You use AI agents to write code. Maybe to debug, refactor, or generate tests. But have you used one to figure out what you actually want?
I don't mean asking ChatGPT "what should I do with my life?" and getting a generic listicle. I mean a structured, multi-round interview with specialized AI experts who listen to your answers, challenge your assumptions, disagree with each other, and gradually uncover things you didn't know you thought.
That's what I did. And it changed my strategy more in one afternoon than months of journaling, whiteboarding, and late-night overthinking.
I've been a developer for over a decade. Computer science degree, currently studying psychology on the side because I'm genuinely fascinated by how people think. I build things — AI systems, distributed architectures, production RAG pipelines, kernel-level security tools. I'm good at shipping, and I love the work.
But I was stuck on the meta-level. I want financial independence. I'm a father — my daughter needs a stable childhood, and I want to make sure we never have to leave Switzerland. The AI wave is reshaping everything, and I'm not sure how safe any job is over the next few years. So I want to earn extra money, build a buffer, maybe transition to independent work eventually.
The problem: I had too many options and couldn't figure out the right strategy. Should I scan freelance platforms? Build an agency? Focus on inbound? What niche? What's realistic? How much do I actually need? And underneath all of that — what do I really want? Financial independence sounds clear until you try to define it. Then it fractures into ten different versions, each with different implications.
I'd been going in circles. So I tried something different.
What if I assembled a panel of AI experts — each with a different specialty — and had them interview me? Not a brainstorming session where AI generates ideas. An actual structured interview where I'm the subject, the experts listen, react to what I say, build on each other's observations, and disagree when they see things differently.
The key insight: I don't need AI to generate a strategy. I need it to help me discover what strategy is already implicit in my own answers.
I'm studying psychology — and one thing I've learned is that people often don't know what they want until someone asks the right questions in the right order. A good therapist doesn't tell you what to think — they create conditions for you to hear yourself think. That's what I wanted from this panel.
I configured four experts in Claude Code, each as a specialized agent with a clear role:
| Expert | Role | Why This Angle |
|---|---|---|
| Lena Richter | Solo Consulting Strategist | Knows the DACH freelance market, pricing, deal flow. The "is this realistic?" voice. |
| Daniel Voss | AI Market Analyst | Tracks what companies actually pay for vs. hype. The "where is the market going?" voice. |
| Sophie Meier | Financial Independence Strategist | Swiss-specific. Knows the math, the tax implications, the timelines. The "here are the numbers" voice. |
| Dr. Markus Stein | Jungian Psychologist | Individuation, shadow work, introvert dynamics. The "what are you really saying?" voice. |
The psychologist was the most important addition — and almost an afterthought. I added him mid-setup because I realized half of what I was wrestling with wasn't a business question. It was an identity question.
The rules matter more than the experts themselves:
This structure prevents the most common failure mode of AI brainstorming: the AI generates a wall of plausible-sounding advice that has nothing to do with your actual situation. By forcing a question → answer → reaction loop, every piece of advice is grounded in something I actually said.
The facilitator (Claude) started with one question. Not "what's your business plan?" or "what are your goals?" Just: What does financial freedom actually look like for you, in concrete terms?
My answer revealed more than I expected. I talked about my daughter needing a stable childhood. About being a German expat in Switzerland for four years and never wanting to have to leave. About the AI wave making me nervous. About wanting enough money to live from passive income so I could "get money out of my mind forever."
Each expert reacted differently:
Lena (the strategist) immediately separated two goals I'd been conflating — financial independence and consulting income:
"You're mixing two goals: financial independence (passive income, never needing to work) and consulting income (active work, trading time for money). These are different games."
Daniel (the market analyst) reframed my AI anxiety:
"Your skillset — implementation-heavy, not CRUD — is actually in one of the safer buckets. AI isn't replacing the people who build MCP servers and multi-agent orchestration. It's replacing the people who glue together forms and databases."
Sophie (the financial planner) pulled out the calculator:
"Even with 5% returns, you're looking at 25-30 years to hit your target from scratch."
And then Dr. Stein said this:
"I hear a man trying to solve an emotional problem with a number. Your target isn't a financial goal — it's an attempt to contain anxiety about displacement, instability, and control."
That one stopped me.
When I pushed back — explaining that I'd still build things even if I were financially free, that I love the work, that I just want the money anxiety gone — Dr. Stein caught something I hadn't seen:
"Here's the contradiction I can't ignore: you say you want passive income so you 'never have to work,' yet your entire self-concept is builder. You're not someone who dreams of lying on a beach. You dream of shipping things that work, alone, without needing permission. So why are you chasing a retirement fantasy?"
Meanwhile, Lena shifted her advice entirely based on my answer. Since I'd clarified that consulting was acceleration fuel (not the end goal), she revised her targets downward — from "build a consulting empire" to "sustainable side income that doesn't burn you out."
This is the pattern working. Each round, the experts refine their model of who I am. By round 2, they were giving advice that was impossible to get in round 1 because they didn't have the context yet.
I mentioned my father being a workaholic. His father too. And that I feel the same drive in myself. That I've always been a builder — "I always do shit" — despite constraints that slowed me down.
Dr. Stein's response was the most precise psychological observation I've ever received from an AI:
"Three generations of workaholics, and now you're at an inflection point. What gets activated when barriers lift? The inherited persona of the tireless worker, or your authentic drive? 'I need to learn to love earning money' sounds like the father's voice, not yours."
He then reframed something I'd always seen as a limitation:
"What if your constraints aren't bugs, but the psyche's way of forcing you toward a different path than your father's?"
