October 30, 2026 | By: Jamie Reingruber, Partner –  Technology, Data & AI Practice

Infrastructure used to be an asset class tech services providers could ignore

For most of its life, infrastructure was the “Steady Eddy” of private markets. Toll roads, airports, regulated utilities, P3 projects. Long contracts, inflation-linked cash flows, a cash yield that arrived every quarter, no surprises. Consistency was king, and the post-close playbook was built to protect value, vs. create it. As one investor put it: “I remember a diligence where the mandate was, literally, to go look at steel beams on a transmission tower portfolio in the mountains. Technology (shockingly) never came up.”

As the definition of infrastructure has expanded, so have the ways infra investors can use technology to create value. Services businesses, operating companies, and asset-light platforms now sit alongside toll roads and utilities in the same fund. An infra investor told our team last week that an animal-tagging business sits in his portfolio. Nobody blinked.

A change in philosophy is under way

The arrival of these operationally intensive businesses, combined with the rapid advance of AI, has caused infra investors to re-examine how they operate their core assets too.  Investors are finding that the technology initiatives that drove value in their core-plus¹ investments, such call center automation, apply to toll roads and towers too.

Not every infra business needs the full treatment. That said, even a tower company with 1,000 sites and an exceptionally lean back office may have a use case: its several hundred leases may span different vintages and carry amendments nobody has read in years. An agent that extracts every escalator, power and property-tax pass-through, and equipment-add clause, then checks each against what is invoiced, could surface revenue leakage in weeks.

Or take a fiber company

A fiber operator sells internet to households. It may do so as a monopoly or a duopoly, at least until more providers build in the same town and both churn and pricing come under pressure. The product does not change, the customers rarely leave, and the business plan is straightforward: build past homes, sign them up, keep them. On the surface, there are few value creation opportunities. Underneath, almost every key metric the business tracks can be improved with targeted investment in data & AI infrastructure.

  • Capex per premise passed: Economics stop working once the cost per home passed crosses a certain threshold. Combining household, competitor and build-cost data in one view lets the operator rank blocks by expected payback before building.
  • Penetration: The build only pays back if enough homes passed become subscribers. Joining the premise list to demographic and competitor data tells sales which households to work first and which blocks are falling behind the curve.
  • Average revenue per user (ARPU): Joining the subscriber files to property and demographic data shows which households are best candidates for targeted upselling via upgrades and bundle offers.
  • Churn: Service tickets, speed complaints, truck rolls and payment history in one place lets a model score every account for churn risk and target retention offers at the highest-risk customers.
  • Cost to serve: Clean access to billing and ticketing data lets an AI agent (Decagon, Sierra, or home-built) handle routine calls (outages, bills, plan changes, router resets) end to end, so the human agents are left with the calls that need judgment (disputes, cancellations, sales, etc.).

What we find when we show up

We spend a lot of time in rooms with infrastructure deal teams and their portfolio companies mapping how the business runs, where it hurts, and what fixing it is worth. As a result, we see three patterns frequently:

  1. A high-ROI data project often unlocks the more sophisticated AI use cases. At an infrastructure services company, we cut DSO by ten days with a simple time-to-invoice alert that flagged completed work nobody had billed, which unlocked a meaningful near-term liquidity improvement. That one use case paid for the platform and made the more interesting ones (predicting client credit risk, recommending payment terms) viable.
  2. Buying tools is easy; process and capability development is much harder: A renewables operator’s rush into LLM tools left business-critical apps built on Codex with no token spend controls, governance to ensure the tools were of production quality, and a sprawl of overlapping subscriptions. We established an AI governance process to mitigate significant legal risk, found considerable savings by rationalizing the tool, and built a market intelligence hub, an agentic workflow designed to rapidly ingest industry and regulatory documents and escalate material updates to the strategy team, helping ensure critical deadlines, such as interconnection queue submission windows, were not missed.
  3. A strong enterprise data model is often critical for realizing data & AI ambitions: An acquisitive waste hauler running contracted and spot lanes across several affiliated businesses had no common way of describing a load, a lane or a customer, so there was no single view of the business. Before any AI use case could be considered, we built a standardized data model that reconciled every load across the affiliates and made margin visible by lane, contract and spot alike. Repricing the weakest lanes to their peers was worth roughly $1M of annual EBITDA, and the same model now sets the foundation for route optimization and for deciding which collections produce the best economics.

There is no silver bullet

Time-pressed infra investors ask us some version of the same question: is this worth it, and where do we start? The honest answer is that the appetite for technology-driven returns in this space has never been higher, and the number of businesses that have captured them remains small.

This gap does not close by buying software. Identifying the highest-ROI use case requires a working command of how the business makes money, what core data assets the company possesses, where processes break, and what the business is seeking to accomplish strategically.

To accelerate this, we often roll up our sleeves and run working sessions directly with infrastructure investors and their portfolio companies to map the process, find the pain, size the value, and leave behind a prioritized plan the deal team can actually run: which use cases first, what data has to be in place to support them, and who owns each one.

Toll roads, towers and fiber are technology companies, whatever sector label they carry. Plenty of factors will drive returns over the next few years, but the owners who understand what their data is worth, and what technology can add to it, will be better placed to capture them.

If you want to touch base on industry use cases or schedule a portco AI workshop, please reach out to Jamie and the Leo Berwick team to learn more.

¹ Core-plus: infrastructure assets that keep most of the contracted, inflation-linked cash flow profile of core infrastructure (regulated utilities, toll roads, airports) but carry some operating, market or growth exposure, such as fiber operators, services platforms and merchant-exposed energy assets.