AI work redesign: why second-order thinking matters more than deployment

By Workplace Strategist · Published 3 September 2026

Most organisations are still asking the wrong question about AI. They ask where to plug it in. The stronger question is what work should look like once the technology is present at all. This article explains what Ethan Mollick’s argument about AI as a general-purpose force means for workplace strategy, and what your team should change before the next decision cycle.

This matters now because AI adoption is running ahead of AI redesign. Tools are arriving faster than the thinking about what they do to roles, workflows, and the physical and social conditions people work in. Workplace strategists who treat AI as a productivity add-on will keep optimising a workplace that is quietly becoming obsolete.

The source claim, stated plainly

In his UNLEASH interview, Ethan Mollick makes one central point: AI is a general-purpose technology, not a point solution. He compares it to electricity or the steam engine rather than to a single piece of software. General-purpose technologies do not simply make existing tasks faster. Over time they reshape what work is worth doing, how it is organised, and where value is created.

From that claim, Mollick draws three practical themes: purpose as the anchor for how AI is used, experimentation as the way organisations actually learn, and second-order thinking as the discipline that separates real advantage from surface-level automation.

The source is an interview, not a controlled study. It offers a well-argued frame and a practitioner’s judgement rather than measured outcomes. That distinction matters, because the value here is the thinking model, not proof of any single result. Our job is to translate that model into workplace strategy practice.

Why "where do we deploy it" is the wrong starting point

When AI is framed as a tool, the strategy question collapses into a shopping list. Which functions get a copilot? Which tasks get automated? This framing feels responsible, but it locks the organisation into its current structure. You end up making the existing workflow slightly faster instead of asking whether the workflow should exist.

A general-purpose force does something different. It changes the relative cost of activities. When certain analytical, drafting, or coordination tasks become cheap, the shape of a role changes. The meetings that justified themselves through information gathering lose their reason to exist. The desk time built around producing first drafts shifts toward reviewing, judging, and deciding.

The implication for workplace strategy is direct. If the nature of work is changing, then the assumptions behind your workplace model are changing too. Space ratios, collaboration patterns, focus requirements, and team rhythms were all designed around a version of work that AI is now dissolving. Deploying AI without revisiting those assumptions means optimising the wrong thing.

Purpose is the constraint that makes AI useful

Mollick’s emphasis on purpose is easy to read as a soft value statement. It is actually a practical constraint. Without a clear purpose, AI use spreads as scattered individual experiments that never add up to organisational capability. People find private ways to save time, but the organisation learns nothing and redesigns nothing.

For a workplace strategist, purpose translates into a prior question: what are we actually trying to make possible? Faster output is not a purpose. Better decisions, more time for judgement work, stronger collaboration, or reduced coordination overhead are purposes. Each of those points to a different workplace design.

This is where the connection to what people genuinely need from the workplace becomes concrete. If AI frees people from routine production, the workplace’s job shifts toward supporting the work that remains distinctly human: sense-making, negotiation, creative synthesis, and trust. That is a design brief, not a licensing decision.

Experimentation is a capability, not an event

The second theme, experimentation, is where most organisations underperform. They run a pilot, declare a verdict, and move on. Mollick’s framing suggests something more demanding: continuous, distributed experimentation where teams try, observe, and adjust as a normal way of working.

The strategic point is that a general-purpose technology cannot be understood in advance. Its useful applications are discovered, not planned. That means the organisations that win are not the ones with the best AI strategy on paper. They are the ones that build the fastest, most honest learning loop about what actually helps.

For workplace strategy this reframes the workplace itself as an experimental instrument. How do you structure teams, space, and time so that useful experiments happen and their results travel? A workplace that isolates people or hides their working methods will slow learning. One designed for visible practice and easy exchange will speed it up. This is a capability question, and it belongs in workplace strategy training and practice-building, not only in a technology roadmap.

Second-order thinking is the discipline strategists keep skipping

The most important idea in the interview is second-order thinking: not just what a change does directly, but what it does next. First-order thinking says AI drafts the report faster. Second-order thinking asks what happens to the skills of the people who used to write those reports, to how managers evaluate quality, and to what junior staff learn on the way up.

This is exactly the analysis that workplace decisions usually lack. A first-order decision cuts desk space because AI reduced individual production work. A second-order analysis notices that the same shift increased the need for high-quality collaboration and mentoring, which requires more and better shared space, not less.

The lesson for practice is that the direct effect of AI is rarely the important one. The second-order effects reshape roles, learning, culture, and the physical demands of work. A workplace strategy that only responds to first-order efficiency will consistently make decisions that look smart this quarter and become expensive within two years.

What organisations still get wrong

Three patterns recur, and all three come from treating AI as a tool rather than a force.

First, they measure success as time saved rather than work improved. Time saved is a first-order metric that tells you almost nothing about whether the work got better or the role got stronger.

Second, they let adoption run without redesign. People use AI privately, individual productivity ticks up, and the organisation’s structure, workflows, and workplace stay frozen in a pre-AI shape. The gains stay trapped at the individual level.

Third, they skip the second-order questions about skills and development. When AI absorbs the entry-level tasks that used to build expertise, the pipeline of judgement quietly breaks. This is a workforce and workplace question long before it is a training-budget line item.

What better teams do differently

Stronger teams start from work, not from the tool. They ask what the work is becoming, then decide what people, space, and process that redesigned work requires. AI capability is treated as an input to that redesign, not the goal.

They also build a real learning loop. Rather than one pilot and a verdict, they run many small experiments, capture what actually helped, and make those findings travel across teams. The workplace is deliberately shaped to support that exchange.

Most importantly, they make second-order thinking a routine analytical step, not an afterthought. Before any structural or workplace decision tied to AI, they ask what the change does next: to skills, to collaboration needs, to how quality is judged, and to what the workplace has to support. This is where a clear method matters. Connecting AI-driven change to a repeatable interpretation model, such as our workplace strategy framework, keeps these decisions consistent rather than improvised.

A clearer decision base for the next cycle

The practical takeaway is a small change in sequence with a large effect. Before deciding where AI goes, decide what the work should become. Before cutting or reallocating space, run the second-order analysis. Before declaring a pilot a success, ask whether the work improved, not just whether time was saved.

A stronger decision base for AI and the workplace answers four questions in order. What is the work becoming? What is the purpose we are optimising for? What did our experiments actually reveal? And what are the second-order effects on skills, collaboration, and space? A workplace strategy that answers these will keep pace with a general-purpose technology. One that skips them will keep redesigning the past.

Source: UNLEASH, "Purpose, experimentation and second-order thinking: HR’s AI blueprint for redesigning work" (interview with Ethan Mollick), 28 April 2026.

Build stronger workplace strategy capability

If AI is reshaping what work is, your workplace strategy method has to keep up with it. Workplace Strategist helps internal teams and advisors turn signals like this into repeatable practice: how to read a change, run the second-order analysis, and translate it into workplace decisions that hold up over time. Explore our courses and learning paths or get in touch to discuss building this capability into your team’s next strategy cycle.