Secret Agents

AI is the fastest adopted technology in history, in both business and the wider world. Within businesses the adoption rate is measured at 88% from McKinsey’s 2025 State of AI report, but for most this is: individual use, experimentation or pilot phases rather than scaled solutions. So, we have unprecedented speed of adoption by users, a lower bar to entry, and a technology that continues to grow – What’s missing to turn use into wider solutions?

What is Agentic AI?

Agentic AI is one of Microsoft’s favourite topics, meaning tools that do tasks without explicitly being asked, capability to assess, reason and execute. An example for tegossuite: A Customer requests a quote for providing a service, it understands the context of the service, and the customer. Is there a contract, pricing, location, caveats, any issues in the past – data available across different systems and users. Then it validates this information and prepares a quote to review for the user, with flagged areas.

This is the best type of agent, a “secret agent”, it does not have to be visible or labelled as agentic. It does work that adds value in the background, understanding context across different areas and saving user’s valuable time.

Platforms are the enablers for this level of agent, making the tech available to all. Microsoft recently released Business Central (BC) Agent Playground, where users can build and deploy custom agents to perform repeatable tasks within BC itself, an opportunity to apply tech directly against system gaps. Workflow platforms like Power Automate and n8n enable simpler connectivity across company tech stacks, letting systems, agents and data talk.

AI in Customer Service

The tech is ready, the challenge is specific application where it’s needed, not following the hype.

Research points to predictions that customer service will be the first area to see significant change, to the point of replacement by AI. In this current moment this is not the right framing, if replaced by visible agents (we see this already with chatbots) it leads to more frustration and worse feedback, so internally there might be time saving, but is it worth losing customers?

Agentic opportunity should remain “secret”, the users are still there but re-defined. “Agents” provide them with context, information, support them and automate where possible, but the users retain the handling and important relationship with the customer.

This is what we’ve done internally as an easily repeatable model, specific AI tools built in nunq per use case: per software, customer, sales and so on. Each tool is built with a RAG approach, to ensure the retrieval of knowledge is efficient. We also regularly engage with the business and encourage feedback loops to improve, by starting small we have live tools that can be continually improved, as we grow and the tech grows.

However, the distinction is critical – Delegating tasks is not the same as delegating judgement. Used well, delegation creates leverage and saves time. Used without intent or governance, it creates risk.

AI Governance: Opportunity & Risk

With any opportunity comes risk, if you have a button that says “I will do your job for you today” would you click it some days? Outputs today are probably approximately correct which means sometimes you’ll get away with clicking the button, but other times you won’t, and over time do you know how to do your job or does the button?

There are other sides to this as well, AI-to-AI creates a new method of communication – this could be customer to customer service or internal systems talking. What might be nonsense from either side can be translated into different nonsense on the other, rather than the truth. This can lead to lack of control and accountability for communication made by the business – as agents are part of it.

This again is where governance matters, what can or should be communicated, accessibility, boundaries and keeping people accountable. These risks are not reasons to stop, but to design carefully with the end goal in mind.

Closing

Agentic AI is not about replacing people with tech, it’s about expanding the capability of users and evolving with the technology.

The best agents are secret, the capabilities continue to increase and they make users lives easier, invisible in the background.