Hg booked $260M of EBITDA impact across 1,600 portfolio AI projects

Operators and investors,

Hg reports more than 1,600 AI projects across its portfolio carrying roughly $260M of budgeted EBITDA impact, a 5x increase since 2024. More than 10 of its businesses now derive over 10% of bookings from AI-enabled products (Hg).

Behind that number sits a headcount decision most sponsors have skipped: about 100 AI builders inside Hg Catalyst, its internal incubator, within a wider team of 150-plus AI value creation specialists.

I covered the business-case data in the July 15 edition, where 95% of funds reported AI initiatives meeting their targets. Underneath that headline, only 36% of portfolio companies use AI in daily operations and 7% describe it as fully integrated, and the gap between funds reporting success and companies running the software is where the money sits.

This issue covers 4 topics:

• What $260M of budgeted AI EBITDA is made of

• Why Hg staffed 150 people against the problem

• What Vista's adoption score changes for a portfolio CEO

• What a mid-market sponsor copies at a tenth of the scale

It is also written for operating partners deciding whether AI belongs at fund level or portfolio level, drawing on the AI value creation work for portfolio companies we run inside PE-backed businesses.

1. What $260M of budgeted AI EBITDA is made of

Budgeted EBITDA impact is a planning figure, and reading it as banked money would be a mistake. What makes it useful is the unit of measurement: Hg reports AI in the same currency as every other value creation lever, which forces each project to carry a number a board can test.

The disclosure breaks the total down by source. More than $130M comes from automation, product innovation, and efficiency gains, and a separate group of businesses is generating bookings from AI-enabled products sold to customers.

The two sources create value differently:

• Cost-side automation shows up as margin inside the existing revenue line

• Product-side AI shows up as bookings and can revalue the company at exit, because a software business selling AI-enabled products trades against a different comparable set.

A sponsor reporting AI progress as pilots completed is measuring activity, and one reporting it as budgeted EBITDA has made the shift Hg made, which turns AI spend into something an investment committee can price.

2. Why Hg staffed 150 people against the problem

The 36% daily-usage figure and the 150-person team explain each other.

Asking 30 portfolio companies to each source AI talent, evaluate vendors, build a data foundation, and ship into production means running that project 30 times with 30 different success rates. Most mid-market portfolio companies have no one whose job includes AI deployment, so the work lands on a CTO already carrying a roadmap or on a CFO carrying a close, which Hg’s structure removes.

A central incubator builds the capability once and pushes it down, which is the pattern across firms getting measurable output in 2026. The 2024 approach had each company running its own pilots; the 2026 approach has the fund building repeatable platforms and deploying them.

A central team sees 30 implementations, so it knows which use cases produce EBITDA and which consume a quarter. A single portfolio company sees 1 implementation and learns that lesson at full cost.

3. What Vista's adoption score changes for a portfolio CEO

Vista scores every portfolio company on AI adoption velocity and ties that score to future capital allocation decisions, reported by the Financial Times in November 2025, which is the mechanism most AI mandates lack.

A sponsor can instruct 30 CEOs to adopt AI and watch the instruction compete with every other priority in the operating plan. Attaching the score to capital allocation changes what happens when a CEO has to choose between the AI programme and a hiring plan, because now the choice has a financing consequence.

I would not copy the scoring model without copying the support that sits behind it. Scoring adoption while leaving each company to build its own capability produces a ranking of which portfolio companies happened to have technical leadership already, and it penalises the ones with the largest gap to close.

The business case is being proven at fund level and never reaching the operators who would run the software every day.

4. What a mid-market sponsor copies at a tenth of the scale

A fund holding 8 companies cannot staff 150 specialists, and it can still copy the structure at proportional cost.

One shared resource covering AI enablement for portfolio companies, whether a fractional AI lead or a small embedded team, removes the duplication that produces the 36% figure. The economics work at a fraction of Hg's scale, because the cost of 1 senior technical operator spread across 8 companies is lower than 8 separate hiring searches that mostly fail to close.

Require every AI initiative to carry a budgeted EBITDA figure, an owner, and a start month, in the same format as any other value creation initiative. Initiatives that cannot produce those 3 items are pilots, and they belong in a research line until they can.

Start with process automation for portfolio companies where the data is already clean and the workflow is already measured, because those convert inside a quarter while an unmeasured process needs a data build before any model helps.

This week, pull every AI initiative across your portfolio into one view with the operating partner who owns technology. Two questions to think about:

1. Can each initiative state a budgeted EBITDA figure, a named owner, and a start month, or does the list read as pilots with no financial line attached.

2. Are we asking each portfolio company to build AI capability separately, and what would 1 shared technical resource across the portfolio cost against that.

Score each answer red, yellow, or green. A red on the first question means the portfolio has AI activity with no way to price it, which is the position 95% of funds report their way out of and 7% actually operate from.

Mario

🧠 LLMs now surface your operating style from how you actually work. Asked to summarize his aptitudes, GPT and Opus returned labels like "leverage-seeking operator" and "scope frontloader" pulled from real briefs and iterations.

