Don’t let AI agents take off in all directions – here’s how to build an AI control tower
AI agents should be managed with the same clear principles as people. That’s why it’s wise to build an AI control tower before agents evolve from subordinates and coworkers into leaders themselves, Juha Kujala and Lassi Kurkijärvi remind us.
The rapid growth phase of a startup is often marked by a frantic rush to recruit new talent. In a scaling company, there’s no shortage of work – and if funding is in place, new people are often hired so enthusiastically that it feels like they’re coming in from all directions.
At some point, this hiring high almost inevitably leads to trouble. So many new people have joined that no one’s quite sure what everyone is doing. Roles, responsibilities, and positions in the organisation can be unclear or overlapping. There’s plenty of energy and buzz, but no shared direction or goals. The company’s staff ends up like a flock of birds with no one leading the way.
Often, this growth high is followed by a hangover – and a course correction where roles and responsibilities are clarified. Some of the newly hired employees may end up leaving the startup.
A similar scenario may soon play out in companies currently captivated by the potential of AI agents. These agents can be deployed quickly and easily to handle various knowledge work tasks – and in the excitement, many rush in without a clear strategy or any real thought about how the agents should be managed. That AI-induced high can easily turn into a hangover.
Build a control tower for the flock before it scatters
Fortunately, the AI hangover can be avoided. The solution is not to abandon AI agents or their deployment, but to ensure that adoption is guided by the company’s strategic goals and done in a structured way. When that happens, AI agents won’t scatter like a flock of birds — they’ll move purposefully toward a shared objective, under clear leadership.
Build your AI control tower by defining these six essentials for your organisation:
1. AI strategy
Define what business objectives the deployment of AI agents is meant to support. Create a roadmap for implementing AI agents across the organisation.
2. AI lifecycle management
AI technology evolves quickly, so the lifecycle of any single agent may be short. However, their deployment, updates, and decommissioning must follow clear, consistent guidelines.
3. Measuring AI value
An agent that’s poorly trained or assigned the wrong task may add little value – or even cause harm. That’s why companies need clear, regularly updated metrics for measuring the value agents generate.
4. AI risk and compliance
The risks associated with AI agents must be assessed regularly, and their management must follow consistent rules. Governance should also cover compliance issues – particularly with regard to the EU AI Act and other relevant regulations.
5. AI inventory
Organisations must have a clear view of what AI agents are in use at any given time. This includes systems that rely on collaboration between agents, as well as the prompts used to train and guide them.
6. AI monitoring and oversight
Beyond knowing how many agents are in use, it’s essential to track their usage levels and the quality of their outputs. Monitoring also includes ensuring data security and overseeing the use of confidential or sensitive information.
Subordinate, colleague, and soon your boss
As AI agents become part of organisations, they need to be managed just like human employees. Increasingly, these agents are becoming team members and coworkers – and soon, they may also take on leadership roles.
We partnered with Haaga-Helia University of Applied Sciences on a nationwide survey conducted in May–June titled “Machine as Manager”. The study explored the kinds of leadership tasks that can be delegated to AI, the benefits and risks involved, and how organisations should prepare for AI in leadership roles.
Most respondents still prefer working under human leadership. However, that may soon change: 49% reported already having encountered a machine acting as their manager. Younger employees are also notably more open to AI leadership than their older counterparts. Many respondents believe that AI could even prove to be a more equal and impartial leader than a human.
Of course, the arrival of AI managers could also create chaos reminiscent of a fast-growing startup in disarray. That’s why now is the time to build an AI control tower.
About the authors
Lassi Kurkijärvi has been building digital business for three decades. Today, he helps Nordic companies navigate digital transformation in his role as Executive CTO at Sofigate. His passion is humanity in all its forms: change is ignited or extinguished by whether people are inspired and engaged by it. Ultimately, everything new should serve a better and more sustainable life.
Juha Kujala is Executive CTO at Sofigate Platforms. He’s passionate about helping organisations use emerging technologies to deliver better services. Juha has extensive expertise in the ServiceNow platform and service development leadership.
Good preparation pays off in acquisitions – 5 ways to succeed in integration
The new technical solution is not all you need – ignore the human aspect and you will lose the benefits promised
