One line in McKinsey's 2026 State of AI survey deserves more attention than it has had. Among large organisations the share scaling AI agents in one or more functions rose from 27 to 40 percent over the past year. Among smaller organisations it stayed essentially flat, at 22 percent. The overall adoption figure of 88 percent hides a market splitting in two, and most professional-services firms are on the side that has stalled.
1. Why mid-sized firms stall
The reasons are not about ambition. Large organisations have integration teams; mid-sized firms have a person who knows where the exports live. Large organisations have a risk function that can write an agent policy; mid-sized firms have a managing partner who is also the risk function. And large organisations can run a twelve-month pilot as a cost of learning; a forty-partner firm needs the pilot to pay for itself. Each of these turns "should we use agents?" into "who has time to make agents work?", and the honest answer is nobody.
2. The three obstacles, named
- Reachability. The agent can read documents but not act in the practice-management system, the ledger or the mailbox, because nobody has connected them. Without that, it is a better search box.
- Ownership. There is no role whose job is how work moves through the firm. IT owns the tools, partners own the clients, nobody owns the workflow.
- Proof. Nobody measured the baseline, so nobody can show the pilot worked, so the second workflow never gets approved.
Mid-sized firms did not decide against agents. They ran out of people to make the decision stick.
3. What the mid-sized firms that moved did
- Bought reachability rather than building it. Platforms that connect to the systems the firm already runs, through standard connectors, so the integration work is configuration rather than a development project.
- Created the operations role, often part time, often an ambitious senior associate or practice manager, with authority over how three or four workflows run.
- Measured before they started. Hours per close, days per diligence pass, write-offs per fixed fee. A baseline makes the second approval easy.
- Started where the data was clean and the work repetitive: the month-end for one client type, first-pass review of one document class, the intake of one matter type.
4. Why this is an opening
A stalled market is an advantage for whoever does not stall. The professional firms that scale agents in the next twelve months will be competing for talent, clients and pricing against peers that have not, and against large firms whose scale comes with slower decisions. The window is wider for mid-sized firms than the survey suggests, precisely because so few of them have gone through it.
Hive Newsroom follows what is changing in AI and professional services. Sources are linked in the text; figures are as published at the time of writing. Comments and corrections: press@get-hive.ai. Back to the .