How to Talk About Your AI Workflow in Job Interviews (2026 Guide)

You are twenty three minutes into a job interview for an operations analyst role, and the hiring manager leans forward and asks: walk me through how you actually use AI, short for artificial intelligence, at work. Not whether you use it. How.

This question is now standard at companies hiring for procurement, supply chain, and operations roles, and most candidates fumble it. They say something like "I use ChatGPT for research" or "I use AI to summarize documents," and the interviewer's face does not change, because that answer tells them nothing. It could describe anyone who has opened a browser tab in the last two years.

The candidates who get the offer answer differently. They name the tool. They name the model. They describe one real prompt, close to word for word. They explain what they checked before they trusted the output, and they end with a number that changed because of what they did. That is the difference between a resume line and a hire.

This guide gives you the exact structure to build that answer, using a real example you can adapt from your own work, even if you have never written a line of code.

Why employers ask this now

Two years ago, asking a candidate about AI usage was a bonus question, something an interviewer might toss in at the end to seem current. In 2026, it has moved to the center of the interview for almost every entry level operations, procurement, and analyst role, and the reason is practical, not trendy.

Companies that hire for these roles, whether that is a Global Capability Center, a Big Four consulting arm, or a mid sized manufacturer, are under direct pressure to cut the time it takes to do routine analytical work. A purchase order reconciliation that used to take a new analyst two days now needs to take two hours, because the tools exist to make that possible and competitors are already using them. The employer is not asking about AI to check a trend box. They are asking because the answer tells them whether you can do the job at the speed the business now expects.

There is a second, quieter reason. Hiring managers have been burned by candidates who claim AI fluency on a resume and then cannot describe a single real use of it in an interview. This happened often enough in 2024 and 2025 that most interviewers now treat the generic claim "I am comfortable with AI tools" as a mild red flag rather than a credential, because it usually means the candidate skimmed an article rather than actually did the work. A specific answer, by contrast, is very hard to fake. You cannot describe a real prompt, a real verification step, and a real number unless you actually ran the workflow.

This is good news for you. It means the bar is not "do you have five years of AI experience," which almost nobody your age has. The bar is "can you describe one real, small, honest use of AI clearly enough that the interviewer believes you actually did it." That bar is genuinely reachable in a weekend, which is the entire point of this guide.

This matters more, not less, in procurement, supply chain, and operations roles specifically, because these teams sit closest to the actual money. A purchasing team that misreads a supplier contract or misses a duplicate payment loses real dollars, not just time, so hiring managers in this space tend to probe harder on the verification habit than a generic tech recruiter might. If your target roles are Global Capability Center analyst seats, sourcing coordinator positions, or entry level supply chain planning jobs, expect this question to arrive earlier in the interview, sometimes in the first ten minutes, and expect at least one follow up question about a time the AI got something wrong.

What "AI workflow" actually means

Before you can describe your AI workflow, you need a working definition of what that phrase means, because most people use it loosely and interviewers notice the difference between loose and precise.

An AI workflow is not the AI. It is the sequence of decisions you make around a tool: what you asked it to do, what you gave it to work with, what you checked before you accepted the result, and what changed because of it. The AI is one component in that sequence, not the whole story. When you say "I use AI," you are describing a tool. When you say "I use Claude Code to pull vendor invoice data from a CSV file, flag anything that looks like a duplicate payment, and then check the flagged rows against the original PDF before I report them," you are describing a workflow, and workflow is what gets hired.

Four things matter more than any others when you describe your workflow.

The tool and the model. Saying "I used AI" is like saying "I used software." It tells the listener almost nothing. Saying "I used Claude Code, built by Anthropic, running on the Sonnet model," tells them exactly what capability you had access to and lets a technical interviewer calibrate their follow up questions correctly. Model names matter because different models are genuinely different tools with different strengths, the same way Excel and a calculator are both technically software but nobody confuses them.

The prompt. This is the part almost nobody prepares, and it is the single easiest way to sound credible in thirty seconds. You do not need to recite your prompt word for word from memory. You need to be able to describe, specifically, what instruction and what data you gave the tool. Compare "I asked it to analyze the spend data" to "I gave it a CSV of 380 vendor invoice lines and asked it to group them by category and flag any invoice number that appeared more than once." The second version proves you actually ran something.

The verification step. This is the component that separates a hire from a pass, and almost nobody mentions it without being asked directly, which is exactly why mentioning it unprompted makes you stand out. Every experienced professional who uses AI for real work checks the output before acting on it, because current AI tools, including very capable ones, still make mistakes, miss context, and occasionally state numbers that look plausible but are not real. When you describe checking the flagged duplicates against the original PDF invoices before reporting them, you are not describing caution for its own sake. You are demonstrating the exact judgment an employer needs from someone handling real financial or operational data.

The outcome. End with what changed, ideally as a number. Not "I saved time," which is vague and unverifiable, but "I cut the reconciliation step from about a full day to under twenty minutes," or "I found twelve duplicate payments worth roughly $3,400 that the manual process had missed the previous quarter." A number this specific is either true, in which case it is powerful, or clearly invented, in which case do not use a number you cannot stand behind if the interviewer asks a follow up question.

Put those four pieces together, in that order, and you have a workflow description that survives real scrutiny instead of a claim that falls apart the moment someone asks a second question.

Practice these interview questions

Reading about how to talk about your AI workflow is useful, but the real test is answering out loud, under a little pressure, without notes. The questions below are the ones most likely to come up once an interviewer knows you use AI tools regularly. For each one, think through your own honest answer before reading the sample, then compare. The goal is not to memorize the sample word for word, it is to borrow its structure and fill it with your own real experience.