How to talk about AI in your behavioral interviews

Meta now directly assess AI impact in their behaviorals

Even large, lumbering talent organizations like Meta’s are reimagining their interview processes as AI changes what companies need from engineers, PMs, designers, etc. The takeaway for you is that you will discuss not only your experience with AI¹ but also your philosophy on it in your next behavioral interview. Questions like these are becoming common:

• How have you used AI to increase your impact?

• How do you safely use AI in your day-to-day work?

• How are you staying on top of the latest developments in the application of AI?

I’ve long predicted the behavioral will become an ever more important round for companies to collect signal like initiative, ownership, and business sense because AI-pilled builders have such a large potential for impact. But it’s now also a place for companies to acquire explicit AI signal, including these kinds of values- or philosophy-oriented signals, since so much about the future of building is uncertain as ways of working change.

If you don’t care about Meta, jump down to the Talking about AI in a behavioral section to learn how to adapt to this trend.

I’ve also updated my AI story building and co-pilot to incorporate these new Meta signal areas and general guidance on talking about AI.

Bonus: it comes with a paperback copy of my book :)

Meta has changed their behavioral evaluation signal areas

Last week, Meta rolled out changes to all three engineering interview rounds, transforming them into so-called “AI-Native” versions. This includes changing their behavioral interview signal areas to explicitly account for AI workflows.

For a long time, Meta has used five signal areas to formally evaluate engineering candidates in their behaviorals:

• Driving Results

• Embracing Ambiguity

• Resolving Conflicts

• Growing Continuously

• Communicating Effectively

PM and other roles have similar but slightly different signal areas. Learn more about how I’ve recommended candidates approach Meta’s behaviorals here.

After April 16th, engineering candidates are receiving behavioral rounds with new signal areas:

• AI-Driven Impact:

  • How do you leverage AI to create measurable impact?

  • How do you execute / drive forward work to achieve results?

  • How do you safely and responsibly use AI to enhance problem solving?

• Continuous AI Learning:

  • How do you seek out and build on opportunities for growth and learning both AI and non-AI technologies and concepts?

  • How do you experiment and iterate with emerging technological capabilities and workflows?

  • Do you take constructive criticism as an opportunity to improve?

  • How have you approached improving your skills?

• Partnering:

  • How do you drive collaboration on cross-functional projects?

  • When faced with disagreements with colleagues and/or managers, how have you resolved them?

  • How have you addressed conflicts directly and approached difficult conversations?

• Embracing Ambiguity:

  • How do you operate in an ambiguous and quickly changing environment?

  • Are you comfortable making decisions and maintaining high levels of productivity when you are missing information or lack clarity?

  • How did you react when you had to quickly pivot away from a project due to a shift in priority?

• Communicating Effectively:

  • How well do you communicate with teams?

  • What about cross-functional partners?

  • How do you tailor your communications based on the work and/or the audience?

I’ve confirmed this with internal sources, but their official SWE prep guide still lists the old signal areas. I imagine that will be updated at some point.

What has actually changed at Meta?

In some ways, this is just remixing and augmenting the previous signal areas, with Driving Results being replaced by AI-Driven Impact and Growing Continuously being replaced by Continuous AI Learning. Notice how some of the old Growing Continuously signal is still being gathered in Continuous AI Learning, things like taking constructive criticism.

But there are significant changes. They’ve made AI front-and-center in these new signal areas, which makes sense as we’ve discussed already—builders leveraging AI are multiples more productive than hand-coders.

Notice the specific signal though:

• Concrete examples of how you’ve leveraged AI to increase output.

• Attention to safety concerns.

• An emphasis on the dynamic nature of AI in today’s workplace.

There are also some other subtle changes to the signal area coverage across the interview:

• Notice the downgrade of top-level conflict resolution signal into a more generic Partnering area. This is something I discussed on Luca Rossi’s podcast, as a hot take: that conflict resolution signal is too often the focus of behaviorals and that interviewers should downplay it vs. collecting ownership, ambiguity, and growth signal.

• One of the effects of this change is to make the Partnering axis more expansive, now making the collaboration and communication signal they’re acquiring more explicit.

• There’s also a more intentional ordering of the signal areas. Continuous AI Learning is close to the top now vs. where Growing Continuously was before. Obviously, AI is changing everything quickly so engineers who can’t adapt will fall behind in their fast-moving teams.

Talking about AI in a behavioral

Even if you don’t care about interviewing at Meta, you’re likely to talk about AI in your next behavioral interview. Here’s how you can adapt if you’re preparing:

Collect Stories About Leveraging AI

• Interviewers will expect you to have a strong story about how you’ve leveraged AI to build something. If you don’t have that story, go get that story.

