Source: Roger L. Martin, 2025
Since the launch of ChatGPT three years ago, almost everyone has used Artificial Intelligence (AI) in some fashion to help with their work and the world collectively believes AI holds potential. I have written on AI subjects three times before in this series — on strategy & AI, on AI investment, and on AI’s impact on entry level hiring. The third was co-authored with friends Ahmad Zaidi, co-founder and CEO of AI start-up TransforML, and Gui Loureiro, Regional CEO Walmart Canada, Central America, Chile and Mexico and co-author of Reinventing the Leader. That team returns for this Playing to Win/Practitioner Insight piece on leadership and AI. It is titled A Leader’s Role in Fostering AI Superpowers: Mastering the Five Levels. And as always, you can find all the previous PTW/PI here.
Substitution vs. Augmentation
Every new technology that has or will come along has the potential to both substitute for humans and augment humans whether wheel, printing press, electricity, internal combustion engine, telephone or digital computer. For example, the printing press put lots of scribes out of jobs, but it also massively augmented the ability of humans to communicate their ideas, starting with the world’s first publishing mogul, Martin Luther! It is irrefutable that the printing press augmented humans to a vastly greater extent than it substituted for them. And that has been the case with every truly important technology the world has seen, including the list above.
The overwhelming public concern about AI is that it will become the first major technology to have its biggest impact by way of substitution. We don’t know how the balance between substitution and augmentation will play out. But it is clear that the easiest path is substitution and lots of people will provide advice on that front. We sincerely hope that those with the power to influence the direction of AI won’t focus their energy on spurring its substitution for humans. One way to guard against that negative outcome is to demonstrate the power of augmentation and that is our greatest interest because augmentation of humans with AI is what will really advance society — and within it, business.
We believe that it will take strong leadership to tilt the balance toward augmentation. To realize the potential value and make AI an augmentation superpower, modern leaders need to master each of the following five layers, each a deeper and more sophisticated augmentation that builds on the layer before.
First Level: How AI Augments Knowledge & Research
AI is now a common first-pass research assistant. It can map a space, surface sources, and draft a structured brief in minutes. Sophisticated deep research tools like iterative question decomposition, source triage, and synthesis passes that are incorporated into AI products help analysts quickly collect information that would have earlier taken days. That even includes the time it takes to verify sources and detect and correct the Achilles heel of AI, hallucinations.
For example, a mid-level manager might prompt AI to: “Size the Mexican hard-discount grocery market and identify three expansion risks.” On a task like this, AI can (1) outline demand drivers and competitor set, (2) pull public data points (market sizes, growth rates, store counts), (3) generate a comparison table, (4) list interview questions for two customer segments, and (5) draft a 1-page brief with assumptions and confidence levels. The human then validates numbers, adds confidential insights, and finalizes the narrative — augmented by AI.
Second Level: How AI Augments through Task Automation
Beyond research, AI can quietly take work off the plate of busy leaders, leveraging their time for other higher-value activities. AI can capture and crystallize notes, summarize meetings, tag decisions, extract owners and due dates, and push them to your tracker. It can cascade strategy, translating top-level objectives into team-level initiatives, propose KPIs, and keep a living “single source of truth” to ensure value delivery. This is the focus of the TransforML platform, which connects strategy choices to projects. It provides risks and weekly updates so leaders see progress, and accumulates blockers and deltas in one place. AI can automate routine operations such as converting emails to tickets, standardizing brief templates, generating weekly roll-ups, and pre-drafting stakeholder communications.
In this layer, leaders get leverage from AI by eliminating the necessity for countless monitoring meetings thanks to more and better machine-assisted follow-through.
Third Level: How AI Augments through Skill Leveling-Up
AI is a force multiplier for uneven skill profiles. Very few of us have completely consistent skill levels across the spectrum required in our jobs. Each of us has our stronger and weaker skill areas. AI can be used to level-up those weaker skill areas to give us a more consistent skill set through AI augmentation.
For example, we have seen a front-end engineer, who is brilliant in user experience design but less confident with logic structuring, use AI to propose architecture options and trade-offs, generate scaffolded components and tests, and help with complex logical flows. As a result, the front-end engineer has been able to add a disproportionate amount of impact through her unique strengths in developing user experiences that truly delight customers, without being held back by her skill deficits.
Fourth Level: How AI Augments through Blind Spot Detection
AI is relentless at checklisting the things we forget. Thanks to Atul Gawande and The Checklist Manifesto, we know that even highly skilled professionals need checklists to avoid blindspots. But AI can go far beyond a simple standard checklist. Point AI at a plan and ask: What’s missing? What could fail? It will probe dependencies, non-obvious stakeholders, compliance constraints, and capacity cliffs. It will test for coverage gaps. Used well, it becomes a second pair of eyes that flags risks early and attaches mitigation options with owner, time requirements, and resulting cost estimates. AI’s ability to sift through innumerable documents in minutes allows it to do this at a scale and speed that augments beyond human capacity.
