Key Takeaways

  • Uncontrolled experimentation has driven software budget blowouts across private equity portfolio companies, forcing finance teams to treat artificial intelligence as capital deployment rather than departmental overhead.
  • Accordion Managing Director Justin D'Onofrio argues that tech licenses fail without parallel funding for data architecture, workflow changes, and organizational adoption.
  • Portfolio CFOs face sharper board scrutiny, requiring every line item to map to specific EBITDA expansion levers like revenue growth, headcount efficiency, or operating expense reductions.
  • Measuring AI success demands upfront operational targets, specifically tracking acceptance rates of automated outputs and capping diminishing returns.
  • Accordion developed the 8 Steps to Building an AI Budget to give operating partners and CFOs a structured financial model for software governance.

The Accordion 8 Steps to Building an AI Budget

Step 1: Know Where Your Sector Sits

Look at industry trends to understand where value capture is occurring for AI in your specific vertical and function before sizing budget numbers.

Step 2: Tag Every Dollar

Tag every AI expenditure across three dimensions: an assigned organizational owner, an EBITDA value lever (revenue increase, productivity gain, or cost takeout), and an overarching program approved through an ROIC process.

Step 3: Fund the Data Foundation

Treat foundational data platforms and master data management as distinct, necessary investments so quantitative and qualitative data repositories are clean and reliable for AI decision-making.

Step 4: Fund Tech, Workflow Redesign, and Change Management Together

Do not fund software in isolation; allocate budget across workflow transformation, employee upskilling, and change management to guarantee user adoption.

Step 5: Put a Name on Every Seat in the Room

Assign explicit governance and accountability to a specific leader or project manager who owns tracking value realization and has the authority to halt underperforming projects.

Step 6: Ask for the Acceptance Rate

Define what acceptable output quality looks like in practice by establishing target acceptance rates for AI recommendations before diminishing returns set in.

Step 7: Quantify the Workforce Question

Map your workforce census into functional cohorts (such as FP&A, customer service, engineering) and align strategic headcount growth projections to the expected operational impact of AI.

Step 8: Test Every Item Against an Unasked Diligence Question

Evaluate every project against the standard private equity investment thesis, ensuring you can articulate why capital was deployed, what returns were generated, and why stalled initiatives were cut.

When This Works (and When It Doesn't)

This system fits portfolio companies entering standard annual budgeting cycles where private equity sponsors demand clear hurdle rates for capital projects. It applies directly to mid-market and enterprise platforms where multiple departments purchase point solutions on decentralized corporate cards without coordination.

It runs into friction in early-stage carve-outs or rapid buy-and-build rollups where underlying ERPs and general ledgers remain fragmented across acquired entities. If core accounting systems cannot track cost center tags cleanly, running an ROIC calculation on individual workflow tools creates administrative drag without providing clear data.

Why It Matters

Sponsors are moving past the initial hype cycle where boards approved exploratory tech budgets without defined return hurdles. When exit multiples contract, value creation plans require measurable operational improvements rather than pilot programs. Tying AI licenses directly to workforce census models and ROIC metrics signals that sponsors now view artificial intelligence through the same cost-takeout lens applied to plant consolidations or procurement renegotiations.