Three months after the portfolio review where AI adoption became a fund priority, an operating partner checks in with the CEO of a $70 million professional services company. The CEO confirms that yes, the team has been looking at some tools, a few pilots are underway, and they are planning to revisit in Q4. The operating partner has heard a version of this conversation at every other portfolio company in the fund. Q4 is ten months away.
Why the AI bottleneck sits inside portfolio companies, not above them
The operating partner’s challenge here is not CEO resistance, and it is not a shortage of available AI tools. Both of those diagnoses lead to the same wrong response: more vendor presentations, more readiness workshops, another Q4 deadline that slips to Q1. The actual problem is structural, and it sits inside the portfolio company rather than above or below it.
Most mid-market service companies at $40M–$150M revenue have some version of a technical function — a CTO or a Head of IT whose mandate is infrastructure: stable systems, vendor management, and whatever the business has specifically asked to be evaluated and built. Their expertise is technical. Their frame of reference is systems stability, integration complexity, and build-versus-buy decisions on technical grounds. When an AI mandate arrives, they do what they are equipped to do — evaluate AI products against technical criteria and shortlist vendors whose demos address the most visible pain.
What no one at this scale has — almost without exception — is a Chief Digital Officer. Not a CTO, but a CDO: the person whose job is to look at the business’s margin structure, identify which operational workflows are consuming disproportionate cost, and translate that into a sequenced, outcome-anchored technology roadmap. The CTO asks what a tool does. The CDO asks which workflows are worth targeting first, in what order, measured against what specific P&L outcome. Those are different questions that require different expertise, and in the mid-market, almost no one is asking the second set.
This is the CDO gap — and it is the structural reason AI mandates from PE funds stall at the portfolio company level. The operating partner provides strategic direction. The portfolio company CEO is willing to act. The technical function evaluates options. Nobody performs the translation step between business priorities and technology decisions, so the mandate never converts into a business case that can be built, deployed, and measured against a verifiable outcome.
Why AI mandates produce expensive experiments rather than EBITDA
Ask an operating partner who has pushed more than one AI mandate through a portfolio, and the story is roughly the same. The fund mandate arrives. The portfolio company CEO acknowledges it and delegates to the technical team. The technical team evaluates AI vendors whose demos impressed someone in a presentation. Pilots start. After six months, the pilots have produced a proof of concept or a workflow that functions for a specific team but hasn’t been extended more broadly. Someone records a learning; the budget line is noted as “AI exploration,” and EBITDA registers no change from the exercise.
The reason this happens is not that the pilots were poorly run technically — most of them run fine on their own terms. The pilots were never anchored to a specific operational cost before they started. Nobody sat with the operations function, identified which process was consuming $800K per year in analyst headcount, expressed that cost in terms a CFO could verify, then asked which AI tool could address it, in what deployment sequence, with what measurable output at what point. That sequence of questions is the CDO’s job. Without someone performing it, the default is to pilot interesting technology and see what sticks.
The hold period makes this costly in a way that individual pilots don’t reveal. A typical PE hold of five years has a finite value creation window. If AI pilots consume the first eighteen months producing experiments with no EBITDA anchor, a meaningful portion of that window has passed. At a $70 million revenue portfolio company, twelve months of unfocused AI exploration might cost $200K–$400K in direct vendor and internal time — modest in isolation, but compounded across eight or ten portfolio companies, and measured against the EBITDA improvement that didn’t happen, the aggregate cost is difficult to ignore when someone runs the numbers at the fund level.
In Digital Forms, we call the discipline that breaks this pattern the 0.5 FTE gatekeeper rule: no initiative gets funded unless it can demonstrate savings of at least half a full-time employee, expressed in a specific workflow, before vendor selection begins. Every AI initiative should be expressible in FTE impact before a pound of budget moves. If the portfolio company cannot articulate, in specific terms, how many full-time employees’ worth of effort the initiative will eliminate or redeploy — and in which workflow, on what timeline — the initiative is not ready. This standard applies before vendor selection, not after deployment. Applied consistently, it kills most unfocused pilots before they consume budget and produces a shortlist of AI use cases that are directly addressable to the cost structure.
What the CDO gap looks like inside a portfolio company
The gap is invisible because it doesn’t manifest as an obvious vacancy. Nobody has a job title that says “CDO.” The operating partner doesn’t receive a report saying “strategic translation layer absent.” What they see instead is a softer pattern: AI investments that have gone live without moving operational KPIs; quarterly reviews where the CEO can describe which tools the company is using but not what those tools have changed about the cost structure.
A useful diagnostic: ask the portfolio company CEO to explain, in FTE savings and margin terms, what the last three technology decisions have delivered. If they can answer specifically, someone at that company is performing a CDO function. If they redirect to feature descriptions — “we implemented a document AI for classifying incoming emails,” “we’ve got a chatbot handling tier-one support queries” — the CDO gap is open, and the AI mandate will produce more of the same experimental pattern until something changes about who is asking the questions.
