AI Explainability — the “Why?”
Every important recommendation can show its work: the exact factors, weights, and trade-offs behind the call. No black box.
4 min read
Open the “Why?”
Any recommended task has a Why? action. Open it and Zovik lays out the full reasoning in plain language — no vague "the AI thinks so". Everything shown is the same deterministic data that produced the score, so the explanation is exactly as trustworthy as the number it explains.
How the score was calculated
The panel shows each of the five factors as a row: its raw 0–100 read, the weight applied, and its share of the final score — so you can see what actually drove the ranking, with the dominant factor highlighted. The shares add up to 100%: it's a genuine split, not a decoration.
Confidence and provenance
A confidence badge tells you how much real signal backs the call — high when the task has a deadline, an estimate, a goal link and a priority; low when it's a best guess from a bare title. A provenance badge says whether the wording came from the deterministic engine alone or was refined by the AI. The score itself is always deterministic.
What would change it — and the alternatives
Zovik names the one change that would move the recommendation (add a deadline, link a goal, raise a priority) so you're never stuck guessing. It also lists the runner-up tasks and how far behind they scored, so the ranking is legible in context. You stay in control — Zovik explains and recommends; it never reorganises behind your back.
Low confidence isn't a bug — it's Zovik being honest that it needs more signal. Add a due date or link the task to a goal to sharpen the call.