The hard question hasn't been asked yet. Nobody has said "the consulting background" or "talk me through the AI piece." You're still in introductions. And already, the person across from you is forming a judgment about whether you believe your own story.
They are reading your verbal patterns. The content comes later, if it comes at all. Trained evaluators score how candidates inhabit their claims. Three verbal cues predicted both interview performance and actual job performance as rated by supervisors: clearly stating a personal opinion, avoiding vague or superficial answers, and talking about emotions. The evaluators were scoring stance.
At Director+, every candidate has a strong background. The differentiator is whether you carry yours like settled fact or like something you're still trying to convince yourself of. And the distance between those two stances shows up as two fundamentally different claims. "I can do this job despite my unusual path" is one candidacy. "My specific path is why I'm the right person for this" is a different one. The behavioral patterns below are how evaluators tell which claim you're actually making, regardless of which words you choose.
This piece catalogs the gap between those two stances. The patterns themselves, which surface regardless of which doubt is activated, and which are legible to anyone who has sat on a hiring panel more than twice. The reframes for specific objections live in individual dossiers. The evidence for any particular company lives there too. Here, the subject is the behavioral layer underneath all of them.
Why the Unscripted Version Is the Real Evaluation Now
Companies are running more reference calls, longer reference calls, and backchannel calls you never hear about. The Wall Street Journal reported that Zapier now requires at least one live reference conversation before making an offer, and that filling a senior role can involve calls to as many as ten references, including off-book ones. One private-equity CEO described spending an hour per call across five to eight references for nearly every full-time hire. Anthropic's candidate AI guidance draws the line explicitly: use AI to prepare and refine your materials, but live interviews are "all you."
AI-polished portfolios and AI-refined outreach have raised the floor on prepared artifacts. The prepared artifact has become table stakes. The unscripted conversation is where separation happens. The recruiter screen where you're slightly tired. The casual coffee where you think you're just getting to know each other. The backchannel call where a former colleague describes what it was actually like to work with you.
Hiring managers who spend an hour per reference call are not checking boxes. They're listening for omissions. Generic endorsements register as filler. Entirely polished praise registers as coached. What registers as credible is specificity, consistency across sources, and balanced detail that includes tradeoffs.
This is the environment your behavioral patterns live in now. You cannot memorize your way to a strong unscripted version. You get one by resolving the doubt underneath, so that what comes out under pressure is the settled version, the one that holds when you're tired and unscripted.
Seven Tells and Their Replacements
What follows are the recurring verbal moves I see in senior design leaders who carry unresolved doubt about some dimension of their background. Each one is legible to experienced evaluators. Each one has a replacement that sounds different because the underlying stance is different.
Scan the bold phrases. Notice which ones produce a flicker of recognition. Most people have two or three defaults, not all seven.
"So my path has been a little nonlinear..." You open with a frame that preemptively explains why your background might look unusual. This feels like honesty to you. To the listener, it reads as apology. You are asking the listener to forgive something before they've decided it needs forgiving. Worse, you've just told them that you see your own sequence as a liability rather than a strategy.
Lead with the claim the sequence supports. "I've spent fifteen years designing systems where errors carry real consequences. That's the throughline." If they want to ask about the path, they will. Most of the time they won't, because you just gave them a frame that makes the path coherent. You explained why the sequence makes strategic sense, which is the only thing a hiring committee at this level actually needs to hear.
"I left that role because the org was restructuring and design was getting absorbed into product, so it wasn't really a fit anymore." Nobody asked. You just flagged a departure as something that needs explaining, which means the listener now wonders what the real story is. The volunteered justification is almost always more damaging than the fact it's trying to soften.
State the move as movement toward. "After three years there I moved to [next role] to [what you went toward]." If they want the departure story, they'll ask. You'll have a clean answer. But you won't have pre-loaded the question.
"It wasn't really consulting in the traditional sense." Or: "I wouldn't say I was doing pure AI work, but..." Or: "My title was Director but the scope was more like a senior IC." Every one of these introduces an objection the listener may not have been thinking about and then half-answers it. Half-answering is worse than either owning the fact or letting someone else raise it. You are telling the evaluator you don't fully believe the title or the role description yourself.
Use the actual words. "I was at BCG DV." "I led the AI integration for the clinical workflow." "I was Director of Product Design." If the listener has a follow-up, they'll ask, and you'll answer with evidence. The hedge is harder to recover from than whatever the follow-up question would have been.
"I know my AI experience isn't as deep as someone who's been at a frontier lab..." The hedge qualifier's more damaging cousin. You are comparing yourself unfavorably to an imaginary candidate who may not exist in this search, and you are doing it for the evaluator so they don't have to. You have just anchored their assessment against a standard you introduced.
Describe what you've built. "I designed the trust architecture for an agentic system where the AI made clinical recommendations and the human needed to know when to override." Whether it happened at a frontier lab or a healthcare startup is a detail.
