Automation and AI in a Service Workflow

Lesson purpose and scope

Lesson: Automation and AI in a Service Workflow. Proposed placement: ATLAS University, AI Foundations for Automotive Teams. This placement is educational organization, not a claim that an accredited course or certification has been launched. Status: REVIEW, newly completed after approval of the three core recording scripts.

Audience: dealer principals, managers, advisors, BDC, sales and other employees who evaluate or use operational technology. No coding knowledge is required. Allow 60 minutes. Use printed case sheets or a shared display, paper, pens and a clock. The lesson uses fictional data and requires no customer records or external AI account.

By the end, learners should be able to distinguish a defined rule, an estimate, generated content and an action; explain why action alone does not prove AI agency; identify a useful source and a permission boundary; and design an exception handoff with a named responsibility. Success requires explanation, not vocabulary recitation.

Teach the distinctions

“A rule specifies what happens under a condition. A prediction estimates an outcome. Generation creates content such as a new summary. Reasoning is our working label for working through a problem; it does not prove human-like thinking. A tool can connect a system to information or actions. An action changes something or communicates it. Human judgment shapes the purpose, boundaries and handling of exceptions.”

“These are overlapping capabilities. AI and automation can be combined, and AI includes logical approaches. We are using a practical teaching model, not claiming a universal technical ladder.”

Fictional case sheet

Jordan writes: “Could I come Thursday afternoon? I need to understand the available options, and I need to know whether I can collect the vehicle by four.”

The fictional scheduling source lists eligible appointment openings at 1:30 and 2:30. It contains no completion-time commitment. The test system may read eligible openings and draft a response. It may not book or send. Maya is the advisor responsible for checking the response and routing unresolved commitments to the service lead. These permissions are fictional case conditions, not an ATLAS-wide operational policy.

The draft says: “You're confirmed for 1:30 and can collect the vehicle by four.”

“A fluent response can still change the meaning of the interaction. We check both the customer's request and the authority behind the answer.”

Pair exercise

Give each pair this changed case: “Jordan chooses 2:30. The tool's booking function is still disabled. A draft confirmation is waiting. The designated service lead is unavailable, and the calendar lookup now returns an error.”

Ask the pair to write five things: what is known; what is missing; what the system may do; what it must not claim; and who owns the next step. Then complete the four-column input, decision, action, exception worksheet. Allow ten minutes for writing and five for exchanging answers with another pair.

Learner worksheet prompts: What input began the task? Which source supports each proposed statement? Which capability is useful and why? What can change in another system? Is that change permitted? What exception stops the normal path? Who receives it? How would you verify that the handoff was accepted? What result would count as completion?

Tell pairs: “You can propose a different design if you explain its authority and evidence. You cannot assume a booking occurred, invent a completion promise, or treat an unavailable person as a completed handoff.”

Measurement exercise

Give these invented times, all in minutes. Current process: find information 2, prepare response 3, check 2, handoff 1. Assisted process: find 1, draft 1, check and correct 3, handoff 2.

Ask: What are the totals? How much did drafting alone change? How much did the full task change? What quality information is still needed?

Transfer and close

Ask each learner to select one real recurring task without collecting private details in the class record. They should name a process owner, a question to investigate, a safe test input and one outcome measure. A transfer case might be a Parts availability enquiry or a sales appointment reminder; the same analysis applies, but the learner must identify that department's actual sources and authority before use.

Say: “Your next step is to understand the task with the people who perform it. You are not authorizing a live rollout by completing this worksheet. Assess the work, train on the actual design, lead with named responsibility, adapt from evidence and define what success means.”

Repeat the opening confidence reflection. Ask what changed and what remains uncertain. Record one unanswered question for follow-up. Close by introducing Episode 04, What AI Could Give a Service Advisor, where the same distinctions will be applied across an advisor's day.