Redesigning Roles Around AI: Lessons from the 9% "Pacesetters"

This week’s article looks at the part of an AI programme that almost every organisation intends to do and very few have actually done. Kyndryl’s 2026 People Readiness Report, a survey of 1,100 senior business and technology leaders across eight countries, identifies a group of about 9% of organisations that it calls pacesetters, and the first of the three things they do differently is that they redesign roles around AI rather than adding AI capability to job structures that have not changed.¹

Those organisations were 1.5 times more likely to achieve AI-related revenue growth and 1.6 times more likely to report stronger innovation in products and services.¹ Role redesign is not the glamorous part of an AI programme, and it is the part that separates the organisations getting a return from the ones still waiting for one.

Nearly everyone intends to. Very few have.

Intent is not the constraint. Mercer’s Global Talent Trends 2026 study, drawing on nearly 12,000 responses from executives, HR leaders, investors and employees, found that 98% of executives are planning organisational design changes over the next two years, and that 65% expect between 11% and 30% of their workforce to be redeployed or reskilled because of AI within that period.² On those numbers, organisational redesign is very close to universal as an intention.

Execution tells a different story. JLL’s 2026 Future of Work Survey, covering more than 2,200 senior leaders across 21 countries, found that only 15% of organisations have reached the optimising stage of AI adoption, while 46% are still tracking trends and 40% are analysing potential impacts.³ Set the two findings side by side and the gap is stark: 98% plan to redesign, 15% are operating at the maturity where redesign actually happens, and 9% are capturing the value. Planning an organisational design change is not the same as making one.

What redesigning a role actually means

Part of the difficulty is that the phrase is used loosely. Adding “AI-enabled” to a job title is not role redesign. Neither is issuing a licence and a prompting guide. Redesign happens at the level of tasks and decisions, which means starting from an honest description of what a person in that role does all day, then asking of each part which of four things is now true.

• The task moves to the system entirely. Nobody does it any more, and someone has to own the fact that it is being done without a person watching each instance.
• The task stays with the person but changes shape. Drafting becomes reviewing. Searching becomes verifying. The time it takes falls, and the judgement it demands rises.
• The task stays exactly as it is. Some work is not improved by automation, and saying so protects it from being automated badly.
• The task is new and did not exist before. Supervising the system, handling its exceptions, and explaining its output to a client or a regulator are all real work that has to be somebody’s.

A role that has been through that exercise looks different on paper and is described differently to the person doing it. A role that has not been through it simply has more expected of it than before, which is how capable people end up quietly absorbing the cost of an AI programme.

The questions a redesign has to answer

The organisations we see doing this well have written down answers to a short and awkward set of questions:

• What is this role now for? Stated positively, in a sentence, and in terms a person would recognise as a description of their job rather than a description of what the technology does.
• Which decisions does it own? Automation tends to strip the routine decisions out of a role and leave the difficult ones, which raises rather than lowers the seniority the work demands.
• What is the span of oversight? A person who reviewed the work of three colleagues may now be accountable for the output of a system running at a volume no individual can meaningfully check. If the ratio has changed, the role has changed.
• Where does the work go when it goes wrong? Escalation paths that assumed a human author need rewriting when the author is a system.
• What is the career path from here? If a redesigned role is a destination with nothing beyond it, the good people will read that correctly and leave.

None of those is answerable by a technology team, which is the practical reason redesign stalls. They are organisational design decisions, and they need someone with the authority to change how work is structured, not merely the authority to deploy a tool.

Why it stalls

Role redesign is expensive in the currency that is scarcest in most organisations, which is senior attention. It requires line leaders to describe their own work honestly, agree what moves, and accept that some of what their teams do will stop. That is slower and more contested than buying a licence, and it produces no demonstration that can be shown to a board.

There is also a sequencing trap. Deployment produces a visible result in weeks, redesign produces one in quarters, so deployment goes first and redesign is promised for later. By the time it comes round, people have already invented their own accommodations with the technology, and those informal arrangements are considerably harder to change than a blank sheet would have been. Preparing the operating model for agentic AI is the same argument one layer up, and it fails in the same order.

