Agentic AI is emerging as the next step in the evolution of procurement. According to a new report from Gartner, 50% of supply chain solutions will use intelligent agents by 2030. How we get there and unlock the business value of the technology depends on what we’re doing now and what our goals are.
For those of you who are reading this blog for the first time, a quick recap: In the first installment, we talked about aligning AI with business processes as the only way for the technology to live up to its revenue potential. In the second installment, we explored what business alignment would look like across procurement-dependent verticals like Manufacturing, Higher Education and the Public Sector. In today’s installment, we discuss basic organizational preparation for Agentic AI.
Part of our focus will be on the cultural and organizational aspects of transformation. The highlight will be the step-by-step process for documenting business processes to train Agentic AI effectively.
Start with Who, Not with What
As with any transformation, everything starts with people. What are they doing now and how can Agentic AI help them do what they do better? Transformation cannot take place in a vacuum. You need to determine which procurement decisions can be automated, which need human approval, and at what threshold this approval occurs.
For instance, if you are using Agentic AI to manage complex sourcing events more effectively, your success will depend on integration with existing systems, strong governance frameworks, and measurable KPIs.
Also critical will be making sure that the team knows how to make the most of the technology. AI literacy is about more than tools. It’s understanding how to collaborate with work alongside AI agents, validate their output, and know how to turn a rate change or some other market development into a revenue opportunity.
Along with your internal team, your suppliers and partners also need to learn new ways of working. In some cases, they will need to adapt their systems to work with your AI agents. Updating contracts to address AI-related questions around data sharing and liability for automated decisions is recommended, as well as establishing service level expectations when machines are handling communications.
Documenting Businesses Processes for AI
Everyone talks about pilot programs. Everyone talks about transformation like it is a light switch we turn on, bringing sudden change to our organization and how we do business.
For anyone who has ever been involved in a pilot program, it is far less dramatic. Transformation begins with contained use cases and specific goals we want to achieve. Maybe we want to start with administrative tasks we want to transform, such as handling supplier inquiries or automating invoice matching.
No matter which area you choose, the steps for documenting business processes for training Agentic AI will be the same:
- Start with Clear Objectives: You need to define what you want the AI to do, whether it is answer questions, automate tasks, or assist team members.
- Use Structured Formats: Make sure you write in a clear format (much like this list!). If humans can’t understand the process, AI certainly won’t, which is why you should break processes into discrete steps, with decision points branching into other steps (If X, then Y) and exceptions that are automatically approved or require the immediate attention of the team.
- Include Context and Examples: In my experience, Agentic AI learns best from rich context. As you document each step, you should explain the “why” behind each step, not just the “what.” Providing real examples of expected outcomes is always a good idea, as is documenting common mistakes and how to avoid them.
- Capture Tribal Knowledge: Training manuals often only cover the most obvious use-cases. In reality, the most valuable information often goes undocumented, which is why interviewing experienced employees about what they do and how they do it is the only way Agentic AI will be effective.
- Make It Machine-Readable: Along with breaking everything down into a step-by-step format, you should make sure to tag content by department, function, or topic, use consistent headings and formatting, and create FAQ sections with clear question-answer pairs. I would also suggest structured formats like JSON or YAML to capture complex workflows (If you don’t know them, a technical team member can explain).
- Maintain Living Documentation: Processes change, old ways of doing things are replaced by new ways. The documentation you create should reflect the latest and greatest methods. It is only with such fluidity that Agentic AI will be able to continue evolving and learning alongside your team.
These steps, and the reasoning behind them, are hopefully enough to get you started. To tap the potential of Agentic AI, we all need to get to work now. Along with the documentation of the business processes we want to transform, we also need to talk about security and governance, which will be the topics in the next installments of this series.
In the meantime, I would like to leave you with the thought that each pilot, informed by the proper documentation, will help you understand how agents behave in a specific context, build trust with stakeholders, and refine your approach before scaling.
Go Deeper
Watch
REV2025 Keynote:
Gopinath Polavarapu (“GP”)
Chief Digital & AI Officer
