Human-AI collaboration starts long before your first AI agent goes live

Enterprise AI has matured faster than most organizations’ ability to change how work gets done. AI initiatives now succeed or fail less because of model performance than because organizations fail to redesign the work surrounding the technology. 

Companies are deploying agents across their workforces or planning to do so in the near future. Yet recent research from TeamViewer shows that even as people become more comfortable using AI, 61% still prefer that it take no independent action. 

Software is becoming an active participant in how work gets done. Organizations that manage this shift well can improve performance while allowing employees to focus their judgement, creativity, and expertise on the work where they create the most value. 

Poorly designed AI rollouts create both inefficiency and distrust. Organizations that redesign work around collaboration between people and AI will be better positioned to realize the value of agentic AI. That process starts by involving employees at every stage — from identifying the strongest use cases to remapping workflows that incorporate digital colleagues. 

Circles, squares and the new shape of work 

At Google Shopping, I frequently designed and implemented workflows that defined how people and technology came together to achieve a shared outcome. At the time, those workflows were built entirely around human roles. When mapping them, I represented each person involved in the process with a circle. 

Agentic AI forced me to redraw the diagram. 

Today’s workflows have two participants: people, represented by, for example, circles, and AI agents, represented by squares. The goal isn’t to replace circles with squares. It’s deciding where each creates the most value. In a well-designed workflow, squares might retrieve information and complete routine actions. Circles provide supervision, resolve escalations, and make decisions that require context, accountability, or empathy. 

Together, they form an operating model for work shared by people and AI. 

Consider an HR support process. A square might answer common employee questions and retrieve policy information before escalating a complex request to a circle. The agent provides speed and scale, while the person contributes the judgment, context, and accountability needed to resolve the issue. 

Design AI with employees, not for them 

For employees to support this operating model, they must understand why the organization is adopting it and have a role in shaping the transition. Leaders should draw on employees’ expertise to determine where agentic AI fits, then introduce the technology in a way that gives the people affected genuine ownership. 

1. Sketch the new operating model 

Before introducing agentic AI into a workflow, leaders need a working theory for how it could improve a business outcome. 

Start with the desired outcome. Then determine which responsibilities are appropriate for agents and where human judgment, approval, or oversight will remain necessary. 

Treat this model as a starting point, not a finished decision. Employees should have a voice in how AI changes their roles, but asking them to respond to a thoughtful proposal is more productive than asking them to design the future from a blank slate. 

2. Involve subject matter experts early 

Employees should not encounter agentic AI as a top-down initiative imposed by the C-suite. 

Leaders can avoid that perception by bringing the right subject matter experts into the planning process soon after creating the initial workflow sketch. Team leaders and frontline employees understand how the work actually gets done, where friction exists, and which tasks are appropriate for AI support. 

When their expertise shapes the operating model, employees are more likely to support the change and help colleagues adopt it. 

3. Give employees ownership of the rollout 

Leaders must clearly explain why the organization is exploring agentic AI. That explanation should connect the technology to a concrete business need and show how their work may change. 

Organizations can also give employees a visible role in the rollout. Hitachi, for example, invited employees to help name its AI coworker. The company introduced Skye, an internal HR companion, at a launch event featuring prizes and fanfare. 

By involving employees, Hitachi made the rollout something they could participate in rather than something happening to them. The approach helped position the agent as a shared addition to the organization. 

The lesson is not that every AI launch needs a naming contest. It is that employees should be able to see where their input influenced the rollout and how the technology will affect their day-to-day work. 

Redesign the work before you automate it 

AI won’t fix a poorly designed organization. Automating a flawed workflow will only reproduce its problems faster. 

Over the next few years, businesses will increasingly compete on how effectively people and agents work together. Those that succeed will give employees a meaningful voice in the transition.  

They’ll preserve the work that requires human judgment and build operating models in which people and agents contribute to shared outcomes. The advantage will go to those organizations that redesign the work before they automate it.