What Would You Say You Do Here?

In Office Space, the two consultants named Bob ask every employee the same question: “What would you say you do here?” It’s played for laughs because the answer should be obvious. Smart, accomplished professionals struggle to articulate the thing that makes them valuable.

They can do the work. They can’t describe the work.

This has always been a career problem. Resumes flatten expertise into bullet points. Interviews reward rehearsed narratives over genuine insight. The people who land the best roles aren’t always the most capable — they’re the most articulate about their capability. But now the articulation gap has a second cost, and it’s more immediate.

The Context Gap

AI tools give you back exactly what you put in. Ask a generic question, get a generic answer. Set the right context — your expertise, your decision-making frameworks, your preferences for how work gets done — and the output shifts from “plausible” to “useful.”

In tools like Claude Code and Claude Cowork, this context takes the form of skills files and project instructions: structured documents the model reads at the start of every session. They capture how you think, what you prioritize, what “good” looks like in your domain. They make a huge difference to the outputs you get.

I published a library of finance skills — documents covering financial analysis, portfolio construction, risk assessment. Over 800 people downloaded it in a few weeks. The demand for well-structured context is real.

The constraint is creating it.

The Blank Page Problem

Writing a good skills file requires a kind of structured self-reflection most professionals haven’t practiced. You know what you know — you just can’t write it down in a way that’s useful to a model or, for that matter, to another person.

We’ve all faced a blank page and had writers block. We write a few generic sentences — “I’m experienced in financial planning” — and stop. The gap between what we do and what we describe is enormous. Not because we lack expertise, but because expertise lives in pattern recognition, judgment calls, and accumulated instinct. Documenting that is hard.

Two Bobs

So I built Two Bobs — a free tool that conducts structured AI interviews to extract your expertise into a portable skill file.

The name is the joke, and the joke is the point. The tool asks you the question the Bobs asked: what would you say you do here? Then it keeps asking. It probes for specifics, follows up on interesting threads, pushes past the generic self-descriptions people default to. “I’m good at strategy” becomes “I evaluate market entry decisions by mapping distribution leverage against switching costs, and I’ve learned the second variable matters more than most people think.”

The output is a SKILL.md file — a structured document that captures your decision-making frameworks, communication style, and domain knowledge. It works across 40+ AI platforms. Pick a mode — quick skill extraction, deep expertise interview, personal brand discovery, or career coaching — and have a conversation. The AI handles the structure so you can focus on the substance.

The Surprise

I built Two Bobs to solve a context engineering problem. The interview was a means to an end — a way to get people past the blank page. In testing with friends and family, the interview became the product. I added personal brand and career coaching modes. The same structured interview that makes AI better also makes people more self-aware about their own value, and documents context in a different way for humans (or their AI assistants) to use.

What This Means

The professionals who thrive in an AI-augmented world will be the ones who can articulate what they know. AI amplifies whatever context you give it. Vague context produces vague output. Precise context produces precise output.

This applies beyond AI. The ability to externalize your expertise — to describe your frameworks, your judgment, your approach — is the same skill that makes you effective in interviews, in leadership, in mentoring, in any situation where someone needs to understand how you think.

The best way to make AI useful is to arm it with context about how you work: your expertise, your preferences, your standards. Write it down. Iterate as you experiment. And if staring at a blank page isn’t working, let the Bobs ask the questions.