Now accepting guest editorial from SAP, Oracle, Workday and ServiceNow practitioners.Pitch a guide
The Sign-Off Table

Lesson 4 of 6

Choosing use cases: the agent discovery approach

9 min

Listen to this lesson

0:00 / 5:40

Most failed AI pilots fail at selection: the team automates something that wasn't painful, or something too risky to trust to a machine. SAP's answer is the Joule Agent Discovery Workshop, published by SAP AppHaus with a free SAP Learning course. It is a two-to-three-hour session for three to ten people from business and IT that ends with one agentic use case described in detail. This lesson explains how it works and how Contoso adapted it.

The four steps

  1. 40 minGet inspiredShort presentation on agentic technology, how it differs from other AI and automation, suitable use cases and examples; Q&A
  2. 20 minIdeate automation scenariosEach participant documents 2 to 3 activities in the Automation Scenario template, using the Agentic Use Case Ideation Cards
  3. 30 minPrioritize use case ideasIdeas go on a prioritization graph; choose at least one in the top-right quadrant
  4. 20 min per caseDetail selected use casesMap the current process and roles, mark the most manual steps, describe how an agent helps, name roles for the follow-up AI Agent Design Workshop
The Joule Agent Discovery Workshop as SAP AppHaus describes it.
The Sign-Off Table · thesignofftable.com/academy/sap-ai-201

The prioritization graph

The prioritization step uses two axes. The horizontal axis is agentic potential, judged from questions about the task's complexity, its variability and its need for human judgment, with Agentic Potential Decision Cards to calibrate low, medium and high. The vertical axis is expected productivity increase, judged from the effort the task takes today. The top-right quadrant (high agentic potential, high productivity gain) is where the first use case should come from.

High gain, low agentic potentialMonthly dunning run review: valuable but rule-based; a skill or standard report may do
Top-right: start herePreparing collection calls and creating disputes from customer e-mails: many steps, varied inputs, real effort
Low gain, low potentialShowing a customer's open items: base Joule skill already covers it
High potential, low gainRare write-off decisions: complex, but too few cases to repay the effort
The workshop's prioritization graph, with Contoso's AR ideas placed on it.
The Sign-Off Table · thesignofftable.com/academy/sap-ai-201

Add risk before you commit

The workshop's axes measure fit and benefit. A process owner also has to weigh what happens when the agent is wrong. Before committing, Contoso adds three risk questions to each top-right candidate:

  • Reversibility: if the agent's proposal is wrong and a person confirms it anyway, how hard is it to undo? A draft e-mail is easy; a cleared item or a posted credit memo is harder.
  • Exposure: does the output reach a customer, a supplier, a regulator or an employee? External and personal effects raise the bar.
  • Volume and value: how many cases per month, and how much money is in each? This decides both the benefit and the size of a mistake.
From idea to pilot decision, with SAP's workshop steps and Contoso's added risk gate.
The Sign-Off Table · thesignofftable.com/academy/sap-ai-201

Preparing the workshop

SAP AppHaus recommends agreeing the focus with the business first: a business area, a department or a specific process. A workshop on "finance" produces a long, shallow list; a workshop on "accounts receivable for fleet customers" produces ideas people can act on. Invite the people who do the work, not only their managers, and bring a few facts: monthly case volumes, time per case and the most common exceptions. The prework instructions and the Automation Scenario template in the AppHaus kit help participants arrive with two or three real activities each.

Check SAP's catalog before building

A top-right idea may already exist as an SAP-delivered agent. Search the SAP Discovery Center AI catalog and the Help Portal before anyone opens Joule Studio. If SAP's agent fits, you inherit SAP's maintenance and its documented controls; if it doesn't, Joule Studio's intent-based flow starts with an AI-driven discovery step that maps your need to SAP's reference business architecture and does a fit-gap against standard assets, which is the same question asked by a tool.

What would you do?

A sales manager proposes an agent that automatically grants price reductions to fleet customers who complain about delivery delays. On the prioritization graph it lands top-right. What should the team do?

Key points
  • SAP's Joule Agent Discovery Workshop: get inspired, ideate, prioritize, detail one use case.
  • Prioritize on agentic potential (complexity, variability, judgment) and expected productivity increase (effort today).
  • Add a risk gate: reversibility, exposure and volume and value.
  • Check the Discovery Center and Help Portal for an SAP-delivered agent before building a custom one.

Check your understanding · not graded

Lesson check

  1. 1. What are the two axes of the prioritization graph in SAP's Joule Agent Discovery Workshop?

  2. 2. Which questions decide a use case's agentic potential in the workshop?

  3. 3. What does Contoso add to SAP's workshop before committing to a use case?

  4. 4. What should the team do before building a custom agent for a top-right idea?