Every firm runs a conflict check. It is one of the first things that happens when a matter opens: the new client's name and the other side go into the practice management system, it searches the firm's records, and it flags anyone the firm has acted for or against. This is bread-and-butter compliance, and most PMS platforms do it well.
So when we talk about a pre-run governance gate that also checks for conflicts, the fair question is: isn't that the same check twice? It is not. They are aimed at two different risks, at two different moments, and a firm using AI now has both risks live. The intake conflict search protects the client relationship. The pre-run governance gate protects the individual piece of work. Missing the second is the newer exposure, and the less understood one.
Why the conflict rule exists at all
Start with what the check is for, because the two versions are protecting the same underlying duty from different angles.
Under the SRA Code of Conduct for Solicitors, you do not act if there is an own interest conflict or a significant risk of one (paragraph 6.1), and you do not act where there is a conflict of interest, or a significant risk of one, between two or more current clients (paragraph 6.2), subject to narrow exceptions where the clients have a substantially common interest or are competing for the same objective, with informed written consent and safeguards. These are not soft expectations. Acting in a conflict goes to the heart of the duty of undivided loyalty a client is owed.
The intake conflict search exists to catch the 6.2 situation: the firm being on both sides, or acting against a former client. That is a relationship-level question, and searching the firm's own database is exactly the right tool for it.
Practical rule: The intake conflict search answers "should this firm act for this client at all?" It is a question about relationships across the firm's whole history. It is necessary, and it is a snapshot in time, taken once, at the door.
What a PMS conflict search actually does
Be precise about the mechanism, because its strengths and its limits both come from the same design.
A PMS conflict search automatically queries the firm's own records: past and present clients, matters, related parties, and adverse parties. You give it names, it returns matches, a person assesses them, and the matter proceeds or it does not. It is fast, it is systematic, and it is far better than memory.
But notice three things about it. It runs once, at intake. It searches only what the firm already knows, its own database. And it governs the decision to take the matter, not anything that happens inside the matter afterwards. Once the matter is open and cleared, the conflict search has done its job and stepped back. Nothing about it is watching the work as it is produced.
That was fine when all the work inside the matter was done by known people drawing on the firm's own files. It is not fine once part of the work is done by an AI drawing on sources the firm did not vet.
What a pre-run governance gate adds
A pre-run governance gate sits at a completely different point: not at the door of the matter, but immediately before a specific piece of AI work runs. It asks a different question. Not "should we act for this client?" but "should this particular task run, on this data, in this jurisdiction, right now?"
That is a check on the work, not the relationship. Before an agent executes, the gate checks things the intake search never touched: whether the task falls in scope, whether the jurisdiction is one the firm is set up to handle, whether the data in play should be exposed to this operation, whether the request itself trips a defined risk condition. It runs every time work runs, not once at the start. And it governs the AI's action specifically, which is the thing the conflict search was never designed to see.
The two are complementary layers, at two moments:
| PMS conflict search | Pre-run governance gate | |
|---|---|---|
| Question it answers | Should the firm act for this client? | Should this specific AI task run, now, on this data? |
| When it runs | Once, at intake | Every time a piece of AI work runs |
| What it examines | The firm's own client and matter database | The task, its scope, jurisdiction, data, and risk conditions |
| What it governs | The decision to take the matter | The AI action inside the matter |
| Blind to | Anything after intake; anything outside the firm's records | Nothing about the work it gates |
Why this matters more as firms adopt AI
Here is the exposure that did not exist five years ago. When a fee earner did the work, the sourcing was implicitly governed: they used the firm's precedents, known databases, their own judgement, all inside the firm's control. An AI agent does not automatically respect that boundary. It can pull from sources the firm never assessed, blend them without flagging it, and produce something that looks authoritative and is not.
That creates conflict and sourcing risks the intake check cannot see, because they arise inside the matter, after the door closed. A research task might surface and lean on material connected to an adverse party. A drafting task might reach for data that should have been walled off. The intake conflict search cleared the relationship weeks ago and is not in the room. Nothing is watching the specific action unless something is built to watch it.
Field note: The mistake I see is treating the intake conflict search as if it covered AI risk too. It does not, and not because it is weak. It is because it runs once, looks only inward, and governs the relationship. AI risk is continuous, looks outward, and lives in the individual task. Different risk, different moment, different check.
This is why our governance layer runs pre-run checks before an agent acts, alongside the post-run verification that ties every output back to a verified source. The Research Agent, for instance, works in a closed world: it cannot cite a source it did not retrieve and quote, which shuts down the sourcing risk directly. We wrote about that anti-hallucination discipline in how to verify AI legal citations, and about the wider evidence duty in what the SRA can actually ask about AI.
This is a second layer, not a replacement
None of this asks a firm to drop its intake conflict process. Keep it. It does a job the governance gate does not do, clearing the relationship across the firm's whole history, and it does it well.
The pre-run gate is a second layer for a second risk. The conflict search governs whether you act. The governance gate governs how the AI works once you are acting. A firm using AI needs the relationship cleared at the door and the work governed as it runs, because a clean intake check is no answer to an AI task that sourced the wrong thing three weeks into the matter.
The honest framing is not "conflict search versus governance gate." It is "conflict search, and then governance on top." One protects the client you took on. The other protects every piece of work you produce for them. Do the first without the second, in a firm now leaning on AI, and you have governed the handshake and left the work ungoverned.
LegalAI Space builds AI agents for legal teams with a governance layer that makes every output verifiable, compliant, and audit-ready. Sign up for early access or book a pilot call with Founder Daman Kaur.
FAQ
Is a PMS conflict search the same as AI governance? No. A conflict search runs once at intake and queries the firm's own database to decide whether the firm should act for a client at all. A pre-run governance gate runs every time AI work executes and checks whether that specific task should run, on this data, in this jurisdiction. Different risk, different moment.
What does the SRA say about conflicts of interest? Under the SRA Code of Conduct for Solicitors, paragraph 6.1 says you do not act where there is an own interest conflict or a significant risk of one, and 6.2 says you do not act where there is a conflict, or significant risk of one, between current clients, subject to narrow exceptions requiring informed written consent and safeguards.
Why isn't the intake conflict check enough once a firm uses AI? Because it runs once, searches only the firm's own records, and governs the relationship, not the work. AI can draw on sources the firm never vetted and produce authoritative-looking output inside the matter, weeks after intake. That risk is continuous and outward-facing, and the intake check is not watching it.
What does a pre-run governance gate check? Before an agent runs, it checks the task itself: whether it is in scope, whether the jurisdiction is supported, whether the data should be exposed to the operation, and whether the request trips a defined risk condition. It governs the AI action, not the decision to take the matter.
Does LegalAI Space replace our conflict checking? No. It is a second layer for a second risk. Keep your intake conflict process to clear the relationship. LegalAI Space governs how the AI works once you are acting, which the conflict search was never designed to do.
Sources
-
SRA, Code of Conduct for Solicitors, RELs and RFLs, Section 6 (Conflict, confidentiality and disclosure). Paragraph 6.1 (own interest conflict) and 6.2 (conflict between current clients), with exceptions for substantially common interest or competing for the same objective, subject to informed written consent and safeguards.
-
SRA, Standards and Regulations, Code of Conduct for Firms, Rule 2.5 (identifying, monitoring and managing material risks), which extends to the risks introduced by AI use inside a matter.
Related reading
- Legal matter management: what it actually means in 2026: the governed matter this check sits inside.
- How to verify AI legal citations: the sourcing risk a governance gate is built to close.
- What the SRA can actually ask about AI: the evidence a firm has to be able to produce.