Influencing a technical decision without authority

Expected question

"Tell me about a time you influenced a technical decision when you did not own the teams involved and could not escalate your way to a mandate."

Variant forms

Interviewers often probe the same competency with different framing — recognize the archetype and answer with your story:

"Tell me about a time you influenced people without authority."
"Describe getting another team to adopt your design via RFC/docs/tooling rather than mandate."
"Tell me about resolving a cross-team technical disagreement without escalating to leadership."
"How do you make the right architecture the easiest path for other teams?"
"Tell me about building a paved-road library or template that replaced three competing patterns."
"Describe aligning incentives so other teams wanted your proposal for their OKRs."
"Tell me about a time influence failed the first time — what did you change?"
"Walk me through Staff-level leadership: how you drove a decision across org boundaries."

The question, as it might actually be asked

"Tell me about a time you influenced a technical decision when you did not own the teams involved and could not escalate your way to a mandate." This is a core Staff+ behavioral signal: durable cross-team alignment via artifacts and incentives, not title. Answer with your own real experience — the case study below is one real example of this competency, not the assignment.

Situation

Across the vpeetla-ai org, multiple demos and platforms were independently inventing adjacent patterns for governance of side effects (publish, notify, tool calls) and for agent skills install — each repo slightly different, each "working," none shared. No single team owned "platform standards," and nobody reported to anyone else across those repos.

Task

Get independent repo owners to adopt a shared approach (AegisAI-style gateway before irreversible actions; shared skills install from vpeetla-ai-skills) without a mandate from above and without blocking anyone's launch calendar.

Action

I treated influence as making the right path easier than the wrong one:

Wrote the constraint down in org agent instructions (AGENTS.md / skills): side effects require gateway or HITL; skills install is one scripted path — so new work defaulted to the shared pattern instead of inventing a fourth variant.
Shipped reference implementations in repos people already cared about (content-factory publish HITL, practice-arena/playbook as graded artifacts) rather than a slide deck asking for adoption.
Cross-linked ADRs and interview playbook entries so the "why" lived next to running code — reviewers could cite a real decision, not a preference.
Avoided forcing rewrites mid-launch — new work adopted the standard; migrations were opportunistic. That kept owners from experiencing the standard as a tax on their deadline.

The mechanism was artifacts + defaults + proof in production-shaped demos, not persuasion meetings.

Result

New agent work in the org consistently routes irreversible actions through gateway/HITL patterns, and skills bootstrap is a single documented install path. Divergence still exists in older corners (honest: influence was partial), but the default for net-new systems stopped fragmenting.

The follow-up question you should expect

"What did you do when a team still refused?" Answer: I narrowed the ask to the next greenfield surface instead of demanding a rewrite, offered to pair on the first gateway integration, and documented the risk of divergence in the ADR so the cost was visible if they revisited later. Forcing a political win that reverts after the meeting is not influence.

What's expected at each level

Mid-level: describes convincing teammates on their own squad.
Senior: cross-team agreement with a concrete proposal.
Staff+: names the mechanism (RFC/defaults/reference impl) and durable adoption without authority.
Principal: org-level standard with honest limits of influence and how dissent was handled.

03 Org-wide security hardening — related influence shape
05 Leading a 0-to-1 AI product