On October 21, we brought finance operating partners, AI and technology leaders, portfolio company CFOs, and value creation specialists together at the Harvard Club of New York for the Finance & AI Value Creation Exchange.
Thank you to the sponsors who made the day possible: Stuut, our Diamond sponsor; Hyphenate, our Platinum sponsor; SAP, our Gold sponsor; and our Silver sponsors Transformworks, Rimini Street, Planr, and KPMG.
Thank you as well to our speakers, Ram Krishnamurthy of Hyphenate, Bill Bowman of WCAS, Ben Winter of Stuut, Lee Goldfarb of Veritas Capital, Robert Rosenwein of H.I.G. Capital, Max Beck of Transformworks, Kristy Khoury of Revelar Capital, Jarrad Berman of TZP Group, and Rebecca Granne of Shamrock Capital, and to every attendee who came ready to talk about what is and isn't working in their portfolios. We ran the day under the Chatham House Rule, so the views below appear without attribution.
One question came up in nearly every session: how do you get from an AI pilot that works to a result a board will accept? The answer had little to do with which tools a company picks. It came down to how the company redesigns the handoffs between its teams and systems, how it governs the data AI puts in motion, how a sponsor sequences initiatives across companies at very different stages of readiness, and how rigorously anyone measures results against a baseline. The room spent far more time on that operating discipline than on vendors, and this recap reflects that.
The Handoff Is Where Finance Transformation Happens
The opening session looked at what large-scale enterprise transformation can teach middle-market and PE-backed finance teams. One data point framed the rest of the day: implementing the technology took a couple of months, but the organizational change around it took years.
Sponsors should budget time and attention for any finance AI initiative with that ratio in mind. You can configure software in a quarter, but changing who owns a process, how exceptions get escalated, and what a controller trusts enough to sign off on moves at the speed of people.
The session also clarified where to look for value. Most finance leaders can name their bottleneck without much thought. The bigger problem usually sits at the handoff between teams or systems, where manual work and tribal knowledge pile up: an exception that passes from collections to cash application, or a close that waits on a reconciliation living in one analyst's spreadsheet. Automate the task on either side of that handoff, and you get a faster step inside the same process. The better question is whether a system can manage the handoff end to end, and you can only answer it by mapping the full process before choosing a point solution. Panelists also discussed accountability as AI takes on more finance work, including the evidence a CFO and the company's auditors should require before relying on AI-generated output.
One model raised in the session pushes the idea further. A firm builds its own AI pilots in-house, deploys them across its portfolio companies to see whether they hold up in live operating environments, and once they prove out, commercializes them with real portfolio results behind them. The upfront investment is significant, but the payoff is a capability that doesn't depend on outside vendors and an asset the firm can reuse across the portfolio.
The ownership argument is gaining ground outside the room, too. A recent Forbes analysis argues that portfolio companies should own their AI rather than rent it, noting that at an exit multiple of roughly 15x EBITDA, every $100,000 in recurring annual cost takes about $1.5 million off the sale value. Few firms have the engineering depth to build everything themselves, but our read is that the core idea, using the portfolio as a live proving ground, holds whether a sponsor builds, buys, or does some of each.
Security as a Condition for AI Value
Every finance workflow AI touches moves more company data, and AI creates value only when the business can trust the data underneath it. The second session treated security as a lever on enterprise value, and one framing landed with nearly everyone: a security function's job is to turn under-informed risk acceptance into fully informed risk acceptance. Every portfolio company accepts risk. The question is whether leadership actually chose to accept it, with a clear view of what data is involved, who can access it, and how it's used.
That framing turns a list of technical findings into a business plan a leadership team can sequence. Panelists talked through how to group security work into priorities, where AI-related exposure fits, and how to separate sensitive data from lower-risk information as companies feed more of it into AI tools. Handled this way, de-risking becomes an ongoing management discipline with owners and regular reporting, rather than a compliance checklist someone finishes the week before a board meeting.
The value case covers revenue as well as protection. Customer trust, business continuity, and the security terms written into enterprise contracts all affect whether a portfolio company wins and keeps business. For operating partners, the practical step is to ask for metrics that show actual risk reduction. A count of tools deployed tells you very little about exposure. Panelists also called for a closer working relationship between the CISO, the CFO, and AI leaders, since decisions about where data flows are now decisions about value.
When the Room Becomes the Panel
For the late-morning roundtable, we handed the conversation to attendees. Eight tables worked through three questions and reported back to the full room after each one. The questions moved from diagnosis to decision, and any sponsor can use them to review its own portfolio.