Fifth Level: How AI Augments through Counterbalancing Groupthink
Groupthink is a well-documented problem (popularized by Yale professor Irving Janis in a 1971 article and 1972 book) by which teams converge too quickly on suboptimal decisions. To counterbalance groupthink, leaders can assign AI the role of devil’s advocate. Prompt it to argue the strongest opposing case, quantify downside scenarios, and stress-test assumptions. Which single assumption, if wrong, breaks this plan? This keeps debate vigorous and constructive without putting the interpersonal burden on a single brave colleague.
What Leaders can do to Move Teams Through the Five Layers of Augmentation
1. Set the strategy for AI
The leader must set the overall role of AI in fulfilling the company’s strategy. For example, Gui’s boss (Walmart corporate CEO) declared that Walmart will be people led, powered by technology. That helped Walmart employees understand that it wasn’t going to be technology first and people second — which would have indicated a substitution rather than augmentation agenda. As strategy evolves, it is important that people know what the company is ultimately trying to accomplish, and how.
2. Set the bar & the rituals
Be specific on ‘what good looks like’ for AI augmentation. For example, mandate AI-first drafts for research memos, require AI-generated action lists after meetings, and add a “devil’s advocate” AI check before big decisions. Make these part of the operating rituals, not optional extras. Once it becomes part of the ritual, people will quickly overcome any initial fear or hesitation.
3. Pair tools with training & metrics
The challenge is less with intelligent use — that will come quickly with practice — and more with adoption. Provide training on the approved tools to ease and sped adoption. Set guardrails to help users avoids unpleasant and unproductive downsides. Set and track usage with simple metrics such as percentage of meetings with AI action summaries or percentage of projects with AI risk reviews. Celebrate wins and coach the laggards.
4. Role-model and de-risk
As with most things in company life, this won’t take root unless leaders set a positive example. That lowers the implied risk for everybody else in the company. For example, leaders should show their own AI-assisted work (redlines, drafts, risk logs) and make it safe to iterate. Create a sandbox for sensitive work, define guardrails (privacy, IP, accuracy checks), and reward teams for using AI to improve outcomes, not just for “using AI.”
Practitioner Insights
Strategy is choice and as a leader (whether at the top or somewhere in the middle of your organization) in the modern era, you have the choice to focus on substitution or augmentation in your utilization of AI. It is a true strategic choice because the opposite is not stupid on its face. Some will focus more on substitution while others on augmentation. And we predict both will succeed in creating value — though in very different ways with differing societal implications.
That having been said, we believe the greatest upside will come setting goals focused on augmentation. That will require cleverness in defining how AI can most powerfully augment your business and then demonstrating the leadership to take advantage of all five levels of augmentation.
AI can give teams superpowers, but only leaders can help unlock them!
This is a really productive framework--thank you for sharing! Lots of great insights here, so I wanted to extend the conversation by playing the (human) devil's advocate.
On #2 (Set the bar & the rituals), I think mandates to use AI can actually make folks feel *more* scared of substitution if the tasks that AI will conduct used to be *their* tasks. If your job is to prepare research reports and a leader says now we'll do AI-first research drafts, you'll likely feel more in danger of being substituted than excited about your potential augmentation. Without clear descriptions of how to continue adding value, and how to do so in the new AI paradigm, folks may feel like they're "training their replacement".
Instead, it could be more productive to provide direct support for the researcher to reflect about their tasks, test different ways to augment them with AI, and allow them to develop their own augmented workflows--willingly. I believe positioning augmentation as a learning process instead of a compliance one will result in better long-term outcomes.
Roger, the augmentation vs. substitution frame is spot-on, and the five levels are useful guidance for how to deploy AI well. But I've seen lots of clients talking about "AI strategy" as something different from the above... Your levels seem to be more about operating imperatives, right? The opposite test: "Don't use AI for research." "Don't automate meeting notes." "Don't check for blind spots." All stupid on their face. That makes them table stakes to play the game, not chips to win it.
The harder question is: where can AI create an edge competitors can't copy in 24 months? That requires a real barrier! Proprietary data that feeds a better model, network effects where more users improve the AI, or process power that compounds over years... Without one of those, AI keeps you in the game but doesn't help you win. Your point about "setting the strategy for AI" gets closest to this: AI should *serve* the strategy, not become the strategy. The five levels describe how everyone will use AI. The strategic question is *where* it can build something defensible... If anywhere at all!