"Tell Me About a Time You Influenced Without Authority": Evidence, Allies, and the Pilot Nobody Could Argue With
Architects have no reports and rarely a mandate, so influence without authority is the job itself. The question tests whether you work from other people's incentives, bring evidence instead of persistence, lower the cost of saying yes with a pilot, and know the line between influence and lobbying. Here is the playbook, the story to choose, and the follow-ups.
An architect’s org chart is usually a lie about their power. Nobody reports to them, the teams whose code they care about have their own managers and their own roadmaps, and the only thing they can actually force is a comment in a design review. Every outcome they are judged on depends on people who do not have to listen. That is why “tell me about a time you had to influence without authority” is in nearly every architect loop: it is not a behavioral question about a soft skill, it is a question about whether you can do the central mechanic of the job.
This is the fifth behavioral post, after the decision that went wrong, disagreeing with a senior engineer, pushing back on a deadline, and your biggest weakness. The disagreement post was about changing one person’s mind on one decision. This one is about changing what a group of people do, over time, when you cannot tell them to. By the end you will have a playbook with named levers, a test for the difference between influence and lobbying, a story structure, and the follow-ups that separate people who have done this from people who have read about it.
The Question
The direct forms:
- “Tell me about a time you had to influence without authority.”
- “Tell me about a time you convinced a team to change direction.”
- “Tell me about a time you got buy-in for something you could not mandate.”
- “Tell me about a cross-team initiative you drove.”
- “Tell me about a time you changed something outside your own team.”
- “How do you get engineers to adopt a standard?”
The variants emphasize different levers. “Convinced” is about evidence. “Buy-in” is about incentives and ownership. “Cross-team” is about allies and the real decision-maker. “Adopt a standard” is about lowering the cost of yes and making the change durable. One story told with the full playbook covers all of them, and the last variant is the one to have a second, shorter story for.
What They Are Really Asking
- Do you understand that this is the job? An architect who expects to be obeyed will fail in the role within a quarter. The interviewer wants to hear that you start from “they do not have to do this” and work from there, without resentment.
- Do you start from their incentives or yours? The team you want to change has a roadmap, a manager, an on-call rotation, and a reason they do it the way they do. Candidates who can explain why the other team said no, in terms the other team would agree with, are the ones who have actually done this.
- Is your influence built on evidence or on persistence? Nagging until someone gives in is a kind of influence, and it is the kind that makes you the person people avoid. A number, a prototype, or an incident changes minds and leaves no resentment behind.
- Did you find the real decider and the people they trust? Most “the team would not adopt it” stories are really “I never talked to the person who decided.” Influence travels along trust; a peer the team already listens to is worth more than a document.
- Did you lower the cost of saying yes? A pilot on one service with agreed success criteria is a small ask. A migration is a big one. Candidates who shrink the ask, do the first piece of work themselves, and let the result argue for the rest are the ones who get adoption.
- Do you know the line between influence and lobbying? Hallway deals, coalitions against a person, and selectively presented evidence get results once and cost you the ability to do it again. The interviewer is listening for whether the other side would be comfortable seeing your whole process.
- Did it last? Changes that depend on you remembering to push are not changes. A default, a template, a lint rule, or a standard outlives you. So does a team that owns the change because they decided it.
A weak answer is “I presented the benefits and eventually they agreed.” A strong answer has the other team’s reason for saying no stated fairly, the evidence that addressed that reason specifically, a pilot small enough to be a safe yes, a result with a number, the moment the decision became theirs, the mechanism that made it stick, and the honest part: what you gave up and who still did not adopt it.
The Gotchas
Gotcha 1: The story where you had authority. “My manager asked me to roll out the new framework, so I did.” If someone with authority sponsored it from the start, you were executing a mandate. The interviewer will ask who sponsored it, and the answer has to be “nobody, at first.”
Gotcha 2: Persistence as the method. “I kept bringing it up until they agreed.” That is attrition. Everyone in the room has been on the other end of it, and it is the single most common way candidates reveal that they do not have a method.
Gotcha 3: Escalation as the first move. “I got the VP to mandate it.” Borrowed authority is a real lever, and sometimes the right one, but it is the last one, used transparently, and it is not what the question asked. A story that starts with escalation says you could not do it without authority.
Gotcha 4: Winning the argument and losing the adoption. “They agreed it was the right approach.” And then? Agreement is free. If nothing changed in the code six months later, you did not influence anything; you won a meeting.
