The Risks

Seven Risks to Examine When Work Changes With AI

Changing work with AI raises questions about learning, workload, oversight and ownership. Use these seven prompts to examine a specific workflow with the people responsible. They are risks to investigate, rather than predictions about every organization.

Based on The Agentic Enterprise (Giridhar LV, Kashi KS, Rajan)

01

Junior development

Which learning opportunities disappear when an agent takes over routine work?

Preparing a first draft or checking source records can help a junior understand the business. If an agent takes over those activities, decide how the person will learn to recognize missing evidence, challenge a result and handle an exception. The effect depends on the responsibilities and supervision in that team.

What it means for you: Keep supervised practice in the redesigned role. For example, ask a junior to investigate a deliberately incomplete invoice record and explain the escalation to an experienced reviewer.

02

Overreliance on outputs

Can the team still investigate when the agent is wrong?

A fluent answer or a clean dashboard can make checking feel unnecessary. Identify the decisions where a missed error matters. People may need access to original records, tool logs and a fallback process to investigate the output rather than simply accept it.

What it means for you: Practise with incorrect or incomplete outputs. Check whether reviewers find the problem, explain their reasoning and take the appropriate next step.

03

Workload and agency

Ask people what the redesigned role asks them to carry.

Changing a role can remove work people value or concentrate difficult decisions in fewer hands. Talk with the people affected about workload, discretion and the responsibilities they want to develop. A task map alone cannot tell you how a change will affect their experience or employment.

What it means for you: Review the proposed role with its manager and participants. Track workload and unresolved cases during a bounded rollout alongside output and quality measures.

04

Improving one step while the queue grows

Where does the work go after the agent finishes?

Faster drafting may move the delay to review or approval. Trace the connected workflow before choosing the change: inputs, dependencies, checks and the owner of each handoff. Some activities can run in parallel; others depend on evidence or approval from an earlier step.

What it means for you: Compare end-to-end completion time, rework and review effort with the existing process. Agree the measures before treating a faster individual step as a business improvement.

05

Incomplete handoffs

The reviewer needs enough context to make the next decision.

An invoice exception needs more than a warning. The reviewer needs the invoice, the matching order and receipt, the rule that failed and the decision requested. Define who receives the case, what happens if nobody responds and what the system is allowed to do while it waits.

What it means for you: Test missing documents, conflicting records and unanswered escalations. Preserve the decision and its reason in the audit trail. Any change to the agent’s rules or behavior needs an agreed review process.

06

Unclear management responsibilities

Who owns the decisions between teams?

Managers may take on new work as agents handle drafting, routing or monitoring. Make responsibility for exceptions, approvals, performance review and staff development explicit. A proposed task split does not determine staffing levels or show that a management role is redundant.

What it means for you: Review cross-team cases with the accountable managers. Agree decision authority, escalation paths and the preparation each owner needs before changing the operating model.

07

Skills without a work context

Connect technical practice to the work someone will build or review.

Agent building, workflow configuration and model evaluation become useful in an enterprise project with relevant records, rules and decisions. Start with the responsibility the person will take on, then identify the technical and domain capabilities they need to demonstrate.

Illustrative project: an admissions coordinator configures a draft intake workflow with technical support, checks missing documents and routing, and rehearses escalation. Clinical and operational owners define the decisions people retain and the evidence needed. This is a preparation scenario, not a delivered hospital result.

What it means for you: Shape the project and assessment around the person’s starting capability and the actual workflow. Check agent behavior separately from the person’s assessment and the operating owner’s acceptance.

Capabilities to practise alongside technical skills

Choose the capabilities the changed workflow needs people to demonstrate.

Judgment

Making a decision when records are incomplete, including when to ask for more evidence.

Empathy

Listening to the people affected and checking what support or information they need.

Creative Synthesis

Using domain experience and different perspectives to propose a workable change.

Ethical Reasoning

Weighing consequences, obligations and competing interests before authorizing a change.

Ambiguity Navigation

Clarifying an unclear request into a problem, a responsible owner and a reviewable next step.

Examine these risks while you redesign the work.

Choose a workflow, a practical work product and the checks it needs. Your team builds or configures the agreed work with scoped Nuvepro support. Existing agents can enter evaluation directly; human preparation and operating acceptance remain separate.

Questions

These are risks discussed in The Agentic Enterprise and prompts for examining your own workflow. Their presence and severity require evidence from your organization. Illustrative preparation scenarios describe possible designs, not measured customer outcomes.
Start with the work and the risk you need to examine. Task Intelligence helps map responsibilities, decisions and handoffs. Scope simulations and environments around your systems, data and rules; agree preparation for people and separate checks for agents. Existing agents can enter evaluation directly. Your accountable owners decide operating acceptance.
Re-roling is the book's framing for redesigning a role around the work people will own. It connects existing expertise and judgment with the technical capabilities the changed work needs, such as configuring workflows, building agents or checking their outputs. Define responsibilities and starting capability before choosing preparation; a new title alone does not establish readiness.
These prompts draw on themes in The Agentic Enterprise by Giridhar LV, Kashi KS and Rajan. They organize a discussion about work design; they are not forecasts from a measured risk study.