6 min. read

To redesign a role around AI, don’t start with the tool. Map the role’s tasks. Decide for each one whether AI should automate it, help the person doing it, or stay out. Redesign the handoffs between people and AI, set the measures, and name the owner. Done in that order, the job changes without the workflow breaking.

In AI 2041, the book he wrote with Chen Qiufan, Kai-Fu Lee argued that beyond relearning skills, “we need to recalibrate what today’s jobs look like with the help of AI” (excerpt). It’s the middle of his three Rs: relearn, recalibrate and renaissance. This post is a practical take on that middle step.

We’ll use one composite example throughout, not a client: an accounts payable specialist at a mid-sized company, whose day is mostly invoices, matching, coding, exceptions and vendor questions.

Step 1: Map the role’s tasks, not its title

A job title tells you almost nothing about where AI fits. The tasks do. List what the person actually does in a typical week, roughly how much time each task takes, what goes in, what comes out, and who receives it next. Include the unwritten parts: the vendor who always sends invoices late, the approver who needs a heads-up first.

That tacit knowledge is easy to miss. In one example from BCG’s Applied AI Index 2026, a multinational consumer company started by documenting how the work was really done, unwritten know-how included, before it redesigned a single process.

Step 2: Decide automate, augment or keep human, task by task

For each task, make one of three calls. Automate: AI does the task, and a person handles only the exceptions. Augment: AI helps, and the person keeps the decision. Keep human: the task stays with a person, perhaps with AI drafting or summarizing in the background.

The choice matters for more than efficiency. Stanford’s Digital Economy Lab finds that employment declines among young workers cluster in occupations where AI mainly automates tasks, while employment holds or rises where AI assists people, especially for experienced workers. We cover that research in AI isn’t ending work. Those are patterns rather than proof of cause, but they’re a reason to make the call on purpose.

A rough rule: high-volume, rules-based and easy-to-check work is a candidate to automate. Work that depends on context, relationships or judgment is a candidate to augment. Anything that carries authority or accountability, such as approving a payment, keeps a person in the decision.

Task (accounts payable)	Call	Why
Capture and log invoice data	Automate	High volume, rules-based, easy to check
Match invoice to purchase order and receipt	Automate; mismatches go to a person	Most matches are clean. The mismatches need judgment.
Code to the general ledger	Augment	AI suggests, the specialist confirms. Errors are costly and context matters.
Resolve exceptions with vendors and buyers	Keep human, AI-assisted	Relationships and judgment. AI drafts the emails and summarizes the history.
Approve payment	Keep human	Authority and accountability stay with a person
Answer vendor payment questions	Augment	AI answers status questions. The specialist handles disputes.
Flag duplicate or suspicious invoices	Augment	AI flags the pattern. A person decides.

Illustrative example. Your own controls, approval limits and regulations decide the real answers.

Your entry could not be saved. Please try again.

We sent links to your email! You should have received an email from us already. If you did not receive, make sure you check your spam folders and add masterofproject.com to safe senders list to receive our emails.

Step 3: Redesign the handoffs

Most redesigns break at the seams, not inside the tasks. For each handoff between AI and a person, write down:

What the AI passes on, and in what form the person sees it.
When it must stop and ask. The thresholds that send a case to a person, such as an amount, a confidence level or a new vendor.
How errors surface. Who hears about a wrong match or a miscoded invoice, and how fast.
Who reviews AI output, how often, and what they do when they disagree with it.

Draw the new workflow on one page. If the team can’t follow it, neither will the system.

Step 4: Set the measures before go-live

Take a baseline first, or you’ll never know what changed. For the accounts payable example: cycle time per invoice, exception rate, errors caught later (by audit or by vendors), rework, and how much of the specialist’s week has moved from routine matching to exceptions and vendor relationships. Decide in advance what result counts as working and what result means you pause.

Step 5: Name the owner

Someone has to own the redesigned workflow after launch: watch the measures, tune the thresholds, approve changes when a vendor updates its model or a rule changes, and decide whether to pause it. Without that person, the workflow drifts back toward the old one, or worse, quietly away from both. What that ownership involves each week is in what makes a company AI-native, and how it fits beside a Center of Excellence and a vendor is in who leads an AI-native transformation.

What the role looks like afterward

In the example, the accounts payable specialist spends far less time keying and matching and far more on exceptions, vendor relationships and spotting problems the system flags. That’s a different job. Treat it as one: update the job description with HR, decide whether it needs new training or a different level, and tell the person what’s changing and why before the system goes live, not after.

Ready to change the role? Five blanks to fill

Which tasks are we automating? Which are we augmenting? Who reviews the AI’s output? What number tells us it’s working? Who owns the workflow after launch?

If any blank is empty, the redesign isn’t ready.

Redesigning roles this way is becoming a job in its own right. If that’s the work you want to do, our guide to how to become an AI project manager covers the paths in, including a Registered Apprenticeship.

Master of Project Academy sponsors the nation’s first registered apprenticeship dedicated to AI Project Management, registered by the Maryland Apprenticeship and Training Council (MATC #2575, September 8, 2026) under the registered occupation Information Technology Project Manager (O*NET 15-1299.09). The registration was announced via Business Wire.

Frequently asked questions

How do you redesign a job for AI?

Map the role’s tasks, decide for each whether AI should automate it, help the person or stay out, redesign the handoffs between people and AI, set measures against a baseline, and name an owner for the redesigned workflow after launch.

What’s the difference between automating and augmenting a task?

Automating means AI does the task and a person handles only the exceptions. Augmenting means AI helps a person do the task while the person keeps the decision. Stanford research finds employment declines concentrated where AI mainly automates, with flat or rising employment where it assists workers, especially experienced ones.

Who should lead AI job redesign?

The people who do the work, their manager and HR all belong in the room. An AI project manager or a similar owner should run the process and own the redesigned workflow after launch.

Put someone in charge of the redesign

The AI Project Management apprenticeship, registered under the Information Technology Project Manager occupation (O*NET 15-1299.09; MATC #2575), develops one of your own employees into the person who redesigns workflows around AI and owns them after launch.

Book a 30-minute curriculum review →

About Master of Project Academy

Founded in 2012, Master of Project Academy has served 500,000+ enrollments across 180+ countries in project management training. It sponsors the AI Project Management apprenticeship, registered under the Information Technology Project Manager occupation (MATC #2575). Learn more at masterofproject.com/p/aipmapprenticeship.

Sources

How to Recalibrate a Role Without Breaking the Workflow