On training myself to make decisions under uncertainty
How my skill-coach AI agent pushed me to stop observing and start deciding
What, if improved, would multiply my impact? I'd never asked myself that.
I used to treat skill development like an endless backlog, diving into whatever looked urgent or interesting, only to end up with a collection of half-baked skills and dust bunnies living in the crevices. The issue wasn't effort. It was never deciding what to get better at, which is fundamentally an allocation problem.
Now I think of skill development as capital allocation.
My colleagues already cover most of my gaps, which lets me double down on my strengths instead of patching weaknesses.
But how could I actually do that?
My skill-coach agent
At Titanium Birch we’ve been experimenting with AI agents for some of our workflows. When my colleague Peter suggested I build one for my own development, I made a simple skill-coach agent with GitHub Copilot.
The first few conversations were generic and useless. Then I gave it real context: my background, feedback from colleagues, instructions on how to challenge me, and everything about Titanium Birch’s values, goals, and how we diligence funds and startups. I even tuned it to sound like someone I’d actually want to talk to (banned: “let’s unpack that”).
Now each conversation starts with a structured interview and ends with one clear priority plus a concrete plan tied to real work on the horizon. I told the agent: find where I could become exceptional, pick a single focus and justify the tradeoffs, define what “good” looks like, and link it to actual work.
Screenshot of my skill-coach.agent.md file.
My role spans hiring, investing, and branding. It would be nice to get my financial modelling up to investment-banking level, but my colleagues have that covered. We also played with the idea of building agents to speed up diligence, but it didn’t feel like the highest-impact bet. After spitballing with the agent and talking through team gaps with Peter, I landed on one thing: getting significantly better at assessing people. After all, I’m the first gate in our hiring process and in many of our investment evaluations.
Getting better at assessing humans
I talk to a lot of people every day. I’d always characterised my instincts for reading people as “pretty good.” I notice body language and facial cues (never smiled → how do I make them feel more relaxed?), catch patterns or inconsistencies in what they say, spot recurring motifs in their life stories (jumps from project to project with little to show → shiny-object syndrome), and sketch rough ideas of how they relate to the world (this person seems deeply afraid of being disliked). That intuition probably comes from my love for storytelling—I majored in comp lit and did some scriptwriting in my 20s—but it lived entirely in my head, and I struggled to explain it to anyone else.
So I’d notice things, flag them to myself, and then… not push. Or I’d take people at face value more than I should and hesitate to form a strong opinion.
In other words, I got stuck here:
Conversation > Observe signals > Interpret
The problem wasn’t spotting signals; it was turning them into actual decisions. The longer I stayed in interpretation mode, the less time I spent actually hiring or investing.
My skill-coach agent suggested I fill out a “scouting report” after every important conversation (it knows I’m a huge NBA fan, so the name stuck). Why not? I already jot down my thoughts after each chat. This would simply force me one step further—from listing interpretations to making a clear judgement.
Now my process looks like this:
Conversation > Observe signals > Interpret > Form a judgement
Here’s the simple template we came up with:
Scouting report
What I noticed (3–5 observations)
Specific thing they said or did
Interpretation: What I think it means
Confidence: High/Medium/Low
What's at stake?
What do I think this person really wants? What's driving them? What would make them walk away? What do they stand to lose?
What would change my mind?
Gut call: 👍 / 👎 / 🤷 and one sentence on why
After the fact (fill in later)
Was my read right? What did I miss? Where can I shift left?
And here’s what it looks like in practice (entirely fictional):
2026-04-23 Scouting report
What I noticed
Competitive sailing (represented country at two Worlds) > engineering degree > early employee at startup (acquired by pharma company) > product lead at eCommerce platform > launched venture fund
Interpretation: She’s operated at a high level in very different environments. Competitive. Network skews toward operators and builders. Could mean a differentiated deal pipeline.
Confidence: High
Talked about her team members with genuine enthusiasm, care, and admiration.
Interpretation: Deeply invested in cultivating her team and giving people room to shine. That should help build an enduring firm. How does this look when the team grows beyond four people?
Confidence: High
Her answers on processes stayed too high-level; my attempts to dig deeper were fruitless.
Interpretation: Possible blind spot on operational rigour. Next conversation: check if she sees it and whether anyone on the team is compensating.
Confidence: Medium
What's at stake?
She wants to prove an operator can build a durable fund vintage after vintage. This is tied to her identity. She’d walk from a superficially attractive deal if she didn’t respect the founder’s approach to building.
What would change my mind?
Total obliviousness to the operational blind spot with no interest in fixing it.
Any inconsistency between how she describes her team and how her colleagues describe it.
Gut call: 🤷 Strong operator with good people instincts, but operational depth still unclear. Next meeting decides whether we move to diligence.
After the fact (2026-04-30)
Partially right. Probed on operations; she said M. “handles all of that.” Later it came out that the M. is leaving for his previous corporate role. When I asked about the plan to replace him, she shrugged.
A critical function is concentrated in one person who is leaving, and she shows no urgency in owning it or replacing that capability. Flag: structural risk with no mitigation plan.
Decision: Pass.
Shift left: in every intro call, ask what they see as the biggest risks to their firm.
What I’ve noticed so far
Documenting these judgements clarifies the thinking behind every decision. Over time I’ll use the collection to spot patterns and biases in my judgement. I’ve already noticed I’m biased toward people with unconventional career paths (like my own).
I can’t yet say for sure whether my people-assessment skills have measurably improved; hiring and investing take time to validate. But the scouting reports forced me to confront something big: my desire for psychological comfort (more data, more discussion) was quietly eroding my decision velocity.
In hiring and investing, good calls compound. Bad calls compound too. Delayed judgement compounds into missed opportunities and diluted conviction (or worse, no conviction at all).
The fix isn’t more information. It’s making a call with explicit confidence levels and iterating as reality unfolds.
The byproducts have been surprisingly useful. I’m now gutsier about probing in real time. Before, I’d let things slide if I didn’t fully understand or if something felt off. Now I push back: “I don’t follow… Can you walk me through that again?” or “That’s at odds with what you said earlier. What do you mean?” Most people engage with it. And when my probing visibly annoys someone, I’ve probably touched a nerve.
I used to think not pushing back was polite. It’s not. If I don’t understand something and let it slide, I’m not being kind—I’m being lazy about clarity.
Most importantly, these reports forced me to ask myself the hardest question: what do I actually think?
A pattern has already started to emerge. The people who stand out to me have a clear, cohesive arc—I can see how what they’re doing now fits into their broader story. After several conversations, I crystallised this into a working thesis and shared it with the team. Writing it down changed what I listen for, and my conversations are no longer checklist diligence but tests of narrative coherence and self-awareness.
The same logic applies to what I'm learning. Treating skill development like capital allocation means I don’t get to hedge. I have to place a small number of bets, commit to them, and let them play out.
What’s the return?
The real return on deliberate skill development isn’t more knowledge; it’s sharper decisions about who we hire, who we invest in, and how we actually build a firm that lasts.
If you thrive in a small, high-trust team that brings this same rigour to people as well as to markets and processes—and where generalists are encouraged to take their skills from good to exceptional—we’d love to talk. We’re hiring.
Disclaimer
The content in this post should not be taken as investment advice and does not constitute any offer or solicitation offering or recommending any investment product.