An assurance review has always run on interviews. You can read every document a site will give you and still not know where the event really stands, because the documents describe the plan and the people describe the plan’s odds. The interviews are where a review earns its fee.
They are also the part that has always been hardest to run well, and the reason is logistics rather than method.
What the interview week actually costs
Picture the usual arrangement. A consultant flies in for a week. Someone at the site spends two weeks beforehand building a schedule around shift patterns, vacation, a unit upset that moves everything, and the fact that the three people you most need to hear from are the three people with the least free time. Slots get compressed to fit. People get combined into group sessions to save an hour.
Then the week runs, and the consultant is sharper on Tuesday morning than on Thursday afternoon, because consultants are people. And the group sessions, which felt like the efficient choice, quietly did the most damage: in a room with their manager, the planner does not say the packages will not be ready.
None of that is a failure of method. The method is sound. It is a failure of scheduling, and we were paying for it in coverage, in consistency, and in candor.
What we changed
Every STOAR now runs AI-led one-on-one interviews as standard, across the whole arc: the readiness reviews on the front end and the closeout review after the event.
The client gives us the roster once. Each person gets their own link and takes their interview when it suits them, by voice or by typing, from a phone or a laptop. There is no interview week to build a site around and nothing for a coordinator to chase.
Each interview knows the seat it is talking to. A fixed-equipment engineer, a materials coordinator, an execution contractor, and a steering sponsor get four genuinely different conversations, each with its own question line, and each anchored to what we already know about the event from the documents and the schedule. When an answer is vague, it asks again. When a real risk appears, it keeps pulling the thread instead of moving on to protect a schedule, because it does not have one.
The part that surprised me
I expected the efficiency. What I did not expect was that the findings would get stronger.
Two reasons. The first is coverage: everyone gets interviewed, so the read no longer depends on who happened to be free that week. The seat you skip is often the one holding the problem.
The second is cross-examination. When one interview surfaces something, it is put, anonymously, to the disciplines that have not been interviewed yet. Operations says the systems will be turned over on schedule; planning, later that week, is asked about the same handover and describes something different. Two people who would never have contradicted each other in a room have just given us the exact disconnect a review exists to find. In a single group session, that disagreement never happens. Someone defers, and everyone leaves believing the plan.
What has not changed
The judgment. The interviews gather and cross-examine, and every finding still reaches the report only when I accept it. Nothing is scored, gated, or written up automatically. That wall is deliberate and it is not moving: the value of an assurance review is an independent professional putting their name on a read, and no amount of good tooling substitutes for that.
What the tooling changes is how much of the truth reaches the person doing the judging, and how little it costs the client to get it there.
Where this sits
We had this on the roadmap as a future capability. It arrived early, it worked better than the thing it replaced, and so it is not a premium option or a pilot. It is how a STOAR runs now.
If you are planning a turnaround and want to see what your own team would tell it, that is a short conversation.