The first 30 days

Pick a topic from a live deal. Ask it two ways, in two separate chats: a generic version (“What should I look for in a sponsor’s revenue growth assumptions?”) and a specific one (“B2B SaaS, 32% historical CAGR, sponsor projects 28% for five years, 18 customers = 70% of revenue. Tell me the three most aggressive things about that projection and what to ask in management meetings to test them.”)

Read both back-to-back. In your log: what was the generic version missing? This is the entire game in one drill — get used to feeling the difference.

Open a new chat. Paste your “About me” block (from setup) at the top even if your settings are already configured, then ask a real question from your day. Notice the system warm up to your role.

On Copilot with a OneNote context notebook? Today, add your firm’s investment criteria and your PM’s known preferences to it, then ask Copilot to use them on a real question.

Pick something you actually need to understand better — a sub-sector, a covenant structure, a PIK toggle. Start with: “Pretend I’m a new analyst and you’re walking me through this for the first time. I want to actually understand it, not just the definition. Use a real market example.”

Push back on anything glossed over: “Slow down — explain that part again.” The single best way to learn anything new in private credit.

Paste a working thesis on a current deal, then: “Find the three weakest parts of this thesis. Don’t be polite. Tell me what a skeptical IC member would say first.” At least one will be something you hadn’t considered. Update your thesis; save it in your log.

Headphones in. Open voice mode. Go for a fifteen-minute walk and talk through a deal you’re confused about, a presentation you owe, or a meeting you didn’t follow. Don’t try to be articulate — just talk. When you’re done: “Organize what I just said into a clear structure.”

Most analysts skip this. It feels weird the first time. By the third it’s the most useful tool in the kit.

Upload a live CIM. Run this sequence: (1) “The sponsor’s actual growth thesis in five bullets — the underlying claim, not the marketing.” (2) “Every quantitative market claim, with page number, and which look softest.” (3) “Every EBITDA adjustment — amount, rationale, and whether it survives committee.” (4) “What’s not in this document that should be?”

That last question is the most important. CIMs are sales documents; the interesting story is what’s missing. On Copilot, do this inside a deal Notebook so the context carries forward.

Open a model, click the Copilot button, run: (1) “Walk me through this — structure, tabs, and the three or four assumptions that drive everything.” (2) “Base-case revenue growth — where is it set and where does the number come from?” (3) “Drop EBITDA margins 200bps in years 2–5; what happens to leverage and DSCR each year?” (4) “The one slide you’d put at the front for committee — draft it.”

Not on Copilot? Paste a simplified version of the assumptions into Claude/ChatGPT and run the same logic in text.

Paste any email or memo you wrote last week: “Cut this to three sentences for an IC member with 30 seconds. Don’t lose the recommendation or key risk.” Decide if it’s better than yours. Then flip it: write a one-line recommendation and have it expand into a full paragraph with supporting logic. You’ve now practiced both directions.

Paste the financial covenant section from a deal you can use: (1) “Translate this into plain English — what must the borrower do, and what happens if they don’t?” (2) “What’s borrower-friendly and what’s lender-friendly?” (3) “Compare to standard market terms for a sponsor-backed unitranche of similar size — loose where, tight where?”

The goal isn’t to replace your reading. It’s to be faster and catch what you’d skim past.

Pick a deal with non-trivial concentration. Pull everything — CIM, model, management deck, call notes — and ask: “Find every claim about customer concentration. Where do the documents agree and where do they contradict? What’s the strongest case the sponsor makes that it’s okay, versus the strongest case I’d make in committee that it isn’t?”

A synthesis drill — the afternoon’s work that now takes ten minutes.

Paste a draft you’ve written: “Make this worse — like a junior analyst trying too hard. Then make it better — sharper, more direct, less hedged. Show me both.” The “worse” version surfaces bad habits you didn’t know you had; the “better” gives you a stretch version of your own voice.

Not Agent Builder yet — the rehearsal. Write a Deal Screener context block: your firm’s actual investment criteria (sectors, EBITDA size, leverage, hold, geography), the three things your firm cares about most, the two immediate disqualifiers, and the output format you want.

Save it. Paste it atop a new chat and feed in the next teaser that lands. You just built a v0 agent — and the spec for your future real one.

Take the deal you’re closest to writing up. Give the AI everything — notes, CIM extracts, model output, thesis — then: “Draft a first-pass IC memo: (1) deal in two sentences, (2) why we like it in three bullets, (3) three biggest risks with how we get comfortable, (4) what we’d ask management next, (5) recommendation. Direct tone, no marketing language, no hedging.”

It’ll be 60% there — four hours saved. Edit the rest yourself. What you fix is where your judgment lives, and where your real value is.

Paste any work product: “Read this as the most skeptical PM at my firm — 25 years in, has seen every flavor of bad deal. What three questions will they ask that I haven’t answered?” Add the answers to your draft. You just ran your own pre-IC dry run.

Pick a portfolio company. Pull the latest compliance cert, quarterly financials, and recent notes. Ask: (1) “Summarize the quarter in five bullets — trajectory, watch-list item, what changed.” (2) “Compare to the original thesis — ahead where, behind where?” (3) “Draft a two-paragraph update for my PM. Direct, no padding. Flag the one thing they should care about most.”

One of the highest-leverage uses of AI in private credit.

Before anything else: “Summarize everything that came in overnight. Group by deal and portfolio company. Flag anything that needs a response from me today.” Then in Teams: “What did I miss in meetings yesterday that affects something I’m working on?” Do it three days running; by day three you’ll wonder how you started without it.

Pick a meeting tomorrow. Paste the subject, the 3–5 docs that’ll be referenced, and your role, then: “Build me a one-page prep doc: the four things I should know walking in, the three questions I should be ready for, and the one question I should ask that nobody else will.” Print it. Bring it. See how it feels.

Pick a sector you might cover next quarter: “Build me a primer: industry structure, the three or four large players, typical unit economics, how recessions hit it, what private credit lenders worry about, and the questions I should always ask management here.” Then drill into the part that surprised you most with three follow-ups. Save it — the prep is done before the deal lands.

There’s a problem you’ve been avoiding — a deal you can’t decide, a memo you can’t start. Headphones, twenty-minute walk, voice mode: “I’m going to ramble for ten minutes. Don’t interrupt — just let me talk. Then pull out the three things I’m actually worried about, what I think I should do, and the thing I’m avoiding. Be honest.”

This drill resolves a real problem in real time. Not metaphorically. Actually.

Open your drill log and paste the whole thing in: “Read all of this. What patterns do you see in how I work and where I struggle? Where am I getting better, where am I still doing generic moves, and what one habit should I build over the next 30 days?” Read it carefully. Argue with it. Update your habits.