I mentioned wanting to write children's books, study psychology deeper, do more active imagination — but only after financial independence. Dr. Stein named this immediately:
"Jung would call this provisional life — the belief that real life starts after you complete some condition. The psyche creates this elegant trap: you postpone individuation until you're 'safe,' but the work of becoming yourself is precisely what creates that inner safety."
By round 4, the panel had converged on a concrete model: inbound, fixed-scope, remote, async implementation sprints. I was energized. "Sounds PERFECT — let's fucking go."
Lena gave me a 5-step 30-day action plan. Daniel validated the market demand. Sophie ran the numbers on a realistic timeline.
But Dr. Stein noticed what the others missed — that I was energized by the doing, not the earning:
"Stress doesn't disappear with security. It changes costumes. Notice what lit you up in your message: not 'I'll finally relax,' but 'let's fucking go' — the building, the doing. That's your real fuel."
I pushed back: "You can't just decide to have a low-stress life. The stress disappears once you have security."
He didn't dismiss it. He validated the legitimate need for a buffer as a father and expat — then challenged the illusion:
"Three generations of workaholics didn't stop at 'enough.' They kept running because the chase itself was the point. Define what 'enough' looks like in concrete terms, so you know when to shift gears. Otherwise you'll be 55, wealthy, and still chasing the next milestone while your daughter wonders why Dad's always 'just finishing one more thing.'"
In the final round, each expert gave their complete recommendation based on everything I'd actually said across five rounds:
Lena defined the business model: fixed-scope AI implementation sprints, 10-15 hours/week max, one active project at a time, rate floor of 150 CHF/hour quoted as fixed-price deliverables. Her warning: "The biggest risk isn't that this fails — it's that it succeeds, you take on 4 projects at once, and you burn out by July."
Daniel defined the market position: AI implementation specialist — RAG, agents, MCP — for DACH companies. "Companies aren't searching for 'MCP consultant.' They're searching for someone who can connect their AI systems to their existing tools. You build the bridge."
Sophie defined three financial tiers — Emergency Secure (survive job loss), Employment Optional (employer becomes a choice), Financially Independent (work only if you want to). "Optimize for Tier 1 first. It removes your biggest fear and costs the least time."
Dr. Stein defined the psychological guardrails. His final recommendation included practices I should start now, not after financial independence: weekly active imagination, building a children's book with my daughter, monthly "shadow checks" asking whether the month felt like joy or grinding. His closing line:
"Individuation doesn't happen after financial independence. It happens when you stop postponing the things that make you feel whole."
After the session, I updated my AI assistant's persistent memory and project strategy files with the consolidated output. This means every future conversation — about code, content, strategy, anything — is informed by what we discovered.
Concretely:
The AI didn't generate this strategy. It helped me excavate it from my own answers.
Here's what makes this more than a one-time exercise: the output went back into the AI's persistent memory.
Claude Code has a CLAUDE.md file (project-level instructions) and a memory directory that persists across conversations. After the expert panel session, I updated both with the strategic conclusions. Now when I start a new conversation — about a blog post, a code refactor, a portfolio update — the AI already knows:
This creates a feedback loop: self-discovery → persistent configuration → better future interactions → more refined self-discovery. The AI assistant doesn't just know my codebase. It knows my strategy, my constraints, and my psychological patterns. Every future conversation is better because of this one.
If you want to try this, here's the template:
Not "what should I do?" but something specific enough that experts can react to your answers. Examples: - "I want to go independent but I'm not sure when or how" - "I have three project ideas and can only pursue one" - "I'm burning out and I don't know why"
Match the panel to the question. 2-4 experts is the sweet spot. Each should have a distinct angle that creates productive tension. Some combinations:
| Question Type | Expert Panel |
|---|---|
| Career strategy | Industry analyst + business strategist + financial planner + psychologist |
| Technical architecture | Backend specialist + security expert + DevOps engineer + UX researcher |
| Product direction | Market researcher + designer + engineer + customer advocate |
| Personal decision | Psychologist + financial planner + domain expert + devil's advocate |
Always include one expert who looks at the human/emotional dimension. The psychologist was the most valuable voice on my panel — and I almost didn't add one.
The facilitator asks one question. You answer honestly. Experts react. This is the engine — question, answer, reaction, repeat. Each round builds on the last.
This is what turns a one-time conversation into a lasting strategic asset. Update your CLAUDE.md, your memory files, your project docs — whatever your AI assistant reads on startup. The conclusions compound over time.
Start with the psychologist, not the strategist. I added Dr. Stein as an afterthought, but his observations were the foundation everything else built on. If I'd started the first round with a psychological question ("what's actually driving this?") instead of a strategic one ("what does financial freedom mean?"), we might have gotten to the core faster.
Fewer experts for simpler questions. Four was right for my situation because I had financial, strategic, market, and psychological dimensions all tangled up. For a more focused question — "which project should I build next?" — two or three experts would be plenty.
Push back harder on the AI. The experts were good at challenging me, but I could have challenged them more. When Daniel suggested a pure MCP niche, I initially accepted it — then questioned it later and we corrected course. The panel works best when you treat it as a real conversation, not a consultation.
We're still in the early days of figuring out what AI agents are actually for. Most people use them as faster keyboards — type less, get code faster. But agents can do something more interesting: they can create structured environments for thinking that would be impractical to set up with humans.
Try assembling four domain experts for a five-round interview about your career strategy. In the real world, that's thousands of dollars and weeks of scheduling. With AI agents, it's an afternoon. The experts aren't real, but the questions are real, your answers are real, and the insights that emerge from the structure are real.
The expert panel pattern isn't about getting AI advice. It's about using AI to ask yourself better questions — and then remembering the answers.