Gotcha 5: Not knowing why they said no. If you cannot state the other team’s objection in a form they would sign, you never understood it, and your evidence was aimed at an objection you invented. The interviewer will ask “why did they resist?” and “they were being difficult” ends the story badly.
Gotcha 6: Making it about being right. The story is not that your approach was better. It usually was; that is why you cared. The story is how a group of people who did not have to care came to act on it.
Gotcha 7: The lobbying story. “I lined up the other leads before the meeting so they would back me.” Told that way, it is a coalition against someone. Told as “I asked the leads who had been burned by this whether they would share what it cost them,” it is evidence. The difference is whether the other side would be fine seeing the whole process, and interviewers are listening for which one you did.
Gotcha 8: The all-at-once ask. Asking for the migration, the standard, and the org-wide rollout in the first conversation. The first ask is the pilot, on the service whose owners are already in pain, with criteria they set.
Gotcha 9: No evidence. “I explained the benefits.” Benefits are opinions. A duplicate-order rate, an incident count, a prototype that handles their hardest case, a cost line: these are the things that move a team that has other priorities.
Gotcha 10: You paid nothing. Influence is credible in proportion to what it cost you. If you wrote the library, ran the pilot, took the first on-call shift for it, or did the first migration yourself, say so. If the other team did all the work and you did the slides, the interviewer will notice that too.
Gotcha 11: No durability and no dissent. Did it stick after you stopped pushing, and who never adopted it? A story where everyone converted and it was perfect is a story the interviewer does not believe. The team that declined, and why that was acceptable, is the detail that makes the rest true.
How to Answer
Step 1: State the philosophy in two sentences
I cannot make anyone do anything, so my job is to make the thing I think is right the easiest and best-evidenced option for the people who decide. I start with what they are optimizing for, bring evidence they would accept, shrink the ask to a pilot they can safely say yes to, and let them own the decision so it outlasts me.
Four ideas: their incentives, evidence, a small ask, and ownership. Everything after it is illustration.
Step 2: The playbook
| Step | What you do | Why it matters |
|---|---|---|
| Find out what they are optimizing for | Talk to the team before proposing anything. What is on their roadmap, what pages them at night, what did they try before and why did it fail | Every “no” is a “yes” to something else. You cannot compete with it until you know what it is |
| Find the real decider and who they trust | Usually a tech lead or a manager, not the whole team, and usually one or two peers whose opinion they already weigh | Influence travels along trust. A document from a stranger loses to a sentence from a peer |
| Build the evidence they would accept | A number from their own system, a prototype that handles their hardest case, the incident they were paged for, a cost line | Evidence is the thing that does not have a rank. Pick the form that answers their actual objection, not the one that is easiest for you to produce |
| Reframe as their problem solved | “You spent two on-call rotations on duplicate orders; this removes that class of incident” rather than “this is the standard I want” | People adopt solutions to their problems. They resist other people’s preferences |
| Shrink the ask to a pilot | One service, a few weeks, success criteria they set, a rollback they control | A pilot is a safe yes. A migration is a career risk for whoever says yes to it |
| Pay the cost yourself | Write the first integration, do the first migration, take the first pager, write the docs | Proves it is not a slide deck, and removes the “we do not have time” objection by not spending their time |
| Let them own it | The pilot’s result is presented by them; the decision to expand is theirs; the credit is theirs | Ownership is what makes it durable. A change that is “yours” dies when you stop attending |
| Make it the default | A template, a shared library, a lint rule, a paved road, a line in the standard | Durability by mechanism, not by memory. The next team gets it without a conversation |
| Close the loop, including the dissent | Report the result, name who did not adopt it and why, and say whether that was fine | The dissent is what makes the report credible, and sometimes they were right |
Step 3: The levers, and what each costs
Influence has sources, and senior candidates can name which one they used. This is the table the interviewer has in their head.
| Lever | What it looks like | When it works | What it costs |
|---|---|---|---|
| Evidence | A number, a prototype, an incident, a cost | When the disagreement is about facts or predictions | Time to produce it, and the risk it disproves you |
| Doing the work | The library, the first migration, the docs, the pager | When the objection is capacity | Your time, which is the point |
| Allies | A peer the team trusts who has seen the problem | When the objection is trust or unfamiliarity | Must be transparent; this is the lever closest to lobbying |
| Lowering the cost of yes | A pilot, a reversible step, a default that can be opted out of | When the objection is risk | A slower rollout, and the pilot has to be allowed to fail |
| Timing | Proposing right after the incident, or at planning | Almost always; the same proposal lands differently in different weeks | Patience, and sometimes waiting for pain you could have prevented |
| Legitimacy | The design review process, an ADR (Architecture Decision Record), a published standard | When the objection is “says who” | Only works if the process is respected; using it on a two-way door burns it |
| Borrowed authority | A sponsor, an escalation, a mandate | Security, compliance, a one-way door, or when evidence was ignored | Used once, transparently, after the others; used first, it ends your ability to influence that team |
The sentence interviewers want: evidence and doing the work are the levers that build credit; allies and borrowed authority are the levers that spend it.