Redesign to what, exactly

Redesign assumes a destination. Mercer’s finding that 65% of executives expect between 11% and 30% of their workforce to be redeployed or reskilled only means something if there is somewhere to redeploy people to, and if the capability exists to move them there.²

That is where the plan most often meets its limit. The Conference Board found that training investment remains concentrated on upskilling people within their current roles, and that most organisations are not yet preparing for reskilling at scale.⁴ A redesign that moves a third of a workforce into roles nobody has been prepared for is a restructuring with a friendlier name.

For South African organisations the constraint binds harder. A survey of 426 Western Cape businesses, published in Development Southern Africa, found more than 70% struggling to find workers with the right skills, and most firms spending only 1% to 4% of their wage bill on staff training.⁵ Where the external market cannot supply the destination roles, redesign and reskilling have to be planned as one piece of work rather than two, because the people already inside the organisation are the only realistic source.

What the pacesetters actually do

Kyndryl’s three pacesetter behaviours are worth reading closely, because they describe a sequence rather than a menu. They redesign roles around AI instead of layering AI onto unchanged structures. They run deliberate change management so that the workforce understands the new operating model and guardrails are in place. And they build workforce readiness rather than assuming it.¹

Redesign comes first because the other two depend on it. Change management with nothing concrete to communicate becomes reassurance, and reassurance is quickly recognised for what it is. Readiness-building without a defined destination role trains people for a job description that does not yet exist. The order is the finding.

It is also worth noting what the pacesetter evidence does not say. It does not say that these organisations deployed more technology, spent more, or moved faster. All three behaviours are change implementation disciplines, and the same report found that 79% of leaders agree the speed of AI will outpace their organisation’s workforce, governance and operating models.¹ The organisations closing that gap are not out-running the technology. They are redesigning the organisation around it.

The reassuring finding, and its condition

There is a genuinely encouraging result in the JLL data. A majority of senior leaders, 60%, expect their workforces to grow rather than shrink, and the same proportion expect AI to reinvent human roles rather than replace them. The organisations furthest along with AI were the ones treating it as a workforce augmenter and actively redesigning roles to be enhanced rather than eliminated.³

That is a better picture than the public conversation usually allows. It comes with a condition attached, though, which is that reinvention is something an organisation does deliberately. A role is not reinvented by the arrival of a tool. It is reinvented by somebody sitting down and rewriting it, and the 9% figure suggests that the sitting down is the rare part.

Where to start

Redesign does not have to begin everywhere at once, and it should not. The organisations that make progress tend to start with one function where the AI impact is already visible, describe the current work at task level, decide what moves and what stays, write the new role definitions properly, and only then decide what training the new definitions require. That produces a worked example, an evidence base and a set of people who can explain the process to the next function, which is worth more than a programme plan covering everything and landing nowhere.

Redesigning the roles

This is the work CCG does, and role redesign is the centre of it:

• Redesign the roles and the workflow. CCG Consult works at task and decision level to translate an AI ambition into changed processes, escalation paths and clearly defined human roles, which is the redesign itself rather than a plan for one.
• Establish what the roles look like today. CCG Analytics measures readiness, adoption and sentiment, so that redesign starts from what people actually do rather than from what the job description says they do.
• Build the capability the new roles require. CCG Learn develops the judgement, supervision and verification skills that redesigned roles demand, so that a new definition is something people can actually perform.
• Settle the decision rights above it. CCG Advisory works with Boards and executives on the oversight and accountability model, which determines what any redesigned role is permitted to decide.

Almost every organisation now intends to redesign work around AI, and the intention is not what distinguishes the ones getting a return. Redesign is slow, contested and unglamorous, and it is the step that the 9% took first. If your AI programme has a deployment plan and no role definitions, you have the sequence the evidence says does not work.

If you are redesigning roles around AI, we would welcome the conversation. Get in touch with CCG.