You speak and a voice on the other end interviews you. You’re asked a question, you’re given room to back off and double down, and the questions keep fitting into what you just said.
You may never notice that you are talking to an agent. Behind her, a second agent is reading against a rubric, figuring out what to look for next.
Improvement president Michael McCullough and director of consulting Michael Ell build such agents for a living. They told a room of technology executives in the CIOCAN Peer Forum last month that these agents are moving from handling single tasks to running all end-to-end workflows in a business, with people watching instead of driving.
The ability is further than you might think. What it has missed is everything about it, the governance, the instruments and the question of who is responsible for the work since an agent is the one who does it.
From performing tasks to mastering workflows
McCullough and Ell classify agents into three types. Capture IQ, the voice interviewer, falls into the category of what they call value streaming agents.
There are chatbots like Copilot and Gemini that are good for personal productivity. Then there are horizontal agents, the same tool that serves the entire company, as one that allows anyone to pull from a common knowledge base of the company. A value stream agent is more specific, targeted at a line of work, such as agents who write software or handle the work of a legal or finance team.
Improvement runs one of these on its own staff. When a project is finished, Capture IQ interviews the team and turns the conversation into a sales case study, from interview to draft to approval to release. Work that used to take a person walking through a script now goes through the agent.
“Essentially, instead of using a keyboard as input, we created agents that you would actually talk to,” Ell said.
All three types describe what an agent does. A separate question is how much autonomy there is, and McCullough has determined that as well.
After all, companies are just experimenting with AI. Agents then tackle a task under close supervision. They then move through departments, with people monitoring rather than approving each step.
At the bottom is full autonomy, where, as McCullough said, agents are “initiating and executing them end-to-end” within the security fence.
Most companies, he said, are nowhere near that conclusion.
The bridge where nobody stands
Technology is the easy part, but McCullough warns that everything around it is where companies get stuck.
When an agent takes over a piece of work, it becomes two new jobs instead of replacing one.
Someone has to tell the agent exactly what to produce, and someone has to check what it gets back. McCullough said that’s where things get stuck right now, and a lot of companies haven’t figured out how to do it well.
The people most determined to define that work, the ones who plan projects and write the business case for them, don’t have good software for it like developers do. Some companies give them an integrated development environment (IDE), the workspace developers write code in and hope it works, which McCullough said doesn’t happen.
His staff has also retired.
When an agent does the job that someone used to do, it raises the uncomfortable question of what that person is for now.
The real obstacle lies beneath all that. Agents need clear guidance from people who understand the business and technical guidance from people who understand how agents work. Those people are often not on the same team.
One has to know the business well enough to say what a good result looks like. The other is knowing how to find an agent to produce it. McCullough said each side tends to wait for the other to go the distance, and while they wait, nothing moves.
Who is responsible for the result?
Many companies already know they need their data in order before any of this can work, a point worth discussing and I have written for before.
The agents raise another question that has not been resolved. When the agent does the work, who is responsible for it?
When one person leads a cross-departmental process, you know who to ask when things go wrong. Run the same process through an agent and that clarity becomes murky. The liability is still there, but the owner is harder to identify.
In practice, it probably falls to the CIO as the technology leader, who takes responsibility without the authority that should come with it. The job involves departments that the CIO doesn’t lead, so being accountable for the bottom line means being accountable for things they can’t directly change.
Losing control of something you’re still responsible for is a problem CIOs already know in a different form.
Raju Vegesna, chief evangelist at Zoho, who warns that running your business through a platform you don’t own means slowly losing ownership of the value it produces. He calls it the difference between a customer and a pawn.
Agents extend this concern from data ownership to workflow ownership, raising the question of who controls the work once the software performs the process.
Working backwards from the customer
This makes ownership more than a matter of governance. If an agent is performing work across departments, someone must decide what the work is ultimately intended to change.
Dave Aeri, CIO and CTO of Shoptravel Group Inc. and a national board member at the CIO Association of Canada, adapted that test around the customer.
For him, the measure of technology’s work is about whether it changes the way business works and reaches the customer.
“If that result impacts horizontally and vertically across the organization, plus it impacts the customer, you’ve done something bigger than just your role,” he said.
This is a useful way to think about a job that now involves responding to agents. A CIO who measures success by organizational impact is already in the habit that the age of agents requires, because the work of an agent appears as an impact on the entire business, not as a regular handover to an owner.
Aeri pairs it with a caution for speed.
“Even if you define a use case today, it can change radically within 90 days,” he said.
An agent can get better and better at getting the job done from one end to the other, and the question of who owns that job still has to be answered by the people around it. McCullough and Ell build these agents and work alongside the companies that deploy them, so the gap they show is one they’ve seen first hand.
That opening voice was convincing because you couldn’t tell who or what you were dealing with. The same goes for a workflow as an agent runs it end-to-end. Ownership is not clear and will not be self-assigned.
A CIO who doesn’t decide in advance who is responsible for the agent’s work is the one who ends up being responsible for it by default.
The last shots
- The hard part of agent AI is connecting the people who know the business with the people who know the agents. It’s an organizational problem, and no vendor sells a solution to it.
- A CIO inherits responsibility for the work of an agent without the authority to change how it operates in departments they do not control. The organizational chart has not caught up with the way work now moves.
- Technology resets on a roughly 90-day cycle, which makes every decision on how far to let agents drive a decision that a CIO will constantly revisit. The question of ownership is the one worth solving first.
