45% of mid-market companies have no single owner for how people and AI work together. That’s the lead finding in The Co-Intelligent Workforce Report: 2026 Baseline, which we released today, and it’s the number that made us want to do this research in the first place.
Adopting AI has never been easier. Getting true value from it remains elusive.
In our work with mid-market leaders, I kept watching the same scenario play out. A company starts using AI with real conviction, people experiment, a few workflows get faster, and drafting and analysis suddenly take a fraction of the time. Then the momentum stalls, the executive team can’t say what the AI spend is returning, and a promising effort shrinks into a tool that writes emails. The technology did what it was supposed to do. The company never decided what AI would be used for, who would own it, or how work should be divided among a workforce of people and AI.
So we asked. Working with Port Research on research strategy and Researchscape on survey design and fielding, we surveyed 484 Director-and-above leaders at US and Canadian companies with 100 to 999 employees between August 17 and September 14. The title borrows from Wharton professor Ethan Mollick, whose idea of co-intelligence treats AI as a collaborative resource rather than a tool, because that describes what the companies getting real value have built. The results come down to five findings, and read together they point to one pattern: the companies getting value from AI decided who owns how people and AI work together, and they designed that relationship on purpose.
45% of mid-market companies have no single owner for how people and AI work together, and only 21% have someone whose main job it is. Another 34% have handed it to someone as one responsibility among many, and 12% say nobody owns it at all.
When it’s everyone’s job, it’s no one’s job. Decisions about which tools to buy, which workflows to automate, and who checks the AI’s work still get made at those companies. They're just made locally and inconsistently, with no one accountable for the whole. Where no one owns the mix, 53% of leaders admit they are hardly or not at all confident in how work should be divided between people and AI. Everywhere else, that number is 7%.
The mid-market didn’t miss the AI wave. 97% of the companies we surveyed have adopted AI, and 92% are still raising their spend. Only 28% say AI is extremely valuable to their business today. By adoption math, the transformation is done. By value math, it has barely started.
Here’s my read on the economics. AI has reached commodity status. Every one of your competitors can buy the same models you can, at the same price, today, and nothing everyone has can be a lasting advantage on its own. Large enterprises are seeing the same trend: McKinsey’s 2026 State of AI survey found that only 37% of organizations can attribute any earnings impact to AI, unchanged from the year before. Most companies have treated buying AI as if it were the same thing as becoming a business that runs well with AI.
Companies with a primary owner of the human-AI mix are 2.5x as likely as everyone else to call AI extremely valuable (55% vs. 22%), and 5.5x as likely as companies where no one owns it (10%). We call that return the Orchestrator’s Dividend.
Ownership shows up in unglamorous habits. Companies with a dedicated owner keep a current inventory of the AI they run (85% vs. 63%) and review AI’s effect on the workforce at the executive level every quarter (76% vs. 48%). That inventory matters more than it sounds. Companies with a fully up-to-date record of their AI tools report extreme value 5.9x as often, and most mid-market companies couldn’t produce that list today, because AI features keep showing up inside platforms they already pay for.
37% of mid-market companies are investing in AI technology alone, and 6% of them report extreme value. Companies that also invest in programs, like training and a written AI policy, and in people practices, like skills assessment and role redesign, reach 50%. That’s an 8.3x difference, and it’s the quantitative version of the commodity point above: the technology returns almost nothing at the top of the value scale until it’s paired with the work only the company itself can do.
The people practices show it most sharply. Companies that have fully assessed which skills their people will need as AI takes on more of the work are more than 6x as likely to report extreme value, yet 45% of companies haven’t done it.
We asked two questions: is the division of labor between people and AI clear, and is it clear who catches and fixes the AI’s mistakes? Where both answers are yes, 51% of companies report extreme value. Where they aren’t, 9%. That 5.7x difference comes from work that costs little more than leadership attention, and half of mid-market companies haven’t done it.
The respondents getting it right described the same pattern again and again, unprompted. AI drafts the customer support reply, and an agent verifies and personalizes it before it goes out. AI generates the code and tests, and a developer owns the requirements and the final approval. Somebody designed those handoffs. Meanwhile, among companies already running agentic AI that completes multi-step work on its own, 61% can’t say that a person reviews every one of those workflows before the work moves on.
When companies do name an owner, the seat goes to IT 48% of the time, and HR holds it fewer than one time in ten. HR handles employee concerns about AI at 44% of companies and runs the training at 41%, but it helps decide how roles get rebuilt at only 29%. That’s an expensive choice. Companies that bring HR in before AI tools are selected report extreme value 58% of the time, compared with 14% when HR shows up after the tools are in use, or not at all. Only 18% of the mid-market brings HR in early, and if you lead HR, that 4x is your business case for a seat at the next tool decision.
Closing this gap won’t come from handing your technologists more accountability. It comes from what I call Workforce Orchestration™. That starts with understanding the capabilities you need across your technology, your programs, and your people. It means moving from everyone experimenting alone to a coordinated effort with an accountable owner, and it means keeping an account of the human and AI resources you’ve deployed and what each one is returning. The person accountable for that work is the Workforce Orchestrator™.
Companies that use AI to cut costs and speed up tasks will get a real but short-lived edge. The lasting advantage goes to the companies that redesign the business around what AI frees their people to do. We hire technology to take on the transactional and the objective work, the drafting, sorting, routing, and reconciling our respondents described handing over with such consistency. We hire people for the relational and the subjective work, and the return on that second half compounds.
This is our first annual baseline. We’ll run the survey again next year and measure who closed the gap. Until then, the question that best predicts which side of the data your company lands on fits in one sentence: who, by name, owns how our people and our AI work together?
Download the full report. It includes contributions from 16 influential leaders and practitioners across economics, academia, and HR, as well as their reactions to the findings. It also closes with five moves for the CEO and six for the head of HR. Beyond the baseline report, if you want to see where your own company stands, take our free Workforce Leverage Diagnostic.
Let’s Orchestrate!