Despite massive investments, up to 95% of AI initiatives never reach real business value. Many get stuck in the pilot stage. Now, a shift is happening: from broad AI initiatives to specialized AI agents working directly within business processes. Implema has now developed five new agents, created by Claes Laine for Dynamics 365.
According to Gartner, up to 40% of all business applications are expected to include AI agents as early as 2026.
What is an AI agent and how does it differ from a standard AI model?
A traditional AI model is essentially reactive. It answers questions, generates content, or analyzes data, but stops there.
An AI agent goes a step further. It can take initiative, make decisions within given frameworks, and perform multi-step tasks, often by using multiple systems and data sources.
When agents are connected to ERP systems, they can not only analyze data but also act directly in processes such as orders, purchasing, and planning. In short: an AI model answers, an AI agent acts within the ERP system.
“It’s only when AI starts driving processes forward that real business value can emerge,” says Claes Laine.
Specialized AI agents provide precision and traceability
At the same time, we’ve seen a large proportion of AI initiatives that never even leave the pilot stage. A crucial shift now taking place is that many are moving away from the idea of an “agent that can do everything.” Instead, it’s about:
• smaller, clearly defined agents
• specific assignments
• high precision and traceability
This also means that AI is no longer a black box, but something you can understand, control, and improve.
Implema’s AI agents: five concrete solutions in the ERP system
What’s often missing in the AI discussion is the concrete aspect. Not what is possible, but what actually works in a real business.
When Claes Laine, Solution Architect for Dynamics at Implema, visited Microsoft’s major AI event in London, he and the team developed five functional AI agents in just three days, directly connected to core processes in the ERP system.
“We deliberately chose not to build anything broad or general. Each agent has a clear mission, linked to a specific friction point in the business. That’s where we see value being created immediately.”
Implema now has five new agents adapted for D365 and built for real-world processes:
- The Product Validation Agent
Ensures that products and configurations meet requirements before moving forward.
→ Fewer errors in early stages, less rework later on. - The Purchase Requisition Agent
Interprets needs and creates requisitions based on structure and policy.
→ Shorter lead times and less manual work. - The Cost and Margin Agent
Analyzes product costs and calculates margins.
→ Better business decisions and reduced risk. - The Order Management Agent
Receives, validates, and drives the order process forward.
→ More efficient flows and less administration. - The Vendor Communication Agent
Handles dialogue and follow-up with suppliers, such as status and deviations.
→ More structured communication and freed-up time.
“What happens when you connect them is the really interesting part. One agent takes over where another leaves off. That’s when we start seeing flows where work actually gets done, not just analyzed,” Claes explains.
From single use cases to automated flows
This is also where the connection to ERP systems becomes crucial.
“When agents work directly in ERP processes, with real data, something else happens. It doesn’t become a side solution; it becomes part of how the business actually functions,” says Claes.
The key is how they interact. When one agent triggers the next, we start approaching true automation. This reflects a clear trend: AI is moving from isolated use cases to connected, multi-step flows where work actually gets done.
Governance determines if AI agents can scale
With AI taking action, the demands for control also increase.
“When you give AI a mandate, you must set boundaries: what is it allowed to do, when should a human step in, and how do we follow up? Only then can it be scaled.”
The challenge is rarely the technology, but the connection to the business, data, and responsibility.
From idea to business value – the next step for Implema
For Implema, working with agents represents a concrete next step.
“We now have five new functional agents that we can demonstrate, test, and further develop together with customers. This allows us to go from idea to real value much faster.”
AI has left the pilot stage and stepped straight into the business and its ERP systems.