Step 4: The line between influence and lobbying
Interviewers at the staff and principal level ask this directly, and candidates who have a crisp answer stand out.
| Influence | Lobbying |
|---|---|
| The other team’s interests are served by the change, and you can say how | The change serves you, and their interests are something to get around |
| You would be comfortable if they saw every conversation you had about it | Some conversations only work because they were not in the room |
| Allies share evidence: “here is what this cost us” | Allies share votes: “back me in the meeting” |
| The coalition is for an idea | The coalition is against a person or a team |
| The evidence includes the parts that cut against you | The evidence was selected |
| The decider makes the decision, with the information | The decision is engineered around the decider |
| Escalation is announced: “I think this is a one-way door and I am going to raise it” | Escalation is a surprise |
The test is the second row. If every conversation could be read aloud to the person you were influencing, it was influence. The rest follow from that.
Step 5: The shape of the change
A picture of what the playbook produces over time, which is also the shape of the story you are about to tell.
flowchart LR
P[Pain: an incident or a recurring cost<br/>in the team's own system] --> U[Understand: their roadmap,<br/>their objection, their decider]
U --> E[Evidence aimed at<br/>that objection]
E --> S[Pilot on one service,<br/>their success criteria,<br/>you do the first work]
S --> R{Result against<br/>the criteria}
R -->|Met| O[They present it.<br/>They decide to expand.]
R -->|Not met| L[Say so. Fix or stop.<br/>Credit for honesty is reusable.]
O --> D[Default: template, library,<br/>lint rule, standard]
D --> C[Close the loop:<br/>who adopted, who did not, why]
style S fill:#2d6a4f,color:#fff
The pilot is highlighted because it is where most influence stories succeed or fail. A pilot with criteria the team set, that you did the work for, that they can roll back, is a yes that costs them nothing. Everything after it is the result arguing on your behalf.
Step 6: Choose the story
| Criterion | Why it matters |
|---|---|
| Nobody with authority sponsored it at the start | Otherwise it was a mandate |
| The other team had a real reason to say no, which you can state fairly | Proves you understood the incentives |
| You can name the evidence and why it answered their objection | Proves the method |
| There was a pilot, with criteria they set | Proves you shrank the ask |
| You did real work: code, migration, docs, on-call | Proves it cost you something |
| The result has a number, and the team presented it | Proves the outcome and the ownership |
| It became a default or a standard | Proves durability |
| Someone did not adopt it, and you know why | Proves honesty, and often shows judgment about where the change did not apply |
Prepare two. One where it worked, told in full. One where you failed to influence, with what you learned about the lever you should have used. Interviewers ask for the failure about half the time, and the candidate who has one ready has already answered the real question, which is whether you can see your own method clearly enough to say where it broke.
Step 7: Tell it in six beats
| Beat | What to say | Time |
|---|---|---|
| The headline | “Three teams handled retries differently and we were producing duplicate orders. I got them onto one approach without owning any of their code.” | 15 seconds |
| Their side | What each team was optimizing for, and why the obvious fix was a no for them: a roadmap, a prior failed attempt, an upcoming launch | 30 seconds |
| The evidence and the ask | The number from their own system, the prototype against their hardest case, and the pilot you proposed with their criteria | 45 seconds |
| The work | What you personally built, migrated, documented, or carried a pager for | 20 seconds |
| The result and the hand-off | The number after, who presented it, and the decision they made to expand | 30 seconds |
| Durability and dissent | What made it the default, who did not adopt it and why, and whether that was right | 30 seconds |
Three minutes, with the other team’s reasons given as much time as your own. The beat candidates skip is the second one, and it is the one that proves you did this rather than announced it.
A worked outline, with details you replace with your own:
We had three services placing orders, each with its own retry logic, and a run of duplicate-order incidents. I wanted them on a shared idempotency approach, and none of those teams reported to me or to anyone I could ask.
The team with the worst incidents had tried a shared library once before, it had broken their deploys, and they had written their own since; their objection was “we do not trust shared code on our critical path,” and that was fair. The second team had a launch in six weeks and would not touch anything. The third had no incidents and no reason to care.
