Will Agentic AI Rewrite the History of ERP?

According to Gartner , by 2028, 33% of enterprise software applications will include agentic AI, up from less than 1% in 2024, enabling 15% of day-to-day work decisions to be made autonomously (Source: https://www.gartner.com/en/articles/intelligent-agent-in-ai).

Let's think about this for a moment... At present, hardly any companies are using true AI agents within their ERP solution, so if Gartner's prediction is to come true, there is a lot of work to do...

So what is agentic AI? Although there may not be a formal definition of the term, most analysts agree that agentic AI provides the following capabilities:

1. Ability to act autonomously.
2. Non-deterministic, meaning they are not programmed with predefined answers, but uses AI to formulate the answer dynamically.
3. Ability to automate work that is currently done by humans.
4. Self-learning.

Recently, Microsoft announced a slew of autonomous agents for the Dynamics 365 application portfolio. You can read more about this announcement here: https://www.microsoft.com/en-us/dynamics-365/blog/business-leader/2024/10/21/transform-work-with-autonomous-agents-across-your-business-processes/. Although Microsoft has announced these as agents, it is clear to me that only a couple of them are real AI agents - the rest are "just" AI applied to specific time-consuming problems, but let's see how it evolves.

These new agents are now slowly reaching public preview, so in the near future we will get the chance to deploy them and try them out.

Is This a Game Changer for ERP?

So back to the headline of this article. The question is whether agentic AI will substantially change how we implement and work with ERP?

If these agents become as "intelligent" as proclaimed, it is not difficult to see, how they can take over a host of repetitive tasks such as:

• Order management.
• Customer service case management.
• Dynamic supply chain scheduling.
• Supply chain disruption impact assess and mitigation.
• Automated bank reconciliation.

However, I believe the most important paradigm shift we will see is a transition to a new user interface where the ERP user interacts with a general AI agent instead of interacting transactionally with a database grid as we do today.

This will clearly have a positive impact on change management in a project where users probably will feel more comfortable with this approach and will require substantially less training. Today, we are relying on individuals to key in the right data in the right fields. This approach is by definition error prone. With an AI agent UI, the agent will ensure that data is correct.

It will also be interesting to see, if we start seeing configuration agents come along. A lot of time is spent in projects setting up interdependent data. With AI agents we may be able to automate this. The same goes for testing. It is easy to imagine a test agent that helps us generate and execute meaningful test cases based on user input.

Will it Happen?

This is the million-dollar question... These agents will certainly start appearing soon and we need to consider how we will use them, but there are some concerns that we need to address urgently as we transition to and an agentic AI future...

• Accuracy, LLMs are still "hallucinating" and that is not a big problem when you generate an image as the one I have used for this article. However, in a business application context, it may be a big risk. Humans are certainly not infallible, but we need to be certain that the AI agent provides highly accurate answers.
• Resistance, in the last 40+ years, ERP has claimed to be able to reduce human effort through automation. Although this may be true in a broader sense, it is not often, in my experience, that a company could substantially reduce their workforce after an ERP implementation. Will this change with agentic AI? I do not think anybody can tell at this stage, but the potential is there. How will that affect the existing workforce? Will they come along or resist?
• Data security, this is a topic that has been widely considered a major risk when applying AI to ERP data. Often we spend a substantial amount of time in an ERP project ensuring that data is secured and only available to the people, who need it. When we start applying agentic AI, it is essential that we operate through a robust and unified security model.
• Data quality, AI agents are relying entirely on data. The reality is that a high majority of ERP data is not of a sufficient quality to be used.

According to the Gartner's emerging technologies hype cycle for 2024 (Gartner 2024 Hype Cycle for Emerging Technologies Highlights Developer Productivity, Total Experience, AI and Security), autonomous agents are still five years from mainstream maturity. However, that also means that those, who want to take advantage as a first mover, need to be starting within the next 1-2 years. It is clear that blockers such as data quality and data security will take some time to remedy for any company, before agentic AI can be deployed.

To me, these are interesting times for ERP. I do not think anybody can say where this will take us in the end, but it is refreshing that the ERP human interaction paradigm we have relied on for 40+ years is being challenged...