First, think of an AI initiative that hasn't paid off yet. What's actually in the way: the data, the people, the process, or the business case? Naming the constraint precisely matters, because each one calls for a different fix and a different owner. Second, carrying the morning's security discussion forward, which AI risks are your portfolio companies accepting today without anyone having decided to accept them, whether in data, vendors, access, or decisions a model is now making? And who should be making that call? Third, if your firm could fund only one finance AI initiative across the entire portfolio next year, which would it be, and how would you prove it worked within twelve months?
We'd argue that the last question belongs in every annual planning cycle. It forces a sponsor to rank opportunities, define success up front, and commit to a measurement window. Those three steps tend to get skipped when AI budgets are spread across a dozen small experiments.
Inside the Finance Function
The first afternoon panel went inside accounting, forecasting, receivables, and payables, where many sponsors have focused their early AI work.
Starting points vary widely. Some portfolio companies are still building data foundations, others are running pilots, and a smaller group has moved to broad adoption. Panelists covered what has delivered, where expectations fell short, and how they define success in business terms.
A key distinction was whether the value comes from a single tool, from automating a task, or from redesigning an entire workflow. That answer drives everything downstream, including the build-versus-buy decision, the choice of implementation partner, and how much change management the CFO and finance team must absorb. In the middle market, where a company may run finance on systems older than its current management team, partners who understand legacy ERP environments and industry-specific requirements carry real weight.
Recent research supports a sequenced approach. A Gartner survey of 160 senior finance leaders, released in late September, found that data extraction, payables and receivables automation, and report creation generally return value within nine to ten months, while forecasting and insight generation take longer to develop. Gartner cautions against letting quick wins crowd out the harder use cases. For a sponsor, that points to pairing early transactional wins, which build credibility with the finance team, with a deliberate plan for the analytical work that eventually changes how the business makes decisions.
Sequencing Across Uneven Readiness
The second afternoon panel took on the judgment calls that sit above any single initiative. A typical portfolio includes companies of different sizes, resources, and maturity, so a uniform AI program rarely fits.
Panelists discussed how operating and value creation leaders decide where to start, how they assess foundational needs in data, technology, security, talent, and leadership, and how they sequence AI alongside the operational improvements already underway.
The readiness gap is real and well documented. An IBM Institute for Business Value study of 1,500 finance leaders, published at the end of September, found that only 6% describe their function as transformation-ready, meaning AI is consistently embedded in workflows and decisions at scale. Most portfolio companies sit well back on that curve, which makes the sequencing decision more consequential than any single tool choice.
A few practical disciplines came through. Set ownership and a baseline before an initiative launches, or you won't be able to measure business impact later. Balance a repeatable portfolio playbook with company-specific execution, so the fund-level team can share what works without forcing it on a business that isn't ready.
Get CEOs, operating partners, functional leaders, and employees aligned on adoption, because that determines whether anyone keeps using a capability after the pilot ends. The panel also discussed how AI is changing finance and analytical work itself, and how firms can redeploy freed-up capacity and develop talent as workflows change. That last point ties directly to value: in our experience, capacity that gets freed up and never reassigned rarely shows up in the P&L.
From Diligence to the P&L
The closing panel followed AI from the first identified opportunity through to measurable financial impact. Successful initiatives start in different places. Some begin with a specific operating pain point, some with a technology investment the company had already planned, and some with an opportunity spotted in diligence before the deal closes. Tracing where past wins came from helps a sponsor spot the next one sooner.
Diligence got particular attention. An AI or operating partner involved before close can assess both the opportunity and the risk, from the state of the target's data to the exposure created by tools already in use. That work is only as valuable as its link to post-close priorities and a realistic first 100 days. When the findings sit in a report, and the plan gets written separately, the sponsor loses its earliest, best-informed view of the opportunity.
Build versus buy came back here from both the vendor and sponsor side, along with a frank discussion of workforce effects as initiatives scale, including when to redeploy, retrain, or reduce. The session's toughest question was also the simplest. When a board member asks whether AI actually moved the P&L, what do you put in front of them? Separating sustained impact from early experimental results takes the same baselines and ownership raised earlier in the day, plus enough patience to see whether a gain holds after the pilot team moves on.
Closing the Gap Between Pilot and Proof
Every session in New York came back to the same gap: the distance between an AI pilot and a number a board will believe. Closing it comes down to a handful of disciplines that surfaced all day: diagnose whole processes before choosing tools, treat data governance as part of the value case, sequence initiatives to match each company's readiness, and set baselines before anything launches.
Most of this work happens well away from the spotlight, and it depends on operating partners, portfolio leadership teams, and implementation partners working closely together. We believe the sponsors who build those habits now will have the clearest answer when a buyer asks what AI is actually worth in their companies.
-1-1.jpg)