• Beyond that, be sure to include AI-related actions in your CARL/STAR stories at each phase of the software development lifecycle—not just coding, but design, review, debugging, documentation, and rollout.

• Talk about why AI was the right choice at that phase, not just that you used it. “I used Cursor to scaffold the service because the API surface was well-defined and I wanted to spend my time on the integration edge cases” shows some kind of judgement and not just hoping the one-shot solution will be good enough.

• Don’t retrofit AI into old stories. You probably have a catalog that predates heavy AI use, and that’s fine. Continue to lean on those stories when they’re appropriate, but understand that whenever you can land a story about leveraged AI usage, do so.

Develop a “Philosophy” of AI in the Workplace

In most IC behaviorals, you won’t find what I call Values Questions, things like, “Tell me about your philosophy on performance management.” Those are typically reserved for very senior or people management loops.

However, since AI is changing so fast, your underlying beliefs have will have a big impact on your productivity in your next role, so companies will start asking more explicitly about AI “philosophy.” I use that work in the loosest sense.

• Know when to reach for AI and when to do something by hand and be able to explain your approach. Understand what current models are actually good at, where they fail, and what that means for privacy, security, and correctness in your domain.

• Narrate the decisions around AI. Interviewers will fill narrative gaps with the unfavorable assumption that you’re now a prompt-typer with no underlying judgment. Close those gaps by sharing why you picked the tool, what you verified, what you overrode, what you refused to hand off. Without this, the signal collapses into “I typed some prompts.”

• Given all the recent hype about token costs and the reality of working in a place with so much AI-generated code, have a mental model of cost and benefit—latency, token spend, review overhead, failure modes—not just “AI makes me faster.” Interviewers at serious companies are starting to probe this.

Have a Structured Learning Approach

• Think about how you learn and how you’ll keep learning. What do you read and why? What is the cadence of trying new things? How do you separate signal from hype?

• Be ready to talk about how you experiment with new technologies. What have you built recently just to learn something? How do you go about choosing what personal projects to build? How do you know when to bring something you’ve learned on personal time into your work environment?

Example behavioral questions about AI

I typically don’t advise you to prepare for behaviorals by drilling questions since the interview is about you and your career and because it’s difficult to predict what questions you’ll receive. I think we can bend that advice here, especially if you are interviewing at Meta, since we know you’re going to get AI-related questions.

Here are some example questions. Work through these the same way you'd work through any behavioral question: decode what signal is being sought, select a story from your catalog that shows it, deliver with CARL. A few of these don't have clean story matches for most candidates yet—that's expected. Use those to identify where you need to build new stories over the next few months.

AI-Driven Impact

• Tell me about a time you used AI to deliver something you couldn’t have delivered otherwise—or couldn’t have delivered as quickly.

• Describe a project where AI was central to your approach. What did you do yourself and what did you delegate to the model?

• Tell me about a time you decided not to use AI for something you could have. What was your reasoning?

• How do you measure whether AI is actually making you more productive, versus just feeling faster?

Safe and Responsible AI Use

• Tell me about a time you caught an AI mistake before it shipped. What made you look?

• Describe how you handle AI-generated code in code review—yours or someone else’s.

• Tell me about a time you had to push back on an AI-driven approach because of privacy, security, or correctness concerns.

• How do you decide what’s safe to send to an external model and what isn’t?

Continuous AI Learning and Experimentation

• How are you staying on top of developments in AI tooling for your role?

• Tell me about a tool or workflow you adopted in the last six months. What made you try it, and did it stick?

• Describe a time your AI workflow stopped working—the tool changed, the model changed, the problem changed—and how you adapted.

• What’s something about working with AI you believed a year ago that you no longer believe?

Partnering (Cross-Functional Collaboration with AI in the Mix)

• Tell me about a time you disagreed with a teammate about how to use AI on a project.

• Describe a cross-functional project where AI changed how the team worked together.

• Tell me about a time you helped a colleague who was skeptical of AI—or overly reliant on it.

Philosophy and Values

• What’s your philosophy on when to use AI versus doing something by hand?

• How do you think about the tradeoffs between speed and verification when working with AI?

• What’s the most important thing you’ve learned about working with AI?

• How do you think AI will change your role over the next two years, and what are you doing about it?

Don’t wait to add AI to your behavioral prep

If you're interviewing in the next few months, don't wait for a company to publish updated signal areas before adapting². Update your story catalog to include how you’re leveraging AI recently, journal about your philosophy in a paragraph or two, and pick one new tool to actually try this week. That's most of the prep. The rest is making sure you talk about AI the way a builder does: as a tool you direct.

1. For the sticklers, I mean generative AI. Come on, you know I don’t mean gradient descent.

2. Besides, most companies don’t publish their signal areas like Meta does.