The great flip

The great flip

Inside the great flip 

The great flip is what happens to human effort when execution becomes cheaper. Traditionally, the bulk of the effort in knowledge work goes into the specialist core – producing the analysis, the design, or the code – and less into the work before and after it. As AI takes on parts of that core, the work on either side matters more: framing the problem at the start, and judging the result and getting it adopted at the end. These activities can become the steps that decide how much value the work creates. In the most affected workflows, the familiar bell curve can even turn upside down. 

As execution becomes faster, bottlenecks may move towards framing, evaluation, and decision-making. If organisations do not redesign the surrounding workflow, they risk replacing one constraint with another rather than capturing the full benefit of AI.

Before the core, the work is framing: interpreting intent, assembling context, making constraints explicit, and deciding what good looks like. Cheaper execution helps here too, because alternatives can be tried before committing to one. After the core, the work is judgement and adoption: testing quality, weighing trade-offs, taking responsibility for the decision, and getting people to use the result. A recommendation creates value only when someone acts on it.

Economists Ajay Agrawal, Joshua Gans, and Avi Goldfarb have shown that cheaper prediction raises the value of complementary activities, human judgement among them,⁵ and that realising AI’s wider benefits means redesigning decision systems, beyond improving individual tasks.⁶ Generative and agentic AI extend this logic across much more of knowledge work. As execution becomes cheaper, framing, judgement, and adoption can become relatively more important. The great flip is our way of describing how that may reshape the distribution of human effort within a workflow. 

Faster execution also changes the workflow itself. When a first version costs less to make, producing it can help clarify the original question. Results from use can then inform the next iteration. When work runs this way, framing, execution, and adoption increasingly come to form a loop, with specialists moving among them rather than handing work along a fixed line.

AI expands the reach of expertise 

Specialist knowledge remains essential. Lawyers, accountants, and engineers must still understand the law, the accounts, and the system, respectively. But that expertise is increasingly applied to framing assignments, directing AI-supported work, challenging results, and deciding what can responsibly be used. 

The T-shaped professional, a long-established idea in management, offers a useful precedent. Traditionally, the vertical stroke represents deep expertise and the horizontal breadth needed for collaboration.⁷ The newly emerged bold T-shaped profile reflects two effects of AI: specialists can do more within their own discipline and perform substantive work in adjacent areas without becoming experts in them. Both strokes become thicker. Deep expertise remains vital for judgement, while some traditional handoffs become less necessary.

A lawyer can spend less time assembling a first draft and more time understanding the client’s objective, challenging assumptions, and advising on consequences. Expertise moves closer to the problem and stays involved after an output is produced. But AI-enabled capability is different from acquired expertise. A polished answer may contain weak reasoning, incomplete context, or a hidden error. Professionals must know what to delegate, what to verify, and, crucially, when to intervene. 

What if the edges compress too? 

Since AI also helps draft briefs, propose framings, and evaluate its own outputs, the edges may compress too. 

For now, three limits hold:

1. Someone still has to answer for consequential decisions
2. Relevant context often sits in relationships and unwritten constraints
3. Adoption depends on people changing what they do. 

Each of these limits may move as AI capabilities improve. What will likely endure longest is accountability and access to proprietary context.

Total effort is a different question. Figure 1 above draws both curves with the same area so that their shapes can be compared, but in absolute human hours the area could differ. It could shrink if AI compresses both the core and the edges; it could stay similar if time saved in execution is reinvested in framing, judgement, and adoption; or it could grow if cheaper execution encourages more experimentation, iteration, and new services.  

Demand matters too: fewer hours per assignment can coincide with more assignments across an organisation. The great flip therefore suggests a different distribution of human effort, while making no general claim about the total. For many consequential workflows, the constraint is shifting from producing an answer towards asking the right question, judging the response, and putting it to use. We treat it as a working hypothesis, and urge leaders to test it in their own work.

