membership when teams change.

    Output
    Updated clearance trail
    Owner
    Conflicts team

What the rules require at each decision point, in England and Wales

This table uses the SRA Code of Conduct for Solicitors. US firms should map the same decision points to the rules of professional conduct in force in their own state.

SituationRule in the SRA CodeCan the firm act?Record the process should keep
Own interest conflictParagraph 6.1: do not act where there is an own interest conflict or significant risk of one1No; the guidance says no exceptions apply, so consent does not cure it2The interest identified and the decision to decline
Conflict between clientsParagraph 6.2: do not act unless an exception applies1Only for a substantially common interest or clients competing for the same objective1Which exception is relied on and why
Exception conditionsParagraph 6.2 conditions1Only with informed consent in writing, effective safeguards where appropriate, and acting is reasonable1Signed consents, safeguards adopted, reasonableness assessment
Confidential information adverse to a clientParagraph 6.5: confidential information material to the matter1Only with effective measures leaving no real risk of disclosure, or the other client's informed written consent1Barrier members, technical measures, consent wording

The SRA's guidance on paragraph 6.5 describes the test for effective measures as high and lists separate teams, separate systems, access controls, conflict-flagging systems and training as examples3. This summary is not legal advice.

A hypothetical corporate firm opens an acquisition matter

Failure modes in AI-assisted conflict searches and the control for each

Over-merging distinct entities

Early signalAnalysts start seeing hits for unrelated companies with similar names.

MitigationRequire evidence for every proposed link and let analysts reject links; never merge entities silently.

Missing a related party because it was never captured

Early signalClearance is given while the intake record has empty role fields.

MitigationBlock the search until required roles are filled or explicitly marked as not applicable by the requesting lawyer.

Treating a match score as a decision

Early signalLow-score hits are closed without anyone reading them.

MitigationRoute every hit above a set threshold to a person and sample closed low-score hits regularly.

Leaking confidential details through the summaries themselves

Early signalHit summaries quote privileged content to people outside the earlier matter team.

MitigationSummarise roles and dates from matter metadata, and link to documents rather than quoting them, under the same access rules as the source.

Questions and answers

Can an AI system clear a conflict on its own?

It should not. Deciding whether a firm can act, whether an exception applies, whether consent is adequate and whether a barrier is effective are professional judgements for conflicts counsel or a responsible partner. AI is useful for preparing that decision: structuring the request, resolving entities, searching records and drafting summaries with linked sources. Design the system so a clearance cannot be recorded without a named person approving it.

Does AI entity resolution need external company data?

Not always, but it helps. The firm's own records already contain many name variants and relationships. Corporate registry and ownership data adds parents, subsidiaries and former names the firm has never recorded, which is where plain text searches miss hits. Check the licence terms for any external data, and keep the evidence for each proposed link so analysts can judge it.

How should lateral hires be handled?

Before a lateral starts, collect a list of their prior matters at the level of detail their professional obligations allow, resolve the parties and run them against current matters. Where the lateral may hold confidential information material to a current matter for an opposing party, the firm needs effective measures or informed consent before the lateral works anywhere near it, and the record should show which was used.

Which systems does an AI conflicts process have to connect to?

Usually the intake or new-business system, the conflicts database, the practice or matter management system, the document management system for ethical wall enforcement and the identity system that controls access. The AI components read from these and write proposals back; the conflicts database remains the system of record for decisions.

Sources

  1. SRA Code of Conduct for Solicitors, RELs, RFLs and RSLs — Solicitors Regulation Authority · checked 10 October 2026
  2. Conflicts of interest (guidance) — Solicitors Regulation Authority · checked 10 October 2026
  3. Confidentiality of client information (guidance) — Solicitors Regulation Authority · checked 10 October 2026

More in Legal Services

Back to Legal Services

Next step

Test entity resolution on your own conflicts backlog

Send a description of your intake and conflicts systems and a few anonymised intake requests that were slow to clear. We will show where structured capture and entity matching would change the path, and where your current process should stay as it is.

Describe your intake process