For Private Equity Firms & Operating Partners
Vertical AI Value Creation for Portfolio Companies
Most portfolio “AI initiatives” are copilot pilots that never touch EBITDA. We help PE firms find the workflows where AI genuinely moves margin — and build the proprietary, defensible capability there. We know it works because we run our own vertical AI ventures.
Book a Portfolio AI Discussion Read: How We Built Our Own AI →
The problem with “AI transformation” in a portfolio
AI theater is everywhere
Every management team now has an AI slide. Chatbot pilots, copilot licenses, an “innovation workshop” — activity that photographs well in a board deck and moves nothing in the P&L.
Horizontal tools don’t differentiate
If a capability arrives via an off-the-shelf license, every competitor has it the same quarter. Generic AI raises the floor for the whole industry; it doesn’t re-rate your asset.
Diligence can’t tell assets from demos
Targets increasingly pitch “AI-enabled” operations. Separating a real data-and-workflow moat from a wrapped API demo takes people who have built both.
Big consulting isn’t built for this
Strategy decks and 18-month roadmaps fit neither a hold period nor a mid-market portco budget. You need working software inside a quarter, with EBITDA math attached.
The vertical AI thesis — why operating companies win
The durable AI opportunities in a portfolio are rarely “add a chatbot.” They are vertical: AI built into one economically decisive workflow of one specific industry. And PE-owned operators hold the three ingredients that matter most:
Proprietary workflow data
Decades of quotes, claims, drawings, work orders, pricing history, and outcomes — the training and evaluation asset no foundation-model lab and no startup can replicate.
Existing distribution
The customer relationships a vertical AI startup would spend five years earning, you already own. Capability shipped into that channel monetizes immediately.
Priced, measurable workflows
In specialized businesses, hours and error rates convert directly to margin. That makes AI impact measurable — and makes the build-vs-buy decision a financial one, not a fashionable one.
Done seriously, the value shows up twice: operating margin during the hold, and multiple expansion at exit — a distributor or services business that owns proprietary AI capability in its niche trades differently than one that rents generic software.
We’re operators of vertical AI, not observers
Kamna Ventures builds and runs its own AI ventures. Aginera, our construction-takeoff platform, turns contractor drawing sets into bills of quantities using a hybrid of frontier models and our own systems — including RouteNet, a proprietary, patent-pending computer-vision model we built when every off-the-shelf approach failed on the workflow that mattered most. Its inbound growth engine — intent-specific landing pages and free tools — grew weekly signups roughly 10× in one summer with zero paid acquisition.
That end-to-end experience — wedge selection, API-first pragmatism, knowing precisely when a proprietary model is justified, and building the acquisition engine around it — is what we bring into your portfolio. The full story is here.
How we engage with PE firms
Portfolio AI screen
A fast, structured pass across portfolio companies (or a diligence target) to rank where AI can actually move EBITDA — scoring workflows on data ownership, error cost, hours burned, and defensibility. Output: a ranked map with build-vs-buy calls, not a vision deck.
90-day proof inside one portco
We build working software on the top-ranked workflow — frontier models where they suffice, custom capability only where it’s justified — integrated with the systems the company already runs, with baseline and impact measured in operating terms.
Scale & moat
Roll out across the company (and sister portcos with the same workflow), formalize the data asset, and where the economics support it, convert the capability into durable IP — the playbook we followed with RouteNet.
Frequently asked questions
How is this different from hiring McKinsey or Accenture for AI strategy?
We ship working software, not strategy documents. Engagements are sized for mid-market portfolio companies, run in 90-day cycles, and measured in operating metrics — hours removed, error rates cut, margin recovered. And we’ve built and run our own vertical AI products, so recommendations come from operating experience rather than frameworks.
Do our portfolio companies need data scientists on staff?
No. The pattern that works in the mid-market is a small external build team working directly with the people who run the workflow. Where lasting internal capability is a goal, we structure the engagement to hand over maintainable systems and train the team that inherits them.
When does building custom AI make sense versus buying tools?
Buy (or call an API) whenever the generic stack meets the bar — which is most of the time. Build only where the workflow is economically decisive, the failure cost of generic AI is high, and your data advantage is real. Our assessment makes this call explicitly per workflow; owning a moat you didn’t need is as costly as renting one you did.
Can you support diligence on “AI-enabled” targets?
Yes. We assess whether a target’s AI story is a genuine asset — proprietary data, evaluation discipline, defensible integration — or a thin wrapper on rented models, and what it would cost to build the real thing.
What sectors do you work in?
Our deepest operating experience is in industrial and built-environment verticals — construction, engineering, manufacturing, distribution — but the wedge-selection and build discipline transfers to any specialized industry with priced workflows and proprietary data.
Bring us one portfolio company
Send us the portco with the most manual, priced, error-sensitive workflow you own. We’ll tell you honestly whether AI moves the needle there — and exactly how we’d prove it in 90 days.
Book a Portfolio AI Discussion For your operators: the CEO’s view →