What it is
AI explainability is an AI system's ability to show the reasoning behind a specific output, not just deliver the output itself. When an AI system scores a lead, flags a signal or matches a candidate to a role, explainability is what lets it answer the follow-up question a consultant always has: why? Not "how sure are you" (that is a confidence score) and not "did you check this" (that is verification against hallucination), but "what led you to this conclusion".
Explainability shows up as the evidence and the chain of reasoning behind a decision: which data points fed a lead score, which detail made a signal look real, which attributes lined a candidate up against a role. It does not have to be a technical trace of the model's internals. For a working consultant it just needs to be a clear, short answer: this account scored highly because of a recent funding round and a headcount spike that match your ICP; this candidate matched because their last two roles line up with the mandate's core requirement.
A score you can't question isn't intelligence. It's a black box with a number on it.
Why it matters
An AI system that only outputs a number or a verdict is a black box, and a consultant who cannot see inside it has exactly one option: trust it blindly or ignore it entirely. Neither is workable when the output feeds outreach that goes out under the consultant's own name, to a client relationship built over years. Explainability is what turns an opaque score into something a consultant can actually verify in the few seconds before they act on it.
It also determines how fast an agency will actually adopt AI. A consultant who has been burned once by an unexplained wrong call stops trusting the whole system, not just the one output, the failure mode that quietly kills AI adoption long before anyone calls it that. A system that shows its reasoning earns the opposite: every correct explanation the consultant checks and approves builds the case for trusting the next one, and eventually for trusting the system with more autonomy.
How boilr handles it
In boilr, every signal, lead score and candidate match your AI sales employee surfaces carries the reasoning behind it in plain language, alongside the confidence score, never instead of it. You see why an account was flagged, which data points drove the score, and why a candidate was put forward, in the same view where you would otherwise just see a number. That is what lets you verify and send a task in seconds rather than re-researching the company yourself to check the AI's homework.
The reasoning also feeds the Company Brain. When you approve a task, you are confirming the logic behind it was sound for your desk. When you edit or reject one, you are correcting it. Over time that reasoning sharpens against what your agency actually knows works, which is the difference between explainability as a one-off feature and explainability as something that gets better the longer you use it.
