How Source Capital Turned AI Training Into $270K in Portfolio Value — Then Shared the Playbook

The Executive Summary

• 3 programs delivered — one 12-week multi-firm cohort, two custom 6-week accelerators
• 123 leaders directly trained — 18 Source Capital team, 59 portfolio-company executives, 46 peer PE professionals
• 52 organizations impacted — 28 portfolio companies across 15+ industries, 24 peer PE firms
• 3 of 5 innovation competition prizes went to Source Capital teams
• Hundreds of thousands in deal risk avoided per flagged bad deal, caught before diligence spend
• 7+ hours saved per deal — 50%+ reduction on initial evaluation
• $270,400+ in documented annual portfolio value, conservative estimate
• $95,000+ in named individual wins across three portfolio-company CFOs and operators
• 82% of peer PE professionals now exploring AI training for their own portfolios

The short version for other PE firms: what started as one cohort seat became a managing-partner mandate, a portfolio-wide capability program, and eventually the AI training ground for two dozen peer firms — seven months, no pilot required.

The AI Gap Most PE Firms Haven't Closed

Before Source Capital spends a dollar on accounting fees or legal diligence, they now know if a deal has a problem. One AI-powered background screen — built inside a 12-week cohort — caught a federal litigation case a standard search missed entirely.

Sam Allen estimates that finding alone could have saved hundreds of thousands in diligence spend. That tool is now in production on every live deal.

Most PE firms don't have that. They have two unsolved problems sitting side by side: their own team needs AI capability — deal sourcing, due diligence, IC memos, portfolio monitoring — and every portfolio company needs it independently, across every industry they hold.

Every firm knows this. Almost none have a model for doing it.

• 97% of PE professionals have had no prior AI training, despite 85% already using ChatGPT in some form.
• Zero firms in our most recent PE cohort had a formal AI policy. 68% had nothing at all.
• Privacy is the #1 concern (41%) — deal data governance is the gate.
• 56% of PE professionals with AI access haven't actually used it in their work. The tool isn't the problem. Structure is.

(All data from client surveys across our PE cohort base.)

Source Capital's three-program sequence is what structure looks like. Each program produced both the capability and the internal proof point that unlocked the next one. Not "AI training" as a line item. AI capability as a compounding asset across an entire PE ecosystem.

"There are startups in every single vertical leveraging AI, running leaner, more nimble businesses. So we've got to play offense and not defense." — Tom Harbin, Managing Partner, Source Capital

How Source Capital Built the Playbook

Phase 1: A Principal Has a Realization

Source Capital entered the story as participants, not buyers. Eighteen of their team — including Managing Partner Tom Harbin, Principal Sam Allen, and Senior Associate Jack Mitchell — joined a 12-week multi-firm AI training cohort alongside three other investment firms.

The goal was simple: get everyone in the firm on the same foundational AI capability, from skeptic to power user, in one shared program.

What happened inside the cohort is what made everything downstream possible. By Week 2, Sam Allen was looking at the landscape differently:

"It was probably week two in the training when I realized, holy cow, there's a lot more capabilities and opportunity here that it would be a shame that companies we're looking to grow aren't taking advantage of these as well." — Sam Allen, Principal, Source Capital

That realization — this is not a search-engine replacement, this is a deal capability — was the spark. Twelve weeks later, Source Capital had two new tools in production:

Executive Screen Deep Dive (Sam Allen + Tom Harbin)

An AI-powered executive background screener that scans public records, historical filings, and legal databases before formal diligence begins. Now in production on live deals:

"There were a couple opportunities where it uncovered litigation from a while ago that unless you really know the specifics of it, you would've otherwise never found that out until much deeper in the actual deal process. You could be hundreds of thousands of dollars in accounting fees and legal due diligence prior to uncovering a massive issue." — Sam Allen

Background screening moved from last step before close to first step before diligence spend. One federal litigation case involving a name change was caught in minutes — a finding standard searches missed entirely.

Source Capital Deal Evaluation (Jack Mitchell + Ryan Johnson)

A custom GPT that automates initial deal screening against Source Capital's investment criteria — roughly 50% time reduction per deal, 7+ hours saved per evaluation — and standardizes output for partner review.

"It's kind of like building a car while trying to drive it. Every week it's a new model, a new capability, a new add-in." — Ryan Johnson, VP, Source Capital

"When I started I was like a one. I would say today I'd probably put myself as like a five. I still think there's lots of room for me to go."

By December, Source Capital had two production tools, firm-wide familiarity, and managing-partner conviction about where this was headed.

"I would say absolutely. It feels like it's a no-brainer." — Sam Allen

Phase 2: Expanding to the Portfolio — 59 Executives. 28 Companies. 15+ Industries.

Having seen what a structured program produced inside their own firm, Source Capital made a bet most PE firms don't make: they bought the same program for their portfolio company CEOs, CFOs, and COOs.

"Investing is only the very beginning of what we do and that we're trying to grow these companies and exit and create value over time." — Sam Allen

A full six-week custom accelerator — 59 senior leaders from 28 portfolio companies spanning manufacturing, real estate, home services, healthcare, equipment rental, specialty contracting, and ten more verticals.

The participant profile mattered:

• 33% CEOs and Presidents. 22% CFOs. 16% COOs. Decision-maker AI training, not middle-management AI training.
• Average AI comfort: 5.4/9. Wide spread.
• Only 4% had any prior AI training.
• 42% flagged privacy as their top concern. The curriculum led with governance and data-zone frameworks before any capability content.

Program innovations — built and documented live in session:

Greg Rosenstein — CFO, Mechanical Concepts (manufacturing). Built a full CFO agent system in Claude. Controller, FP&A, and Power BI agents as Claude Skills.

"I think the logic has been very sound, and the work has been very acceptable and very high-quality work."