🔁 Continuation vehicles are becoming the default exit when timing does not cooperate. Verdane just closed a €635M continuation vehicle with Coller as sole lead to hold a performing portco rather than sell into a weak market.

📉 Bookings up 26% hid a gross-to-net conversion of 73% against a 90% thesis. Contract amendments, early terminations, and discounts left net revenue 14% below bookings with nobody escalating it, which is why gross-to-net conversion and pricing realization belong in every operating review.

📊 Sponsors rarely get a number they trust because every system reports a different one. Ask a portco for qualified leads and marketing says 340, sales says 287, and the CRM shows 412, a gap one reconciled definition closes before any new dashboard.

🎙️ Buy a facilities-management company and the real asset is the operational data underneath the cleaners, contracts, and schedules, the reframing most sponsors miss. - by Lee McCabe

🎙️ Asking whether people are "AI ready" is the wrong question, because the bottleneck in transformation sits with management long before the model. - by Praneet Gill

🎙️ In home services, every discount handed out is profit you never get back, which is how undercharging caps the margin a roll-up is buying. - by Justin White

🎙️ Longer hold periods are reshaping the CPO role, with $1 trillion in US dry powder waiting and multiple expansion no longer doing the work. - by Sophie Gorvett

🎙️ A buyer can expose in three weeks what a business has learned to live with for three years, since the current team already knows where the pricing and process quirks are buried. - by Douglas Pudney

🎙️ Most quality problems charge you over and over: one defect turns into sorting, rework, overtime, and premium freight. - by Alicia Motomura

🎙️ Every company has a rainmaker allowed to be cruel because of the revenue attached to them, and few name what that tolerance actually costs. - by Greg Montgomery

🎙️ Multiple expansion drove 37% of value creation in 2011-2013 deals and just 0.5% in 2020-2022, which is why higher rates made existing weakness impossible to hide. - by Kristian Graff Rasmussen

🎙️ Five new exits in Nigeria in 2026 saw TLG's Naira portfolio earn roughly 40% annualized in USD terms after doubling down through the 2024 currency low. - by Dr Ola Brown

🎙️ Most acquisitions lose customers when the commercial message changes even though the product did not, as one sales team calls the combined company premium and another positions it differently. - by Dieunor Michel

🎙️ At a $50M e-commerce portco squeezed on margin, the customer-service fix came from raising capacity per head rather than adding headcount, which freed budget to reinvest in growth. - by Sereena Pathiratne

🎙️ Every PE firm already has an expert bench sitting in its partners' inboxes, a network built over years that most firms never systematically tap. - by Kate Hopkins

AI's software-productivity gains now force an operating-model rethink. Astorg reports engineering productivity gains of 3x and up to 10x on cycle times, which moves the constraint from writing code to redesigning how teams, staffing, and margins are built around it.

Software multiple compression is now flowing into PE marks. HarbourVest data shows the software sell-off reaching portfolios as global buyout returns turn negative, raising covenant and refinancing pressure on any tech portco still carried at last cycle's multiple.

Private credit stress is showing up in the marks. Blue Owl cut its Loparex loan valuation as bankruptcy risk mounts, a signal for mid-market operators that lenders are actively repricing risk on leveraged borrowers.

Enterprise-software contracts are a governance exposure. Oracle Health executives were subpoenaed after a VA EHR contract's costs nearly tripled, a reminder to price implementation and change-management risk into any large software commitment before signing.

HappyPetHQ premium pet ecommerce brand: fast-growing brand on $973K revenue with authorized premium labels, clean ops, 10-13x Google and Bing ROAS, and B2B and DTC scale headroom. $763,000

At-home clear-aligner service: 10-year-old teeth-straightening service with virtual support and tailored treatment plans, throwing off about $286K a year in profit at a 3.0x multiple. $869,932

Federally-mandated ELD compliance SaaS for trucking: 4-year SaaS with $200K ARR, 300+ trucking clients, 80% margins, and a bundled TMS platform, pairing recurring revenue with a regulatory moat. $1,050,000

Automated AI-run digital agency: profitable agency with recurring clients, proven lead-generation systems, a trained team, and scalable fulfillment, running near $1.28M a year in profit at a 1.8x multiple. $2,273,879

For PE partners and operators seeking alpha

🌐 Scaling $50M - $500M+ mid-market companies with value creation through RevOps, data engineering, and WordPress. DevriX provides full RevOps consulting + delivery with GTM enablement for PE-backed portfolio companies, traditional tech, healthcare, finance, and professional service businesses pacing toward revenue growth initiatives. Our standard retainers between $10K and $60K include revenue lifecycle services for marketing and sales leaders, FP&A for financial teams, pipeline enrichment through websites and dozens of lead sources, automations and delivery integrations, CRO and ongoing testing, product delivery and platform integration solutions, and more through our consulting solutions.

🚀 1:1 Advisory retainers. Supporting operating partners, private equity funds, family offices, and mid-market executives in different capacities, from value creation through due diligence to portfolio digital GTM management in my async advisory programs via Growth Shuttle.

📈 GTM while scaling. European and international businesses can opt in for doola LLC and their “Business in a Box” model. Scaling founders can find smaller digital opportunities on Flippa. And additional opportunities across my investments can be found here.