Lead an AI Workforce Discussion Through Operating Questions

Executives often open an AI workforce discussion with a prediction about jobs. Move first toward work. A useful session asks where service slows, where employees repeat effort, what information managers lack, and what experience leaders want to protect. Those questions make a broad request concrete. They also keep a meeting from treating a technology label as a decision.

Begin with one operating outcome. A leader might say, “We need an AI strategy.” Ask for a recent customer or employee experience leaders want to improve. Perhaps a customer waits for a reply because staff search several records. Perhaps a supervisor lacks a current view of incomplete requests. State an outcome in observable language: “Reduce avoidable waiting between complete intake and routing.” Avoid starting with a proposed tool or a headcount estimate. A group needs a shared view of work before judging possible changes.

Map one process from trigger through completion. Record who receives a request, information arriving with each request, judgment required, system or record used, handoff, and exception path. Use a recent fictional case during practice instead of an ideal diagram. A case with a missing document exposes a different path from a routine request. Ask employees, “What reached your desk first?” “What did you look for?” “Who decided when an exception appeared?” “What sent work back?” Their answers show where a process description leaves out real effort.

Words deserve attention. An executive might call a step review while frontline staff call the same activity cleanup. Neither label settles meaning. Ask each person to describe action and output. Does someone compare two records, request missing material, select a route, or approve an exception? A precise work description creates better discussion later. “Staff compare submitted details against two internal records” tells a group more than “staff handle intake.”

After mapping work, describe a proposed AI role one step at a time. A proposal might sort incoming material, locate a record, draft an internal note, or flag an incomplete field for human review. Write four parts beside every proposal: input, expected output, reviewer, and next action. For example, a fictional team might test whether a suggestion list helps staff locate duplicate submissions. Input is two submitted names and reference numbers. Output is a suggested internal search path. An intake employee reviews suggestions, then decides whether a manager needs to see a possible duplicate.

Do not place every work step in one category. Drafting an internal checklist differs from sending a customer response. Finding material in records differs from deciding an exception. A group should ask what source material would enter a proposed workflow, who sees output, and where a human decision occurs. Technical, privacy, security, procurement, legal, people, or policy questions belong with client specialists authorized for those subjects. A consulting partner helps organize questions and options. Leaders make decisions through their own process.

Decision rights need their own page. Ask who selects work for study, who approves a limited practice, who decides what employee communication says, and who accepts operational changes. Name one owner beside each unanswered question. “Team will investigate” hides responsibility. “Operations director will supply two fictional cases for mapping by Tuesday” shows an accepted action. Avoid naming an owner unless a person has agreed to take work.

Choose participants for named questions. Operations describes flow, bottlenecks, and service priorities. Frontline employees describe work beyond a slide deck. Technology staff describe current systems, access patterns, and technical questions. People leaders raise job design, training, and communication questions. A finance or risk representative might identify internal review needs. An invitation should say why a person attends. Broad invitations often produce a room full of observers and no shared record.

A fictional workshop offers a practical sequence. Chief executive Noor says, “We need to know which roles AI will replace next year.” Partner Celeste answers, “Which customer process creates most avoidable delay, and what service outcome should stay protected?” Noor selects claims intake. Two recent fictional cases guide mapping. One arrived without a required document. Another had duplicate files under two customer names. Staff searched records, requested missing material, and routed complete cases for review.

Celeste writes each step on a working page. Customer submission starts intake. An employee checks submitted material. The employee searches two internal records. Incomplete cases receive a request for information. Complete cases move to a review queue. A manager examines exceptions. The operations lead identifies duplicate searches as a frequent source of delay. The technology representative asks for security review before any record retrieval practice. This meeting ends with questions, not a technology promise.

Noor asks, “Does this mean we need fewer intake employees?” Celeste answers directly. “Leadership needs a view of task time, quality checks, exception volume, and service goals before evaluating work options. Employment decisions deserve their own process.” Her answer neither evades staffing nor claims a future outcome. She places a serious issue after facts leaders need.

Build scenarios instead of predictions. In one fictional scenario, document checks stay unchanged while staff use an internal checklist draft. In another, complete cases receive route suggestions, with a manager reviewing exceptions. In a third, work focuses only on missing document messages. Give every scenario an operating aim, participating group, information source, review point, and question leaders expect to answer. Scenarios let leaders compare work changes without treating an early idea as a commitment.

Define observations before a session closes. A group might record elapsed time from complete intake to routing across a fictional practice set. Employees might note where they reject a suggestion or where a checklist misses a needed detail. A client owner chooses appropriate records for any future review and asks relevant specialists to assess proposed activity. Record assumptions beside unknowns. An assumption such as “duplicate files use consistent reference numbers” needs later checking before a group relies on such consistency.

Employee communication deserves early planning. Explain what leaders are studying, which work sits within scope, who receives updates, and which questions remain open. Do not label a temporary examination as a job change. Give managers a route for employee questions. Capture questions by theme, then respond only after owners have a current decision. A vague assurance harms trust when scope changes later.

End with a decision record. Separate observed facts from proposals. “Staff searched two records during a fictional case” belongs with observations. “A suggestion list might reduce search effort” belongs with proposals. Name missing information, owner, due date, and next meeting. Continue Noor’s workshop by scheduling a 45 minute mapping session for Tuesday. Before Monday noon, Celeste sends Noor one page with six work steps and three unanswered questions. Noor marks which manager owns each answer, then returns the page before session invitations go out.