AI Agents: From Buzzword to Business Value

AI is the buzzword of the moment. Everyone’s talking about it. The hype is massive—but what does it actually deliver for businesses? That’s the question many organizations are asking themselves.

We’ve all seen the examples of AI generating cool images or writing song lyrics in the style of Queen. Fun, but what are the real-world use cases that can actually speed up, improve or optimize our business processes?

That’s what companies are really waiting for. That’s why at McCoy, we focus on showing how AI can truly add value. Not someday. But today.
We’re doing just that with our AI-powered operational procurement agent.

What is an AI agent, really?

AI agents are autonomous computer programs capable of reasoning and taking action independently. They receive input, determine what needs to be done, and use the tools you provide to carry it out. This marks the beginning of truly autonomous AI.
But where did these agents actually come from?

Evolution: Where do these autonomous agents come from?

A short history in 4 eras

The history of AI is long—its first forms appeared as early as the 1950s! Still, the current wave of innovation can be traced back to four recent and relevant developments:

1. The birth of LLMs
For the first time, we had access to computers that could understand human input, like natural language. These models were predictive but didn’t offer real interaction yet.

2. The GPT era
Users could finally talk to AI. The introduction of interactive chat (like GPT-3.5) sparked massive adoption. Companies such as OpenAI (with ChatGPT), Anthropic, Perplexity, Grok, Google, and DeepSeek started popping up everywhere.

3. The AI assistant
Early chat interfaces were rich in knowledge, but often too generic. AI assistants gained the ability to add context and combine it with domain expertise. This led to task-specific assistants capable of providing answers that truly match your situation.

4. The AI agent
The logical next step: AI that not only thinks along with you, but takes action. AI agents combine contextual understanding with access to tools and systems. They can operate independently in applications and orchestrate processes. This creates a new generation of autonomous software that you can actually deploy within your organization.

The building blocks of an AI agent

Now that we know where AI agents come from, let’s look at what you need to build one yourself. An AI agent consists of four essential building blocks:

1. Instructions – behavioral rules and goals
The guidelines that define how the agent should behave. Think of it as a work instruction for your digital colleague: what is it allowed to do, when, and how?

2. LLM – the brain that understands language
A Large Language Model powers the agent’s reasoning. It enables communication and information processing. Choose the size and provider that fit your use case.

3. Tools – taking action in the real (or digital) world
The integrations needed to perform tasks. Think of APIs connected to SAP, Jira, Outlook, HubSpot or websites. These enable the agent to carry out actions independently.

4. Knowledge – context and data for smart decisions
Access to relevant information sources such as business data, policy documents, contracts, or catalogs. This allows the agent to make decisions that typically require human judgment.

AI agent in practice: supporting Procurement

Free-text requests are a thorn in the side of many procurement departments. In indirect procurement, the responsibility for submitting a correct request often lies with hundreds or even thousands of requesters. Catalogs and forms help, but there’s always a significant pile of free-text entries left that require manual checking.

Wouldn’t it be great if those checks could be automated?

The solution: an AI-powered operational procurement agent

At McCoy, we’ve built an AI agent using n8n that checks free-text requests in SAP S/4HANA in real-time for:

• The correct material group

• Whether the request should’ve been made via a catalog

This AI agent is already working with transactional SAP data. The result?
Fewer manual checks, faster turnaround times, and more time for strategic procurement.

Even better: you decide how many checks the agent performs and at which stage in the process. The two checks we’ve built so far are just the beginning—they showcase the potential of this technology in automating operational procurement.

How does it work? The step-by-step workflow

• Submit request
  The requester describes the required goods or services in SAP S/4HANA.

• Retrieve approval task
  An OData service retrieves the corresponding approval task.

• Retrieve request
  The free-text request is pulled from the system.

• AI review
  The AI agent checks the request for completeness and compliance.

• Return advice
  The agent’s advice, including substantiation, is automatically sent back to the system.

• Approve or reject
  The approver receives clear feedback and can approve or reject the request.

We’ve built three versions of this proof of concept:

• A version for S/4HANA

• A version for SAP Ariba

• A version built on Azure

Curious if this solution fits your business processes?

Feel free to reach out—McCoy is happy to explore the possibilities with you.

What’s next?

We’re seeing a lot of enthusiasm to get started. At the same time, many customers have questions:

• Am I ready for this?

• What steps do I need to take internally?

• Who should I involve?

Many organizations don’t yet have a concrete plan for this new technology.
In a next blog, I’ll dive into these questions and help you get started.