Value-creation teams have spent the last year watching AI move from a line in the investment thesis to a line item in nearly every portfolio company's budget. The opportunity is real, and so is the risk of treating each portco's AI effort as its own one-off project: a dozen companies, a dozen vendor relationships, a dozen different answers to the same questions about spend, data governance, and which models are even allowed. This is a practical playbook for capturing AI value across a portfolio without asking every portco to solve governance from scratch.
Start with the quick wins, but design past them
Every portfolio company can find a first AI win quickly: drafting support responses, summarizing diligence documents, speeding up a finance close, generating a first pass on a deck. These wins are genuinely useful, and they are also the easy part. The more durable value shows up later, once a portco has settled into which workflows actually benefit and has the spend and adoption data to prove it. A playbook that stops at the quick win treats AI as a one-time productivity bump instead of an operating improvement that compounds across the hold period.
Standardize the layer, not the model
The instinct to pick one model and mandate it across the portfolio is understandable and usually wrong. Model quality shifts every few months, and the model best suited to a support team's tone is rarely the one best suited to a finance team's spreadsheet work. What is worth standardizing across portcos is not the model; it is the layer that sits in front of every model: how spend is attributed, which models are allowed for which categories of data, how provider credentials are held, and how usage is measured. Standardize that layer once, across the portfolio, and every portco gets the same governance foundation while staying free to choose whichever model actually serves its own workflows.
- A single allowlist policy for which models can touch which categories of data, adjusted for each portco's regulatory profile.
- One method for attributing AI spend to a team, project, or workflow, so a report means the same thing at every portco.
- Centralized handling of provider API keys, rather than credentials living wherever an engineer first set up a pilot.
- A shared view of what each model's data residency, retention, and training practices actually are.
Measurement is what turns a mandate into a value-creation story
A portfolio-wide directive to use AI produces activity, not necessarily value, and activity is hard to defend in an investment committee meeting. What separates a mandate from a value-creation initiative is measurement: knowing what each portco spends, on which models, for which workflows, and connecting that spend to a qualitative improvement, a faster close, a lighter support queue, fewer hours on a first draft. Without that connective tissue, AI spend looks like an unexplained cost line. With it, AI spend becomes evidence in the value-creation report the deal team is already building for every other initiative.
The two failure modes to design against
Left alone, portfolio companies drift toward one of two problems. The first is reinvention: each portco builds its own governance from scratch, at its own pace, with its own blind spots, so the tenth portco is solving the same allowlist and budget questions the second one already solved. The second is shadow AI: without a sanctioned, genuinely good option, individual employees at a portco sign up for tools on personal cards, and sensitive customer or financial data starts moving through accounts nobody at the firm or the portco can see. Both failure modes have the same fix: give every portco a governed starting point rather than a blank page.
A sequence that works from diligence through the first 100 days
During diligence, ask what AI tools a target already uses, who pays for them, and whether any sensitive data has left the building through an ungoverned account; this is fast to check and often reveals more about a target's operational discipline than the answer itself suggests. In the first 100 days, stand up the governed layer before expanding access broadly, so the portco's AI usage starts inside a spend and data-governance framework instead of retrofitting one after adoption has already spread. From there, measurement across the hold period is what turns a portfolio-wide AI initiative into a line in the value-creation report, rather than an assumption inside it.
How Switchboard helps
Switchboard gives a deal or value-creation team one governed AI layer to stand up at every portfolio company, rather than reinventing governance a dozen times over. Each portco's teams reach ChatGPT, Claude, Gemini, and open-weight models through one login, with per-team budgets, model allowlists, and centralized provider-key management built in from day one. Every request is attributed to a person, a team, and a portco, so spend rolls up into one view the deal team can actually use, and each model's data residency, retention, and training practices are surfaced so governance decisions at every portco are informed rather than assumed. The result is an AI initiative that shows up in the value-creation report as measured, governed progress, not as a collection of unrelated portco pilots.