What McKinsey's State of AI Report Really Means for Mid-Market Operators

Your company is using AI. So is everyone else's. That part is solved.

The unsolved part, according to the data, is the revenue impact. McKinsey's 2024-2025 State of AI report found that adoption is now near-universal across enterprises and mid-market players alike, but value capture is stuck at the back of the room. A significant majority of companies report using AI in their business, yet struggle to convert that into measurable EBIT or strategic advantage. The problem isn't the AI. It's that you're measuring the wrong things, governing the wrong way, and redesigning the wrong workflow.

This matters acutely for mid-market operators, where capital is tight and every 50 basis points of margin counts. If you're running a $20M to $200M business, you can't afford to be a dabbler. But you're also not a Fortune 500 with a dedicated AI center of excellence and three hundred people. So where do winners actually pull away?

The Adoption-to-Value Gap Is Not Closing Itself

McKinsey's data shows a chasm between deployment and return. Companies are running pilots, launching agents, deploying copilots inside internal tools, setting up data infrastructure. All of this is happening. None of it is reliably translating into a line item on the P&L.

The standard story is that everyone is doing AI, so the gap must be execution or maturity. The real story, confirmed by practitioners across both consulting and operations, is starker: companies are measuring progress wrong, and that blindness is the root of the value problem.

If you're measuring success by "pilots launched" or "percentage of headcount trained" or "AI tools deployed," you will hit those targets and see zero revenue gain. That disconnect is not an accident. It's because those metrics reward activity, not value creation. A mid-market company we spoke with last year launched AI-driven customer segmentation in their CRM, trained the entire go-to-market team, and called it a win. Eighteen months later, nobody was using it. The tool was sound. The segmentation engine was right. But it required sales reps to change how they qualified leads, and the organization never redesigned the workflow to make that the default path, not an optional extra step. Result: excellent pilot, zero revenue impact.

Workflow Redesign Is the Actual Constraint, Not AI Capability

Here is where McKinsey's finding lands hard for mid-market operators: real value comes from redesigning workflows, not buying tools. And that's organizational work, not technical work.

The implication is uncomfortable. It means that you can have a world-class LLM, perfect fine-tuning, clean data, and it will still produce zero net value if the humans using it haven't actually changed how they work. AI doesn't create value in a vacuum. It creates value when you've pulled levers in process, measurement, staffing, and decision rights to make the AI output the default input to a decision or action.

This is why agentic AI is the next frontier for mid-market, not because it's a hype word but because it removes human workflow as a constraint. An agent that automatically routes tickets, ranks leads, or flags anomalies doesn't require a user behavior change. It just works. But even agents require upstream decisions: What is the SLA for agent action? Who reviews? What are the rollback conditions? Those are governance and workflow questions, not AI questions.

McKinsey's report flags this as a differentiator between companies capturing real value and those stuck in pilot purgatory. Winners aren't better at picking models. They're better at redesigning the human process around the AI's output.

Governance and a Named Owner Separate Winners From Dabblers

McKinsey emphasizes governance and organizational ownership as the core lever separating companies that pull real value from those that don't. For mid-market, this is a relief and a challenge in equal measure.

It's a relief because it means you don't need to solve the unsolved problem of enterprise AI governance at the scale of a 50,000-person corporation. You need to solve it for your core workflows. That's tractable.

It's a challenge because every mid-market operator we talk to is still figuring out where AI governance lives organizationally. Is it the CTO? The COO? The CFO (increasingly common)? Should you hire a named AI lead? Most mid-market companies don't have an "AI officer," and many don't think they need one. McKinsey's data suggests otherwise.

Here's the operational decision: You need one named person or a small team (not a committee) with decision authority over what gets built, how it gets measured, who owns each workflow change, and what triggers a rollback. That person is accountable for both the technical execution and the business outcome. In large enterprises, this is a role. In mid-market, it's often part of someone's job along with five other things. And when it's split across five people and four departments, nothing gets redesigned and nothing ships.

The governance gap isn't a compliance box. It's a decision-speed box. Companies that move fast on AI have clarity on: (1) who decides if an AI initiative gets funding, (2) who owns the outcome, (3) how we measure return, (4) what happens if we're wrong, and (5) who has authority to stop it if we are. That's it. Formalize that and you're already ahead of most peers.

Your Metrics Are Almost Certainly Wrong

This is the finding with the most immediate operational teeth. McKinsey's research flagged that companies use the wrong metrics to judge AI progress, and the window to course-correct is shorter than anyone thinks.

If you're measuring by adoption rate (percentage of team trained), cost of infrastructure (capex sunk), or number of use cases launched, you're optimizing for the wrong thing. Those metrics tell you about velocity and investment, not value. They're also backward-looking and easy to game.

Mid-market operators should measure: (1) decisions accelerated or automated, (2) revenue impact per workflow redesigned, (3) cost saved or headcount freed (with specificity about what that person is now doing), and (4) customer outcome improved (if customer-facing). These are harder to track. They require you to actually know your baseline before the AI change. They force you to redesign workflows explicitly, not accidentally. That's exactly why you should use them. The companies winning on AI are the ones ruthless enough to shut down initiatives that can't articulate a line-item return in six months to a year.

Agentic AI is the Next Adoption Wave for Mid-Market

McKinsey identified agentic AI as the frontier. For mid-market, this is where you should be thinking, even if you're still ramping up your first generation of AI tools.

Agents require a new governance model. They have decision authority. They act without human intervention in the loop. That means you need to define: What can this agent do? What requires a human check? What's the audit trail? If it causes harm, how do we know? Who is liable? These aren't hypotheticals for high-budget enterprises. They're immediate for mid-market because you can't hire expensive compliance teams to sort it out. You need to think it through yourself, which forces clarity.

The upside is that agents are the fastest way to convert AI into workflow redesign at scale. They automate the human constraint. The downside is that governance failures are also more visible and costly. A busted copilot is an inconvenience. A busted agent that shipped fifty bad orders is an incident.

For mid-market operators, the message is: start thinking about agents now, but frame it not as "cool new tech" but as "how do we remove this human workflow as the bottleneck." That framing ensures you're solving for value, not novelty.

The Takeaway: Go Narrow, Then Scale

McKinsey's data points to a clear pattern: companies that win on AI do three things in sequence. First, they pick one high-impact workflow that's currently constrained by human review or decision-making. Second, they redesign that workflow explicitly to use AI as the primary input, with governance and measurement baked in from the start. Third, they hold themselves accountable to an actual revenue or efficiency metric, not an activity metric. If it hits the return target in six months, they replicate the model to the next workflow. If it doesn't, they kill it and move on.

That discipline is brutal, but it's also what separates mid-market companies that compound AI value year-over-year from those that run endless pilots and call it progress. Your AI journey isn't about building the most sophisticated models or training every employee. It's about redesigning work so that AI is the default, governance is clear, and every dollar of infrastructure spend maps to a decision or revenue outcome you can trace.

We've written before about how to measure AI ROI as a mid-market operator, and how to navigate the path from strategy to production. McKinsey's report is a validation that the companies asking those questions first are the ones pulling real value. If you haven't yet, that's your next step.