When competitive advantage shifts 

Companies have long competed partly through their ability to recruit, coordinate, and scale scarce specialist expertise. As AI makes parts of this execution cheaper and more widely accessible, specialist capacity alone may become less differentiating. In that sense, intelligence is becoming a commodity that almost anyone can access.

When competitors can reach similar capabilities, gaining commercial advantage depends increasingly on an organisation’s ability to identify valuable problems, apply proprietary context, make sound decisions, and turn results into customer outcomes. Of these, framing and judgement are the hardest to commoditise, because they depend on context, relationships, and accountability that sit inside the organisation and cannot easily be bought off the shelf. That makes it important to decide which intelligence to own or control, and which can be developed or sourced externally.

The great flip therefore has a further effect beyond individual workflows. It may change where companies capture value and what they should invest in to defend their position. This is a strategic hypothesis to test across business domains, not a claim that specialist expertise or execution capacity will cease to matter.

The three Rs: strategic decisions for CEOs 

As AI changes the economics of execution, leaders face choices that extend beyond productivity. The opportunity is to reconsider where the company competes, how work is organised, and how value is created – and three decisions can help companies seize it: 

Rethink advantage. Identify domains where cheaper specialist execution could weaken existing advantages or create new opportunities. Consider which capabilities will become widely accessible and what competitors will struggle to replicate. Decide which intelligence to own, develop, and rent. Proprietary context and customer relationships may become increasingly important sources of differentiation. 

Redesign the operating model. Reconsider functional boundaries, specialist handoffs, and management structures built around expensive execution, and challenge the assumption that existing workflows and roles should remain unchanged. Smaller cross-functional teams may be able to take greater end-to-end responsibility, with specialist expertise applied where it creates the most value. Design for people and AI agents working together from the outset, with clear decision rights, accountability, and continuous feedback. Reliable work across organisational boundaries depends on shared enterprise context, supported by suitable platforms and governance.  

Reallocate capacity. Decide whether released capacity should reduce costs, increase throughput, improve quality, accelerate innovation, or strengthen customer adoption. These choices have different implications for growth, profitability, and competitive positioning. Measure resulting business value rather than treating faster execution as the objective.   

The risk for established companies is that they improve individual activities while competitors redesign how the business operates. As the economics of execution change, existing structures may increasingly become constraints rather than advantages. If you were designing your company today, with AI agents capable of performing much of the specialist execution, would you build the same organisation? 

How to start: redesign one workflow 

Start with a business-critical workflow in a domain where cheaper execution could change competitive advantage. Use the FLIP approach (Find, Learn, Improve, Prove) to test that strategic hypothesis in practice: measure how human work shifts, what constraint emerges, and whether the redesigned workflow creates greater business value. 

• Find the core. Map the activities before, during, and after execution, including waiting and rework. Record the current distribution of human effort and identify the current bottleneck and where it may move. 
• Learn what AI can compress. Use representative cases to establish what AI can reliably execute. Include the human effort required for context, evaluation, corrections, and exceptions. 
• Improve the workflow. Redesign the sequence of activities, team boundaries, decision rights, and feedback loops. Determine where specialist judgement adds value and how cheaper execution enables new ways of working. 
• Prove and scale. Compare the redesigned workflow against the existing approach on time, cost, quality, and adoption. Measure how human effort has shifted. Scale only when the benefits hold beyond a demonstration. 

While the three Rs above can help establish the strategic direction, FLIP provides a practical way to test it. By starting with a business-critical workflow, leaders can discover where value moves, which organisational assumptions no longer hold, and what an agentic operating model requires. The lessons can then inform broader decisions about where to compete, how to organise, and where to invest.  

From automation to redesign 

The great flip suggests that AI’s greatest impact may come from redesigning how work is organised, rather than simply making existing activities faster. It is an emerging pattern to recognise and test, not a prediction that every workflow will follow it. But where it holds, the distribution of human effort and the sources of competitive advantage may both change.

Leaders should therefore ask what their organisation should look like when specialist execution is no longer the primary constraint. And as AI makes execution faster and more accessible, the time to address these questions is now, before competitors answer them for you.