Tell me about a time your manager overruled you and you had to carry out the decision anyway.

After an outage where our scanner missed a vulnerable container image on a client estate, my manager decided the whole practice would standardise on the single vendor we already had under contract. I had run the bake-off, and my read was that this tool was weakest exactly where we had just been burned: it missed seven of the twenty-two deliberately vulnerable images we seeded into the test. I made that case once, properly, in a one-to-one, with the seeded-image results in front of us. He didn't dismiss it. He told me the part I couldn't see — the alternative meant reopening terms on every client contract, plus a retraining cost the practice couldn't carry that year. Then he made the call. Two things I did next. I wrote the decision up in the team channel in his reasoning rather than mine, so nobody had to work out where I stood, and I didn't re-open it in standups. And I asked for one thing in exchange: a compensating image-scan step in the pipeline, and a date to look at the numbers again. We cut over in about a sprint and a half. Criticals settled around thirty-one instead of spiking the way I had predicted, so my objection was overweighted. The image gap was real, though — the extra step caught two criticals in the first month and got funded properly at the review.

I led a six-person security practice inside the agency. After a client-visible outage traced back to a rushed remediation, our delivery director introduced a sign-off gate: a named security lead had to approve every production change on every account. My objection was specific — that put one person's calendar in front of remediation work, and I expected our open critical count to climb. I put it to her once, with the queueing numbers behind it, and I asked for something narrower than a reversal: a figure that would count as the gate failing. We agreed on one. Above forty criticals for two consecutive weeks, we would look at it together, and nothing reopened before then. Then I committed, and the part that mattered was how visibly. When my team asked whether I had fought it, I said the call was hers, that the reasoning was sound given what the client had just watched happen, and that we would argue for changes with data and not with grumbling. I rostered sign-off across three of us so it wasn't one calendar, and I instrumented the queue properly. Week seven, we hit forty-four. I brought the chart, not the argument I had made originally. She narrowed the gate to internet-facing tiers inside a fortnight, the count came back under thirty, and the gate stayed where it genuinely bought us something.