AI or automation? What really lies behind most ‘AI solutions’

In practice, the picture is clear: an estimated 80% of solutions marketed as AI solutions are essentially based on traditional rule-based automation. In other words, they rely on predefined, logical workflows that process structured processes reliably and quickly. Only around 20% of applications utilise genuine AI capabilities, i.e. systems that can make decisions independently based on facts and patterns.

In Dynamics 365 SCM and Finance, many business processes are based on clearly defined rules and structured data. This is precisely where automation delivers its greatest benefits and achieves a high level of effectiveness and efficiency. Typical areas of application include invoice posting, triggering purchase orders, reporting, and master data reconciliation. Such workflows can be automated reliably and cost-effectively without the need for complex AI models.

Important: Automation follows clearly defined rules: if condition A is met, action B is carried out. These processes are characterised by being deterministic, which means that the outcome is predictable and reproducible. However, this requires that the underlying rules are defined fully and correctly.

AI, on the other hand, becomes relevant where data is ambiguous or decisions must be made based on patterns and probabilities. In Dynamics 365 SCM, this applies to areas such as demand forecasting, traditional MRP processing, the generation of intelligent purchasing recommendations, supply chain forecasting in procurement, dynamic pricing in sales, order matching and the execution of stock movements.

So-called agents are playing an increasingly important role: these are software-based units that carry out tasks independently, prepare decisions and interact with other systems.

In ERP contexts, such as Dynamics 365, agents can, for example, analyse incoming queries via the MCP server, initiate appropriate processes and, in cases of uncertainty, consult AI models for decision-making. In doing so, they combine automation (for structured workflows) with AI expertise (for complex assessments). The key advantage lies in orchestration: agents control when a process is executed based on rules and when AI support is required.

Companies should therefore not only distinguish between automation and AI knowledge, but also assess where agents can be usefully deployed as a connecting layer.

A common mistake in practice is to tackle automation problems with expensive AI solutions – or, conversely, to attempt to map complex decision-making processes using rigid rule sets. Both approaches require different technologies, skills and investments. Making the correct diagnosis at the start of a project saves considerable costs and effort.