Every UK firm knows it must run a conflict check before taking on a matter. The SRA Code of Conduct paragraph 6.2, in force in its current form since 11 April 2025, is clear that you must not act where there is a conflict of interest or a significant risk of one, save in narrow circumstances with informed consent in writing. What is less obvious is that the tool most firms rely on, a search box over a spreadsheet or a case management system, quietly misses the conflicts that matter most.
This is a buyer's guide to automated conflict check software: what it should actually do, the near-misses that separate a real check from a name lookup, and where artificial intelligence genuinely helps.
A name-match list is not a conflict check
Most systems marketed as conflict checking are really a search over a list. You type a name, the system returns exact matches. That is useful, but it answers the wrong question. A conflict check has to answer "is any party on this new matter the same real person or company as a party we have already acted for or against", and real names never line up neatly.
The same company appears as "Zenith Holdings Ltd" on one file and "Zenith Holdings Limited" on another. A client is entered once with a middle name and once without. A paralegal types "Jon" where a colleague typed "John". A search box treats all of these as different parties and returns nothing. The conflict is real; the tool says the file is clean.
The near-misses most conflict tools miss
Strong automated conflict checking is defined by the cases it catches when the text does not match exactly:
- Legal suffixes and abbreviations. "Ltd" and "Limited", "and" and "&", "Co" and "Company" should collapse to the same entity.
- Typos and fuzzy matches. "Zenyth Holdings" should still surface "Zenith Holdings" for review.
- Sound-alike names. "Katherine" and "Catherine", "Smith" and "Smyth" sound the same and are frequently the same person entered by two people.
- Accented and pasted names. A client pasted from a Word document as "Zénith" or "Müller" must still match "Zenith" and "Muller". A tool that only reads plain ASCII drops the accented version on the floor.
- Initials. "J Smith" as an opponent should raise "John Smith" the existing client for a human to check.
- Company numbers and email addresses. Where you hold them, an exact company number or email is the strongest signal of the same entity and should override a spelling difference.
If a conflict tool cannot demonstrate these, it is a search box, not a conflict check.
Firm-wide by default, not just your own matters
A conflict of interest is a duty on the firm, not the individual. Yet many systems, and almost every manual process, screen only the records belonging to the person running the check. A junior opens a matter against "Acme Ltd", their own files show nothing, and the check passes, while a partner two desks away already acts for Acme.
Automated conflict check software has to screen the whole firm: every fee-earner's matters and clients, current and former, because acting against a former client can also be a conflict. Just as important, when it does find a match, it should tell you which colleague holds the conflicting record, so the person opening the matter knows exactly who to speak to before proceeding.
Where AI helps, and where it must not
Artificial intelligence has a genuine role in conflict checking, but a narrow one. The value of AI is entity resolution: deciding whether "Zenyth Holdings Limited" and "Zenith Holdings Ltd" are really the same company, or whether two unrelated people simply share a common surname. That is a judgement a deterministic rule struggles with and a language model is good at.
The danger is letting the model make the compliance decision. A model that can clear a flagged conflict can also erase a real one, and it can be wrong. The safe design is asymmetric: AI may confirm a match and raise a finding, but it must never remove a finding the engine has already flagged. Every flagged conflict stays visible until a qualified solicitor reviews it. Missing a conflict is far worse than a false alarm a solicitor dismisses in a second, so the software should be built to fail towards showing you more, not less.
The waiver audit trail the SRA expects
Finding a conflict is only half the job. Paragraph 6.2 allows a firm to act in defined situations, for example where clients have a substantially common interest or are competing for the same objective, but only with informed consent given or evidenced in writing and appropriate safeguards. When something goes wrong, your professional indemnity insurer and the SRA will ask how the conflict was handled.
That means the software must record the decision, not just the detection. For every flagged conflict there should be a permanent note of who reviewed it, when, whether it was cleared as a false positive or waived, and the reason. A conflict register that only records detections, or that lets a later decision quietly overwrite an earlier one, will not stand up to scrutiny.
A checklist for automated conflict-check software
Before you commit to a system, confirm it can:
- Run automatically the moment a matter is opened, not only when someone remembers.
- Match on fuzzy, phonetic and accent-aware variants, not just exact text.
- Screen the whole firm, current and former clients, and name the colleague who holds the match.
- Use AI to confirm entities without ever clearing a flag on its own.
- Grade each result so definite conflicts stand out from possible ones.
- Keep a tamper-evident audit trail of who cleared or waived each conflict and why.
- Process data in the UK or EEA and never train models on your client data.
How Writford's AI Conflict Check works
Writford builds all of this into the matter workflow. The moment you open a matter, the AI Conflict Check screens the client and opposing party against every party your firm has ever acted for or against, using exact, fuzzy, phonetic and accent-aware matching, plus company number and email signals. Amazon Nova AI then adjudicates the close matches, confirming genuine same-entity overlaps and raising them, while never clearing a flag itself.
Every matter carries a clear status: definite, potential or none. A persistent banner warns on every screen until a solicitor clears the finding as a false positive or waives it with a recorded reason, giving you the informed-consent audit trail the SRA expects. Each finding also names the colleague who holds the conflicting record. Data is processed in the UK and EEA, and we do not train models on your matters.
Summary
Automated conflict checking is not about replacing the solicitor's judgement. It is about making sure the judgement is made on complete information: every close name, across the whole firm, with AI to resolve the ambiguous cases and a solid audit trail behind every decision. A search box cannot do that. If you want to see conflict checking that catches the near-misses and keeps the record the SRA expects, Writford starts at 59 GBP per seat per month with a 14-day free trial and no minimum contract.