So I started with the first team’s own numbers: eleven duplicate orders in a quarter, two of them refunded manually, one on-call rotation spent on it. I built a prototype against the case that had broken the previous library, their double-submit on a retried gateway timeout, and showed it handled it. The ask was a pilot on their one write path, for a month, with their criterion: zero duplicates and no deploy regressions. I wrote the integration myself and took the first week of on-call for it.
After the month, duplicates on that path were zero, and their lead presented it at the platform review, not me. The second team adopted it after their launch, because by then it was a template default in the service scaffold rather than a request from me. The third team never did, and that was right: their writes were naturally idempotent and the library would have been ceremony. Two years later it is still the default and I had moved teams.
Count the levers: understanding, evidence aimed at the actual objection, a pilot with their criteria, work you paid for, ownership transferred, a default that made the next adoption free, and a dissent that was correct. The idempotency post is the technical half of that story if you want it.
Step 8: When to use borrowed authority
The question is about doing without it, but a senior answer knows when it is the right lever, and says so before being asked.
| Situation | Why influence alone is wrong | What to do |
|---|---|---|
| Security or compliance | Consent is not the standard; the requirement exists whether or not the team agrees | Use the mandate, and spend your influence on making compliance cheap |
| A one-way door another team is about to walk through | The reversal cost is paid by the organization, not by the team deciding | Say you are escalating, escalate with evidence, and accept the outcome |
| Evidence was accepted and ignored | The disagreement is no longer about facts; it is about priorities someone above both of you owns | Ask that person to decide, with both sides present |
| A standard that only works if everyone follows it | A protocol, a tracing header, an auth scheme: one holdout breaks it for all | Influence the pilot, then ask for the mandate on the rollout, with the pilot as the evidence |
The rule: borrowed authority is used transparently, after the other levers, and with the evidence they produced. Used that way, it costs nothing. Used first, it is the last time that team takes your call.
Follow-Up Questions to Expect
- “Why did they resist?” The test of whether you understood their side. Answer in terms they would sign: a prior failure, a launch, a trust problem with shared code. “They were set in their ways” fails.
- “What if they had still said no after the pilot?” For a two-way door, accept it and keep the pilot result for the next conversation. For a one-way door, say you would escalate transparently with the pilot as evidence, and that the pilot is what makes the escalation fair.
- “Tell me about a time you failed to influence.” Have it ready. The best versions name the lever you used and the one you should have: “I brought a document when the objection was trust, and what I needed was a peer who had seen it work.”
- “How do you know it was you, and not that they would have done it anyway?” Honest answer: you do not, entirely. Point to the thing that would not have existed without you: the prototype, the pilot criteria, the template default.
- “How do you influence someone more senior, or another organization?” Same playbook, with more weight on evidence and timing and less on doing the work yourself, because you cannot. And a sponsor, found transparently, who owns the relationship you do not.
- “How do you keep this from becoming politics?” The read-aloud test. If every conversation could be shared with the person you were influencing, it was not politics.
- “What did it cost you?” Weeks of your own time, usually, and some credit given away. If the answer is “nothing,” the interviewer will wonder who did the work.
Key Takeaways
- Influence without authority is not a soft skill adjacent to the architect role; it is the role. Start from “they do not have to,” without resentment.
- Every no is a yes to something else. Find out what the team is optimizing for before proposing anything, and be able to state their objection in words they would sign.
- Evidence and doing the work build credit; allies and borrowed authority spend it. Name the lever you used.
- Shrink the ask to a pilot: one service, their success criteria, a rollback they control, and the first work done by you.
- Let them own the result. A change that is yours dies when you stop attending; a change that is theirs becomes a default.
- Durability is a mechanism: a template, a library, a lint rule, a standard. Not a reminder.
- The line between influence and lobbying is whether every conversation could be read aloud to the person you were influencing.
- Borrowed authority is a real lever: used last, transparently, with the evidence the other levers produced.
- Tell the failure story too. It proves you can see your own method.
Further Reading
- Allan R. Cohen and David L. Bradford, Influence Without Authority (the book the question is named after; the currencies-of-exchange model is the “find out what they are optimizing for” step)
- Robert Cialdini, Influence: The Psychology of Persuasion (reciprocity and commitment, which are why doing the work and the pilot work)
- Will Larson, Staff Engineer: Leadership Beyond the Management Track, on operating without authority at the staff level
- Tanya Reilly, The Staff Engineer’s Path, on sponsorship, pilots, and making change stick
- Michael Nygard, Documenting Architecture Decisions: the legitimacy lever, in one page