AI Catalyst Partner Matthew Simons shares his insightts and practical steps into how mid-market businesses can prepare for the effective adoption of AI agents.
McKinsey CEO Bob Sternfels revealed in January 2026 that the consulting firm now employs 25,000 AI agents alongside 40,000 human staff. Speaking at the Consumer Electronics Show and on Harvard Business Review's IdeaCast, he stated the firm expects to reach parity, equal numbers of humans and agents, by year end.
This is not a glimpse of a distant future. It is a statement about the present, from one of the world's most sophisticated professional services firms.
For mid-market businesses, this raises an immediate question: if McKinsey is restructuring its workforce around AI agents, what does that mean for your organisation?
The answer is not that you need to rush to deploy thousands of agents. It is that you need to start thinking about agents differently and preparing your organisation for a shift that is coming whether you are ready or not.
Terminology is confusing. Chatbots, copilots, assistants, agents: the labels overlap and marketing often obscures more than it clarifies.
The key distinction is autonomy. A chatbot responds to questions. A copilot assists with specific tasks when prompted. An agent can be given an objective, break it down into steps, make decisions about how to proceed, use tools to accomplish sub-tasks, and work toward the goal with limited human intervention.
At McKinsey, agents are handling entire job functions: creating slide decks, compiling research, synthesising information from multiple sources. The firm reports saving 1.5 million hours in 2025 through AI automation of tasks that junior consultants previously performed. Their agents have generated 2.5 million charts in the past six months alone.
This is the shift from using AI tools to managing AI workers. It requires a fundamentally different mindset.
When you deploy a traditional software tool, you think about configuration, training, and adoption. When you deploy an AI agent, you need to think about delegation, oversight, and accountability.
This is closer to hiring a new team member than implementing a new system. Consider a concrete example: you want to deploy an agent to handle initial customer enquiries and route them appropriately.
With a traditional chatbot, you script responses and define decision trees. With an agent, you define the objective (qualify enquiries and route to the right person), the constraints (what it can and cannot promise, when it must escalate), and the resources it can access (CRM data, product information, team calendars). The agent then decides how to achieve the objective within those boundaries.
The questions become managerial rather than technical. What authority does the agent have? How will you know if it is performing well? What happens when it makes a mistake? Who reviews its work?
For leadership teams, this is a significant shift. The questions are not primarily technical. They require input from operations, HR, compliance, and the business functions where agents will operate. This is why new roles are beginning to emerge.
Just as the growth of websites created demand for UX designers and the mobile era spawned app developers, the agent economy is generating new specialisms.
Agent Operations Managers oversee the day-to-day performance of AI agents. They monitor reliability, manage exceptions, and ensure agents are meeting their objectives. This is a blend of process management and technical oversight.
Agent Experience Specialists design how agents interact with humans. When should an agent escalate to a person? How should it communicate uncertainty? How do you build trust between human workers and their AI counterparts? These are UX questions applied to a new context.
For mid-market businesses, these roles may not require dedicated hires initially. But someone needs to own these responsibilities. The skills involved—a combination of process thinking, technical literacy, and human factors understanding—are not typically found in traditional IT or operations roles.
Not every task is suitable for an AI agent. The best candidates share certain characteristics:
A practical starting point is to audit your operations for tasks that meet these criteria. Where are your teams spending time on work that is high-volume but relatively predictable? Where would errors have limited consequences while you are learning?
Equally important is assessing your organisational readiness. Do your leadership team understand what agents are and how they differ from traditional automation? Is there clarity about who would own agent deployment and management? Do you have the governance foundations to manage the risks that come with increased autonomy?
There is a tendency to assume that larger organisations will lead in AI adoption. In many ways, the opposite is true.
Mid-market businesses can move faster. You have fewer layers of approval, less legacy infrastructure, and more direct lines between strategy and execution. When you identify an opportunity, you can act on it without navigating corporate bureaucracy.
The MIT research on AI deployment found that large enterprises take an average of nine months to scale AI pilots, compared to just 90 days for mid-market firms. That speed advantage compounds over time—early adopters build capabilities and institutional knowledge that later entrants must work harder to develop.
You also have the ability to be more deliberate. Enterprises often deploy agents at scale before fully understanding the implications. Mid-market businesses can pilot an agent in one function—customer service, for example—learn from the experience, refine the approach, and then expand. That controlled learning is harder to achieve in large organisations with multiple stakeholders and competing priorities.
The key is to start thinking about this now. Not necessarily deploying agents tomorrow, but building the understanding, the governance, and the skills that will allow you to move confidently when the time is right.
The transition to an agent-enabled workforce will not happen overnight. But for mid-market businesses, waiting until agents are unavoidable means ceding the advantage to competitors who prepared earlier.
A practical starting point involves four steps:
The organisations that will lead in the agent economy are those that treat this as a people and process challenge, not just a technology decision. That perspective is where mid-market businesses can genuinely differentiate.
If you are thinking about where AI agents might fit in your business, or want to assess your organisation's readiness, our Solutions Accelerator is designed to help you move from concept to implementation with confidence.
Sources: