How a Conflict-of-Interest Matter Should Actually Start

2026-10-09 · Philip Choo · AI9OS

Someone is bound by a restrictive covenant: a non-compete in an employment contract, or a promise in a sale agreement by the person who sold the business not to compete with the buyer. They are suspected of running or joining a competing venture, and perhaps of taking staff or clients with them.

The investigation itself is usually straightforward. Company registrations, directorships, web presence, staff movement, all laid against a date.

What decides whether it is worth doing happens before a single search is run.

The question is almost never "whether". It is "when"

Whether the person is active in the sector is rarely contested. What is contested is when that activity started, or escalated, relative to a date in a document. Before the covenant began, after it expired, during it. Everything in the investigation is ultimately measured from that one date.

Which is why the first request is the most important one.

Produce the clause

Not a description of the clause. The text, and the date the agreement was executed.

A summary of a covenant, in a brief or an email, is somebody's reading of it. The investigation will be measured against the words, so the words have to be on the file before anything else happens. If a date in the brief is incomplete or truncated, that is the first thing to resolve, because every timeline, every comparison of archived web pages and every conclusion about whether the covenant had expired will depend on it.

A brief is a starting point, and it deserves to be read closely whoever sent it. Its job is to describe the matter. The investigator's job at this stage is to identify what the brief does not yet establish.

Is the covenant still running, and could it be enforced?

This is a question for the lawyers, but the investigator should ask it out loud before taking money. An expired covenant, or one so broad in scope, duration or geography that it is unlikely to be enforced, makes the matter unwinnable, and every hour spent on it is wasted. The investigator does not decide it. The investigator makes sure somebody has.

Name the single blocking fact

Most matters of this kind have one unknown that decides whether the matter is winnable at all. Often it is a date. Sometimes it is whether a particular person was ever employed by a particular entity.

Name it, in writing, before mobilising. Refuse to start collection until it is resolved, or until the client accepts in writing that the work proceeds without it. Starting without it is how a well-run investigation ends up answering a question nobody needed answered.

Split the objectives honestly

For each objective, say plainly which of three routes can deliver it:

• what open sources can establish: registrations, domain records, dated archived web pages, public marketing;
• what needs work in the field;
• what only the client's own documents can prove: employment records, internal emails, the departing person's device if the client owns it.

That last category is the one most often forgotten. If a load-bearing fact already sits in the client's own files, say so before taking a fee to look for it elsewhere. It is the most useful thing an investigator can tell a client at this stage, and the least commercially convenient.

Do the same for the anticipated defence. Which half of it can open sources actually reach? Say so in the proposal, rather than letting the client assume the whole of it.

Confirm who is who

Where subjects are identified by name and matched to photographs, the match must be confirmed in writing by the instructing party. A mapping inferred from the order of documents in a bundle is an inference, not an identification. It stays marked as inferred until confirmed. Following, photographing or approaching the wrong person is a catastrophic failure, and the prevention costs one email.

Settle the rest before signature

• Is a claim filed, and is there a hearing, disclosure or limitation date? A litigation timetable reorders every priority.
• Where did the client's own supplied material come from, and did they have the right to pass it on? The same provenance question as a signed declaration.
• Run the conflict check across every party, before signature.
• Lock the list of subjects and the list of people not to be approached before any collection begins.

And score the matter against the acceptance criteria, recording the number. A written score is what makes it possible to decline later without an argument, which is the discipline set out in why a practice should be willing to say no.

Then, and only then, the investigation

When the work does start, the central product is a correlation of dates: entity formation, domain registration, changes in web presence, staff movement and marketing shifts, all laid against the one date in the agreement. Dated, third-party-held records such as a domain's registration date or an archived page are often the strongest material in the case, precisely because they are not anyone's word against anyone else's.

If the matter is a reopen of earlier work, read the raw records before the earlier report. Why that matters is the subject of a case closed as no links found.

If you are instructing on a covenant matter

Send the clause, not a summary of it. And ask the investigator: "What is the one fact that decides this, and do we already have it?"

(General information for practitioners, not legal advice.)