Hello, {{hc_sub_firstname | friend}}! Param here.
Most companies racing to add AI to their products are asking one question: "Does our software have AI features?" Constellation Software – the $37B compounder that has acquired more Vertical Market Software (VMS) businesses than any company in history – is asking a different question entirely.
That inversion is the whole story. And it's a field manual for every holdco investor underwriting acquisitions in 2026.
CARVE-OUTS
The Lynnfield Investor Program
At Lynnfield, we acquire cash-flowing businesses in the $1M–$10M EBITDA range and offer co-investment opportunities to qualified investors. Join our investor list to receive deal flow as we evaluate new acquisitions.
Continue reading the main story below ⬇️
The Disclosure Map
At Constellation's May 15 annual meeting, president Mark Miller laid out what the company has been building across its portfolio over the last 18 months.
Four disclosures, taken together, add up to something worth studying.
First, the AI accelerator programs. Constellation has pushed AI capability development down to the business unit level – thousands of developers across operating groups trained in AI-augmented coding, with more teams scheduled through 2026. Joint programs across operating groups. Events in Denver, in the U.K. The model is consistent with everything Constellation has ever done: no central mandate, no CIO deciding what platform to use. The business units closest to their customers run their own experiments and share what works. Mark Miller put it plainly at the annual meeting: "There's no CIO who decides what AI platform we use."
Second, the explicit AI lens on acquisitions. Constellation has added AI exposure as a documented criterion in its underwriting process. Targets are now assessed for AI disruption risk and potential AI upside, modeled accordingly. This is not a philosophical position – it's in the capital allocation process, applied to every deal.
Third – and this is the part most analysts find disappointing and most operators find clarifying – Constellation has produced no significant new revenue from AI. Miller said it directly: the company has not "really seen a lot of new revenues from that." The thesis is not AI as a revenue line. The thesis is that the durable moat is "deep vertical knowledge, a genuine understanding of customer workflows and processes, the data inside their solutions and the trusted relationships they built." That sentence came from the Q4 2025 earnings call and was reinforced at the annual meeting. It is worth reading twice.
Fourth, the PEMS thread. Mark Leonard's step-down from the board, effective May 15, coincides with Constellation formalizing its Permanent Engaged Minority Shareholder (PEMS) strategy. The first position: a 12.7% stake in Sabre Corp, acquired for ~$86M, with a board seat secured through a governance agreement. The AI lens applies to PEMS targets too – Sabre operates in global travel technology, a vertical with its own AI exposure questions to work through.
The Underwriting Reframe
Here is the shift that matters for acquirers outside Constellation's vertical software world.
Most buyers in 2025–2026 have been asking a binary question: does this target use AI? Does it have an AI roadmap? Constellation is asking something more useful: how exposed is this business to AI disruption, and is that exposure priced into the deal?
Three dimensions of exposure:
Automation risk. Some businesses have a work product that AI can replicate or accelerate. A bookkeeping firm delivering a monthly close package faces direct price pressure from AI-augmented tools today. A residential pest control operator does not – the core work is physical, route-based, relationship-driven. These two businesses require completely different underwriting logic, and conflating them is how acquirers get mispriced.
Data defensibility. Some businesses sit on proprietary transaction histories, customer records, and workflow data that become more valuable as AI tooling improves. A vertical software business with 20 years of niche industry data has a fundamentally different AI posture than a general-purpose service business. The analog for Main Street: a pest control operator with 15 years of route and outcome data has a data asset worth preserving. An HVAC installer with undocumented job histories does not.
Customer adoption rate. Some customer segments are adopting AI quickly – mid-market and enterprise buyers are pulling forward. Main Street customer segments – small construction firms, family-owned restaurants, independent clinics – are years behind. A business serving slow-adopting customers gets to compound through the AI cycle. A business serving fast-adopting customers has to race it.
Constellation scores acquisition targets across these three dimensions before bidding. It is a simple framework. The point is not to score every business as high or low. The point is to score consistently – and price the difference.
How the Lens Translates to Main Street
The natural objection here: Constellation buys software businesses. Lynnfield buys fence companies and HVAC operators. Does any of this translate?
It translates – and the asymmetry actually favors Main Street acquirers.
For software businesses, the AI lens is mostly defensive. Most software targets carry some AI exposure, and the underwriting question is how much of that risk is already priced in. For Main Street businesses, the lens is mostly offensive. Direct AI disruption risk is lower. But the operational upside from AI deployment inside these businesses – scheduling, dispatch, customer communication, financial close – is real and largely untapped by the owner-operators selling them.
Walk through three sectors from Lynnfield's hunting grounds:
A residential HVAC operator: low automation risk (the work is physical), low data defensibility (limited proprietary records), slow customer adoption (homeowners are not AI buyers). Score: low exposure across all three dimensions. Underwrite without AI adjustment. The opportunity is internal – AI for dispatch optimization and customer follow-up is pure margin lift, no thesis change required.
A specialty business-to-business (B2B) distributor: medium automation risk (purchase order processing is automatable), medium data defensibility (transaction history has value), medium customer adoption (mid-market buyers are moving). Score: watch which customers might eventually self-serve via AI tooling. The opportunity is longer-horizon – the proprietary transaction data becomes strategically interesting over time.
A regional bookkeeping or accounting firm: high automation risk (the core deliverable – a monthly close – is exactly what AI tools are being built to replicate), faster customer adoption (small business owners are the first to try AI bookkeeping products). Score: high exposure. Underwrite at a meaningful discount, or pass. Constellation's AI lens would treat this as a higher-hurdle target.
CARVE-OUTS
The AI Exposure Scorecard for SMBs
Before Constellation bids on a target, every acquisition gets scored across three dimensions. Here is the same scorecard translated for Main Street.
Low exposure across all three: underwrite as normal. No AI premium, no AI discount.
High exposure on even one dimension: either document the moat, adjust the price, or pass.
The asymmetry worth noting: most Main Street businesses score low. While software investors are stress-testing every target for AI disruption, the HVAC operator and the fence installer are quietly compounding. The AI lens does not threaten the Main Street thesis. It only confirms it.
Continue reading the main story below ⬇️
The Lynnfield Framework
Lynnfield is applying a version of this lens in 2026 underwriting.
We use three components:
Every letter of intent (LOI) now gets a brief AI exposure assessment – automation risk, data defensibility, customer adoption rate. Targets with high exposure require either a documented moat, a price discount, or a pass. Targets with low exposure are underwritten as before, with no AI premium and no AI discount.
Operationally, we are rolling out a small set of high-leverage AI deployments across portfolio businesses – dispatch, customer communication, financial close. These are not thesis changes. They are margin improvements that compound quietly, the same way everything else we do compounds quietly.
The third piece is the least urgent but potentially the most valuable: the proprietary data sitting inside our portfolio businesses. A business with 15 years of customer history, route records, and job outcome data has an asset that becomes more interesting in an AI-enabled world. We are not selling that data. We are organizing it and preserving it as a long-horizon option.
Thanks for reading!
The single most useful thing Constellation disclosed in 2026 is the thing the market did not particularly want to hear: there is no significant new AI revenue yet, and the durable moat is still vertical knowledge and customer relationships. That moat is the same thing Lynnfield underwrites on every deal – businesses with deep customer ties, undocumented institutional knowledge, and cash flows that have compounded for decades without anyone paying much attention.
Constellation, at a market cap of ~$37B, reaffirms that thesis and is one of the more useful validations the permanent capital strategy has gotten this year.
Talk soon,
Param
P.S. If you want to go deeper on the Constellation story, the original profile ran last week. You can read that story and all our other pieces via our archive.