This matters particularly for AI because AI tools carry high feature-visibility and low outcome-clarity. It is straightforward to demonstrate that a language model can summarise contracts. It is considerably harder to demonstrate that the summarisation saves three analyst-hours per case and therefore £400K per year in a claims function handling 15,000 cases annually. The second demonstration requires a business lens applied before the technology decision, not after it — and the absence of that lens is precisely what makes the CDO gap consequential at this stage of AI adoption.
The CDO gap also compounds with what becomes, at portfolio-company scale, a Human API problem: staff manually bridging between the six, eight, sometimes ten disconnected software systems sitting in the company’s stack, none of them integrated with each other. When an AI mandate arrives, these businesses often discover that the workflows most amenable to automation are the same ones most dependent on staff performing the bridging function informally. An AI tool deployed on top of a workflow that nobody has fully mapped — because the mapping has always been done by whoever happens to know the exceptions — tends to automate the visible layer and leave the human bridging intact underneath. The failure is diagnostic — the workflow was never properly mapped before the tool arrived. Without someone whose mandate is operational design first and technology second, AI sits on top of the existing complexity rather than reducing it.
The non-technical CEO guide to technology decisions addresses one dimension of this from inside the portfolio company: the experience of technology arriving in technical language when the CEO’s frame of reference is P&L. The operating partner’s challenge runs a layer above — ensuring that portfolio companies develop the capacity to make that translation consistently, rather than continuing to delegate technology decisions to people who don’t have a business-outcome mandate.
Why filling the gap with a permanent hire doesn’t work at this scale
The direct response — hire a CDO — runs into two problems that make it impractical for most portfolio companies at this revenue level. The first is cost. A CDO capable of performing the function properly typically costs £250,000 to £400,000 per year in base plus benefits, a meaningful EBITDA impact before they have produced a single outcome. The second is timing: recruiting a CDO takes around six months in a competitive market, and the ramp period before the hire has enough operational context to make good decisions adds several more months on top. In a five-year hold, that is a significant slice of the value creation window consumed before any substantive work begins.
There is also a structural fit problem with full-time CDO hires at mid-market scale. The role requires intensive bandwidth during the diagnosis and initial deployment phases — call it the first eighteen months — and then a substantially lighter governance function once the roadmap is established and the first automations are running. A company at $70 million revenue pays a senior executive salary year-round for a job that is front-loaded and intermittently demanding thereafter. Most companies at this revenue level find the economics unattractive once the initial deployment cycle is complete, and the CDO role either expands into adjacent functions to justify its cost or becomes politically awkward as the pace of new initiative work slows.
Operating partners who have navigated this describe a different model: deploying an external CDO function on a fractional or project basis, with the engagement structured around specific EBITDA outcomes rather than an advisory retainer. The external partner enters, performs the diagnostic work to map operational costs and identify AI-addressable workflows, produces a roadmap expressed entirely in P&L terms, then runs the first deployment cycle with outcome accountability built into the arrangement. The portfolio company gets the CDO function for the period it needs it most, without the permanent overhead. The operating partner gets an EBITDA-anchored AI roadmap rather than a report on adoption rates. And the digital transformation strategy conversation — which in most portfolio companies happens at too high a level of abstraction — gets grounded in the specific operational cost structure of the business rather than a benchmark about what percentage of the industry is “using AI.”
What to do in the next 90 days
For operating partners who recognise this pattern, the most useful next step starts with a diagnostic conversation with the portfolio company CEO — not the CTO — focused entirely on operational cost. Where are the workflows that consume the most manual FTE? What is the loaded annual cost of each? Which of them has a consistent, documented process that could be a candidate for AI-assisted automation? These questions don’t require technical knowledge to answer; they require P&L fluency and access to the operations function. If the CEO cannot answer them, or immediately routes the conversation to the technical team, that routing behaviour signals the CDO gap directly — the business-outcome framing for technology decisions doesn’t exist at the leadership level, and any AI mandate will land in the same place it has been landing.
From there, the question is whether to address the gap through an external CDO model or build toward an internal hire, with the decision anchored in cost structure and timeline rather than preference. What the decision should not be is another AI vendor evaluation cycle launched before the business problem has been defined in financial terms — because the selection criteria will default to features, the pilots will default to proofs of concept, and the Q4 check-in conversation will default to the same shape as the last one.
The Quick Wins framework that Digital Forms deploys in the first six weeks is structured around the same discipline: the first deliverable needs to be a measurable operational saving — something the CFO can verify in the same quarter — rather than a strategy deck or a proof of concept. At the portfolio company level, the difference between an AI mandate that reaches Q4 unchanged and one that delivers EBITDA within the hold period often comes down to whether the first ninety days produced a verifiable financial outcome or an expanded set of possibilities.
The tools exist. Portfolio company leadership is generally willing to act. Without the CDO function in place, the AI mandate arrives at Q4 as a list of pilots rather than a P&L entry. What stalls it, in nearly every case, is the absence of someone whose job is to ask the business question before the technology question — and to hold the outcome accountable to a margin line rather than a feature demonstration. The operating partners who have broken through this describe the same turning point: not when the AI worked, but when the portfolio company CEO could explain, in margin terms, what it had actually changed.
