When AI frees up time, redesign the job

Executives are starting to see slack in their organizations as AI takes on routine work. In BearingPoint's 2025 survey of 1,010 senior executives, 92% reported up to 20% workforce overcapacity in legacy roles. The reflex is to turn that into a headcount number, and in Microsoft's 2025 survey a third of leaders said they were considering headcount reductions.

I think cutting is usually the weaker option. It captures the savings once, and competitors can do exactly the same thing. Using the same people to do more, or to do better work, is much harder to copy.

So far more companies seem to be choosing that second route than the headlines suggest. In EY's autumn 2025 survey of 500 senior US decision-makers, only 17% of companies seeing AI productivity gains said those gains had led to reduced headcount. Far more were putting the gains into existing AI capabilities (47%), new AI capabilities (42%), cybersecurity (41%), research and development (39%) and upskilling their people (38%).

Capacity is different from redundancy

Redundancy means more people than you need for what you're trying to do. Overcapacity means more capacity than you're currently putting to good use. If AI makes a sales team 20% more productive, you don't have 20% too many salespeople. You have time for deeper account planning, better market intelligence and a tighter loop back to product. So before asking who to cut, ask what higher-value work you can now take on. The closest analogy I know is the spreadsheet, which changed what finance people spend their days on.

Decide what AI does and what people do

I'd sort each role's tasks into three layers.

Automate: tasks AI can handle on its own, such as data processing, first drafts, routine scheduling, first-pass quality checks and summaries.

Augment: tasks where judgment and AI combine. AI synthesizes the data and a person interprets it. AI generates options and a person weighs the trade-offs. AI brings the account history to a customer conversation and a person brings the empathy.

Keep human: work where people remain clearly better and where your advantage lives, such as building trust, handling organizational politics, ethical calls in grey areas, strategic judgment and influence across teams.

The redesign itself has three steps: map how time is spent now, find the tasks AI can take, and decide what the freed time is for. As an illustration, take a customer success manager who spends 30% of the week answering routine questions, 25% on data entry and CRM updates, 20% analyzing customer health, 15% on account planning and 10% building relationships. With AI handling most routine questions, the data entry and the health dashboards, that might become 5%, 5% and 10% on those, with 30% on account planning and 50% on relationships. The title stays the same. The job moves from reactive support to proactive account management.

Four ways people and AI work together

Four patterns cover most of what I'd design for:

• Researcher and analyst: the person frames the questions and decides what's significant; AI gathers data, finds patterns and proposes hypotheses.
• Architect and builder: the person designs the system and keeps it coherent; AI produces the implementation and the repetitive construction.
• Strategist and executor: the person sets direction and handles exceptions; AI runs routine workflows, watches for anomalies and escalates edge cases.
• Conductor and ensemble: the person orchestrates several AI systems or agents and resolves conflicts between them.

The last one is quickly becoming ordinary. Microsoft's 2025 Work Trend Index predicts that every employee becomes an "agent boss": someone who builds, delegates to and manages agents. In all four patterns the person keeps the judgment and the relationships, and AI takes more of the execution.

Put the freed time to work

Freed capacity can go in three directions. Growth means serving new segments or adjacent markets with the same people. Innovation means more experiments and faster product cycles. Quality means deeper client relationships, better service and more careful work.

I don't have good evidence on which of these pays best, and I'd be wary of anyone who quotes precise multiples. The EY data does show some companies spending their gains on growth directly: 29% said they were cutting prices to win market share.

Change what you reward

Redesigned roles need redesigned incentives, or people keep optimizing for the old metrics. Activity measures such as tickets handled, documents produced or hours billed fall once AI absorbs the volume, so people look less productive at exactly the moment their work becomes more valuable. Reward outcomes and judgment instead, and reward people who find AI uses that others adopt, whatever their level. I've written about how that can work in compensation for the AI era.

Someone has to connect the two sides

Most organizations have AI expertise and domain expertise but few people who connect them: someone who understands a workflow well enough to redesign it and AI well enough to know what it can take on. The role goes by several names. Among the AI roles leaders told Microsoft they were considering hiring for were AI agent specialists (30%) and AI business process consultants (28%). My view is that you mostly grow these people. Look at domain experts who've built AI fluency through hands-on practice, technical people who've spent time in the business, and process-improvement specialists who've added AI to what they do.

Say what you're doing, plainly

This lives or dies on communication. People need to hear that AI is creating capacity and that the organization is investing that capacity in them, with specifics about how their roles will change. Phrases like "evaluating efficiency opportunities" or "optimizing the workforce" signal layoffs, and people will spend the freed time updating their resumes instead of doing the new work. Give them a say in how their roles change and real support in learning the new parts. The same autonomy and competence that drive AI adoption matter here.

I'd start with two or three roles that have measurable outputs. Map the time, redesign with the three layers, try the new split for a few months, and measure both the capacity freed and what it went into before touching the next roles.

What I can't tell yet is whether reinvesting survives tighter budgets. Redesigning roles pays off over years, while cuts show up in the next quarter's numbers. I don't know how many companies will hold the reinvest-first line when that trade-off gets sharp.