Estimated annualized value: $30,000+.

Daniel Krieger — World Water Works (manufacturing). Built four production systems. Claude connected to bank + billing for payment matching. IT ticket monitor with missed-response alerts. Customer-sentiment classifier across phone + email. Built the company website in Claude and ranked it #1, #2, #3 on Google for target keywords.

"I've been able to get my website ranked 1, 2, and 3 on Google using Claude."

Estimated annualized value: $50,000+.

Martha Parker — CFO, CBS Enterprises / ServiceMaster Restoration (services). Used AI for percentage-of-completion accounting across hundreds of open construction jobs — simpler method, equally accurate, with storyboard explanations built for Source Capital reporting.

"Not only did it give me the right accounting standard, but it gave me its interpretation. Lots of time saved there."

Estimated annualized value: $15,000+.

Alan Jenkins — The Management Trust (real estate services). Used AI to reconstruct bank balance sheets from fragmentary public data — research no Google search could surface — saving 30+ minutes per analysis.

Chris Go — PropertyRate (fintech / real estate). Connected Claude to a read-only reporting database for on-demand SQL queries. Explored multi-model routing via OpenRouter.

Cheryl Ventola — Head of Operations, Kitchen Brains (manufacturing). Switched from Copilot to Claude mid-program and ran live platform comparisons that shaped what the rest of the cohort tried next.

Program outcomes — Executive Accelerator

Note on methodology. Conservative = 26 certified participants × 2 hours/week saved × $100/hour. Moderate = 40 active × 3 hours/week. Named individual wins ($95K+) are bottom-up, reported directly in session.

Phase 3: Source Capital Opens the Playbook to Their Peer Network — 46 Professionals, 24 Firms

Phase 2 would have been a strong story on its own. Phase 3 is what makes Source Capital's pattern unusual in private equity.

Source Capital extended the program to their peer network. Forty-six PE professionals from 24 lower-middle-market PE firms — Appian Way, Astria Elevate, Black Lake, Calidant, Canopy, Chartwell, Cornerstone, Crofton, EQ Equity Partners, FALX, Georgia Oak, Heritage Growth, Kilroy, Lockeland, LP First, Platt Park, Red Dog, Resurgent, Sea Pine, Seneca, Southstone, Third Century, Tricorner, Tuckerman.

Partners, Managing Directors, Principals, Analysts. The people who make investment decisions.

• 80% Partners or Managing Directors
• Average AI comfort: 4.4/10 — the lowest of the three programs
• 97% had zero prior AI training
• Zero firms had formal AI policies
• 38% already using Claude — a signal of Source Capital's influence already rippling through the network

"Folks just appreciate that we're trying to help. It's a very timely moment for a lot of those folks to level set." — Tom Harbin, Managing Partner

The adoption evidence started early:

Justin Chen, Partner — Kilroy Partners. Came in self-identifying as "a 1 out of 10." Three weeks in, he opened a session with "I'm 102 now" — describing a firm-level HubSpot integration he'd personally built.

Alex Mammen, Partner — Heritage Growth Partners. Lowest self-assessed comfort in the cohort at start. By Week 2 he was volunteering first, sharing a portfolio-company job-costing tool built with Claude. Perfect attendance through the midpoint.

John Ehlinger, Partner — Astria Elevate. Self-described "once bitten, twice shy" skeptic. By Week 3, the longest exchange in the program was his — nine minutes on AI quality, privacy, voice-clone risks, and local LLM options. Full skeptic-to-convert arc in three sessions.

What Phase 3 validates

Source Capital didn't have to sell their peer firms on whether AI training works for PE. They demonstrated it across Phases 1 and 2, and the network pulled in.

82% of participating PE professionals expressed interest in running AI training for their own portfolio companies. If even a fraction of 24 firms × 5–15 portfolio companies each convert, that's a cascade of 25–75+ additional programs from one seed cohort.

Firm-Level AI and Portfolio-Level AI

"We use AI to model, to underwrite, do research, to make decisions on investment… And I would describe that as the beginning of our journey. The second step is how are the companies where we're already invested… how do we help them on their journey? That's perhaps more important in terms of value creation." — Tom Harbin

Two lenses, two kinds of return:

Most PE firms have a plan for one or the other, not both. Source Capital built both in seven months.

The Pattern That Made It Work

Four things repeated across all three programs:

1. Managing Partners participated personally. Tom Harbin co-authored the tied-1st-place innovation competition project. CEOs and Presidents made up a third of the portfolio-company cohort. Partners and MDs made up 80% of the peer-firm cohort.

2. Security led the first week, every time. In all three programs, privacy was the #1 concern at 41–42% of respondents. The curriculum led with data governance, the Three Zones framework, and deal-data handling before any capability content.

3. Real tool-building, not awareness. Every program produced actual production systems — not slide decks, not frameworks.

4. PE-specific, not generic. Every exercise mapped to an actual PE workflow — CIM analysis, deal screening, IC memos, portfolio monitoring, LP updates.

What's Next for the Source Capital Network

• 82% of the peer-firm cohort are exploring AI training for their own portfolio companies — a potential cascade of 25–75+ additional programs.
• The Executive Screen Deep Dive and Deal Evaluation GPT remain in active production use.
• Portfolio-company adoption continues independently.
• A Source Capital AI Champions network is under discussion.

Is This Possible for Your Firm?

Source Capital ran this in seven months. Here's what it takes:

1. One cohort as ignition. Get partners and analysts through the same program at the same time.
2. A portfolio-company program within 90 days, before momentum fades.
3. A security-first curriculum built for deal professionals. Generic AI training burns trust faster than it builds capability in this industry.

Your portfolio companies are already competing against leaner, AI-enabled businesses. Don't wait, book a call with AI Operator today.