The three-minute answer to a one-sentence question. Someone asks about a role and you cover the org structure, the political dynamics, the constraints, the reason the scope was what it was, and then the outcome. The outcome arrives last, buried. This pattern activates around the parts of your background you feel most uncertain about. You over-explain because you're trying to preempt every possible misreading. The listener hears someone who needs a lot of scaffolding to get to the point.
Lead with the outcome. "We shipped in three months, zero to MVP, 4.8-star rating." Then add one sentence of context if they need it. Assertiveness mediates the relationship between anxiety and perceived interview performance. Assertiveness here means the willingness to make a claim and let it stand.
"BCG DV wasn't really consulting. Every engagement shipped into production." The content is strong. The structure is defensive. You are leading with what the experience wasn't, which means you are still operating inside the listener's potential objection rather than outside it.
Drop the first sentence. "Every engagement shipped into production with real users, real data, and real operational consequences." Now the employer logo is irrelevant. You didn't defend against the consulting question. You made it disappear.
Filling the silence. You finish an answer and the interviewer pauses. Maybe to write a note. Maybe to think. You keep talking. You add a caveat, a qualifier, a second example that dilutes the first one. The silence felt like doubt, so you filled it with more material, and the additional material almost always weakens the answer that was already complete.
Stop. The answer was done. Let the silence exist. Silence after a strong claim works as emphasis.
The Grain of Truth Underneath
Some of these patterns persist because there is a real thing underneath them. Your path is nonlinear. Your AI work is application-layer, not frontier-research. Your consulting-adjacent experience does carry a specific set of assumptions in the listener's mind.
I am not going to tell you to paper over these facts with better delivery. That will fail in the interview and it will absolutely fail in the backchannel, where a former colleague won't use your positioning language. They're going to describe what they saw.
The resolution is genuine settlement. You have to actually believe that the nonlinear path produced a specific, rare combination of capabilities that a linear path would not have. You have to actually believe that application-layer AI work, where you're designing for trust under consequences that cannot be rolled back, is harder than research-layer work where the cost of being wrong is a retracted paper. Harder because the system has to be right enough for a human to stake a clinical decision or a financial outcome on it, and you have to design the interface that makes that judgment possible. You have to actually believe that shipping under federal oversight in six months is a more demanding speed test than iterating in a low-consequence environment.
If you believe these things, the behavioral patterns resolve on their own. You don't hedge because there's nothing to hedge. You don't pre-justify because the fact doesn't feel like it needs justification. You don't apologize for your AI experience because you've genuinely internalized that designing trust architectures for agentic systems is AI work, full stop.
If you don't believe them yet, no amount of scripting will hold under pressure. The recruiter screen might go fine. The first interview might go fine. But somewhere in the process, in the fourth interview or the casual dinner or the backchannel call your former manager gets, the unscripted version will surface. And the unscripted version is always the real one.
The Reference Test
Here is the test for whether a behavioral pattern is resolved or just managed:
Would someone else say the same thing when you're not in the room? They won't use your exact words. But they'll tell your version of the story.
If you describe your consulting work as "every engagement shipped into production with real users and real consequences," would your former project lead say something that rhymes with that? If you describe your AI work as "designing the trust layer for systems where the machine makes recommendations and the human needs to know when to override," would your engineering counterpart corroborate the framing?
If yes, the pattern is resolved. You are telling a true story and choosing which true story to lead with. That is positioning, and it is legitimate.
If no, if the version you're telling in interviews would surprise the people who were there, you have a factual problem. No catalog of replacement phrasings will fix it.
Before the Next Live Conversation
Read the bold patterns above. Identify the two or three that produce recognition. Then ask the honest question: Do I actually believe the settled version?
The question is whether you believe it. Saying it convincingly is a downstream effect.
If yes, the fix is mechanical. Catch yourself in the first three words of the old pattern and stop. Start over with the replacement. You will feel exposed for about two seconds. That exposure is what settled authority feels like from the inside.
If not yet, go back to the actual project outcomes. What you shipped, what it changed, what happened to users or to the organization because of the work. Talk to the people who were there and ask them what they saw. You are arriving at the conversation already knowing what's true, so that what comes out under pressure is clean.
The hard question is coming. It always does. But by the time they ask it, they've already decided whether they believe you'll answer it from strength or from defense. That decision happened in the first two minutes, in everything you didn't say.
- Anthropic's live-interview line: Their candidate AI guidance says to create your first draft yourself and use Claude to refine it, because they want to see the candidate's real experience and real-time thinking in live interviews.
- AI skills help, but skepticism varies: A study of 1,725 recruiters found AI skills increased interview-invitation rates by 8–15 percentage points, but effects were weaker for graphic designers, consistent with more skeptical recruiter attitudes toward AI in creative work.
- Authenticity cues that actually predict: Heimann and Schmitz-Wilhelmy's study of 181 employed individuals found that verbal authenticity cues correlated with supervisor-rated job performance at r = .24, driven by stating personal opinions, avoiding vague answers, and talking about emotions.
- Explanation design complicates trust claims: Chen, Liao, Vaughan, and Bansal found that feature-based explanations did not improve decision outcomes and increased overreliance on AI, which matters if your positioning centers on designing trust into human-AI systems.