How to Build an AI Business Case for Your LP Presentation

The majority of AI pilots in PE-backed companies I saw didn’t die in implementation, but in the business case.

An interesting demo, a few enthusiastic operators, a vague "this could be big" — and then it stalls at the investment committee because the numbers don't speak the language of value creation. EBITDA uplift. Payback period. Risk-adjusted return. Replicability across the portfolio.

And the risk is real. A pilot needs clean data and security hardening for anything SOX-relevant or customer-facing and touches live operations while it's being built. The IC knows this, so a demo isn't enough. They need a case built like any other value creation decision.

Building the case well once gives you the template every portco in the fund can reuse, and eventually the operational story that backs your strategy. Here are the four steps we use with operating partners and value creation teams to build one that holds up under scrutiny.

1. Start with the right process

Not every process is a good first pilot. The top choices share three traits:

• High volume
• Rules-heavy
• Tolerant of a human-in-the-loop checkpoint

The common entry points are predictable. Process and areas we suggest to start with are: order processing, invoice handling, AR collections with direct DSO and working-capital impact, CRM hygiene, shipment communications, customer and vendor onboarding, compliance reviews, L1 customer service, sales ops augmentation, demand forecasting, financial reconciliations.

Three more deserve their own line, because they hit margin directly: Pricing optimization, so a portco can raise prices without losing customers. Margin protection, by monitoring the costs that feed into pricing before they erode it. Procurement streamlining, so a company uses the buying power the group gives it, central purchasing instead of local, sometimes discounts brokered at the PE-portfolio level.

These rarely show up on the first list, and they're often where the cleanest money sits.

Filter your shortlist with two questions. Does it have a direct impact on cash or EBITDA? And will it lay the data and team foundations for larger initiatives later? For a fuller method on where to look for ROI, we wrote a separate guide.

2. Shortlist and model the ROI

Before committing scope, rough-cut the economics for each candidate. You need:

• FTE count on the process today, fully loaded cost
• Cycle time and volume
• Expected productivity recovery post-deployment (40–60% is a defensible range for most back-office use cases, so don't inflate it)
• Run-rate savings, one-time implementation cost, payback period

One thing teams forget: total cost of ownership and the effort the portco itself has to put in to get there.

An important framing to remember: don't model this as headcount elimination. The case holds up better, with the IC and with operators, when you show capacity redeployed to higher-value work and a human-in-the-loop kept for exceptions and oversight. It also de-risks the change management conversation, which is where pilots often struggle. The frame to sell is: grow without hiring.

3. Pick the winner and map the target-state workflow

Once you've picked the process, document current-state and target-state side by side. Systems involved, data flows, handoffs, who owns which decision, where the AI layer sits, and what it does not do.

You'll likely need a pragmatic execution partner for this, and it starts with workshops and understanding the business as it really runs. Don't cut that corner. AI adoption takes time, energy and money, and there's no silver bullet here, only gradual change you commit to.

The result is the artifact the IC will scrutinize. It shows you've thought past the demo.

4. Scope the roadmap, and the scale story

Build the rollout as a phased program:

• Pilot
• Validate
• Expand

Tie each phase to a decision gate with pre-agreed success metrics, so every gate is a defensible stop-or-proceed call. Only then extrapolate. This is the part most first-time cases miss.

One portco, one process, saves $X and lifts EBITDA by Y basis points. The same process exists in eight other portcos in the fund. Two adjacent processes in the same portco add another layer. Across the portfolio and across markets, the compounding effect will turn one pilot to the fund-level story.

A single-portco case is a pilot. A fund-wide case is a thesis. Don't bury the portfolio number on the last slide. Make sure it is on the cover of the LP update.

##What we've learned about cases that get approved

A few things hold true across the cases that clear IC: Target payback inside six months for the first pilot. Longer windows trigger harder questions and slower approvals, and the political cost of a drawn-out review usually outweighs the scope you're defending. Pick something unglamorous. Don't let the first pilot depend on a CRO changing how their team sells. Pick the boring, high-volume, operator-controlled process. Credibility compounds. Put every number in EBITDA terms. OPs and ICs read a case in margin. Convert the savings into basis points of margin expansion and run-rate dollars annualized into EBITDA before it reaches the committee. State the critical path. Name what has to be true for the projected number to land: data readiness, integration access, change management bandwidth. Expect the data to disappoint. It's rarely as good as anyone thinks, and it's never as good as they say. No AI produces good outcomes on bad data, so budget for cleanup. Have patience. The first steps in the right direction tend to reveal the gaps you'll need to fill. Once they're filled, everything after moves faster. Start with the vision. Insist on AI readiness and adoption as a priority — it'll be a key driver of productivity in the coming years. All the basic jobs that bring questionable value will be replaced by AI. Lead that way.

And one myth worth debunking: that rebuilding the data platform first is a must. You can pick an isolated process that delivers meaningful results without revamping the whole data foundation. Start with a pilot that doesn't need strategic CAPEX. Prove value on a few instances, and you'll earn the buy-in and budget for the large-scale projects.

Key takeaways

• The business case is the gating step. Build it once, reuse it across the portfolio.
• Start with an unglamorous, high-volume, rules-based process that touches cash or EBITDA.
• Model conservatively, including total cost of ownership and the portco's own effort.
• Lead with the portfolio-scale number. A pilot is one portco; a thesis is the fund.

If you're building the first AI business case for your IC, or standardizing a playbook across the portfolio, we help value creation teams pick the right starting process, model numbers that hold up, and scope the pilot end to end. Contact us to book a discovery call.