PDF Summary:Irreplaceable, by Pascal Bornet

As artificial intelligence rapidly transforms our world, many of us wonder how to remain relevant in an automated future. In Irreplaceable, former McKinsey executive Pascal Bornet tackles this question, drawing on his 20+ years of experience in the field. He explains that as AI spreads, our success will lie in learning to work with it—by augmenting our distinctively human strengths with this powerful technology.

This guide explores Bornet’s strategy for becoming indispensable amidst ever-better AI technologies, starting with his view of AI’s capabilities and trajectory (as of 2023). We detail his advice for developing key human capabilities that AI can’t (and won’t be able to) replicate—clear reasoning, creativity, and interpersonal connection—and we cover his recommendations for using AI in your work and at home. In our commentary, we’ll look at the broader discussion around AI’s impact on society, from techno-optimism to AI skepticism. We’ll also connect Bornet’s ideas to the work of experts like Gary Marcus and Ernest Davis (Rebooting AI) and Cal Newport (Digital Minimalism).

(continued)...

Capability #1: Creativity

Bornet asserts that current AI technology can’t compete with genuine human creativity. That is, AI can produce the appearance of creativity, but the appearances it produces lack real depth—they’re only a form of mimicry. This is because generative AI models (like ChatGPT) work by predicting the most probable next element based on statistical patterns learned from vast datasets. The model doesn’t truly “create” like a human does—it makes statistical predictions about what should come next in a given sequence of elements (like words or pixels). These predictions create things that accurately resemble human-made works, but which lack their emotional evocativeness, substance, or depth, according to Bornet.

For instance, say you ask ChatGPT to “write a short story in the style of Ernest Hemingway.” To do so, it’ll reference the billions of patterns it’s been trained on to predict a more or less appropriate sequence of words for that prompt. But it doesn’t know what any of Hemingway’s writing means, so the output will look like his work but lack its substance.

(Shortform note: In Rebooting AI, Gary Marcus and Ernest Davis contend that statistical learning fundamentally can’t produce real intelligence. This key machine learning approach has produced impressive results by using massive datasets to create algorithms with powerful pattern-matching abilities—like LLMs that have “learned” human languages. However, Marcus and Davis say, teaching machines to statistically correlate patterns in data doesn’t teach them what that data means. Lacking understanding, AI can make harmful mistakes that developers have to fix issue-by-issue. For instance, social media algorithms tend to amplify misinformation because it drives clicks, and they’re designed to correlate “more clicks” with “good content.”)

In contrast to AI’s statistical mimicry, humans can be truly creative. Bornet writes that this is because true creativity comes from lived experience. To create, a human draws on their personal experiences, emotions, and intuition—their unique sense of what’s meaningful. They channel that into their work, imbuing it with intention and feeling that evokes emotional responses in others. Their work is meaningful because it speaks of and from their personal lived experience. This distinguishes their work—say, a song—from an AI version that sounds just like it. Bornet says that because AI isn’t alive in a world of emotion, meaning, and context, it can’t and won’t ever be able to create as humans do.

(Shortform note: Philosopher Alva Noë argues in Strange Tools that art—a classic example of creativity—isn’t just about creating beautiful objects. It’s a way humans reflect on and organize our experiences, giving them meaning beyond their surface appearance. Making art involves investigating the sorts of activities we do naturally (like dancing or singing, as children do) by exploring the forms they take and could take. This helps explain why AI-generated images, while technically impressive, often feel hollow: They lack the layer of exploration and meaning-making that transforms mere arrangements of elements into art.)

If you want to develop your creativity, Bornet recommends that you practice mind-wandering. Instead of reaching for some entertainment, let yourself be bored during moments of daily downtime. This stimulates your mind to wander freely and make unusual associations. In turn, this encourages creative insights or ideas, since your mind can go in any direction and connect thoughts it normally wouldn’t. You can also encourage this sort of free-associative thinking by going for walks in nature and by slowly falling asleep or waking up without your phone nearby.

(Shortform note: Research supports Bornet’s emphasis on the benefits of mind-wandering. A 2009 study found that when we’re daydreaming, our brains are actually more active than during focused, analytical problem-solving. When our minds wander, multiple brain regions associated with complex problem-solving work in unison, and when insights arise, our brains exhibit a surge of gamma wave activity (a type of brain wave involved in associative thinking). That we’re unaware of this activity doesn’t mean it isn’t happening; mind-wandering isn’t as idle as we think. It’s also credited with sparking some of history’s biggest “aha” moments, like the discovery of penicillin.)

Capability #2: Critical Thinking

Next, Bornet explains that humans can think critically in ways that AI can’t. Critical thinking means using reasoning skills to scrutinize information and come to sensible, well-informed judgements. This involves identifying and questioning any premises, assumptions, and biases in context to learn and improve your understanding.

For instance, imagine you’re a new business owner selling candles online, and you see a dip in revenue during your first summer. You’ve been running a sale on new summer scents, but it hasn’t brought the returns you’d hoped for. To think critically about this, you reflect on the reasoning behind your decision as well as relevant information like market data. In the end, you find out that people don’t buy as many candles in the summer—something you’ll keep in mind for the future. This thinking process requires the ability to consider context and nuance, which AI struggles to do. It also asks you to think about your own thinking, another capacity AI lacks.

(Shortform note: Humans can think critically, but we don’t do so by default. In Thinking, Fast and Slow, Daniel Kahneman describes how our thinking uses two systems. System 1 is fast and intuitive—it uses heuristics, or thinking shortcuts, to reach quick, usually accurate conclusions. System 2 is slow and analytical. It involves consciously examining your assumptions, the framing of a problem, your logic, and more. In other words, System 2 involves critical thinking, whereas current AI excels at rapid pattern recognition, similar to System 1 thinking. While AI can’t yet replicate true System 2 thinking, techniques like chain-of-thought prompting can get AI to show its work by reasoning through a problem step by step.)

One key step Bornet recommends for thinking more critically is to learn about cognitive biases. These are built-in errors, or “glitches,” in our thinking. We’re all vulnerable to them, and they affect our thinking on an unconscious level. Start by learning about the most common cognitive biases (like hindsight bias—where we believe after the fact that we “knew something all along”). By becoming more conscious of these biases, you can avoid falling prey to them.

(Shortform note: In Kahneman’s view, cognitive biases tend to affect System 1, since it uses quick shortcuts rather than rigorous reasoning. He explains how to avoid biases in your thinking by distinguishing between intelligence and rationality. Intelligence refers to computational power (or brainpower). Rationality means thinking deliberately and reasoning things through. This is the key to avoiding cognitive bias: Take what feels right according to System 1, and then use System 2 to rationally work it out. This way, you’ll avoid the errors (cognitive biases) caused by taking mental shortcuts.)

Capability #3: Connection

The third AI-proof skill is connection. According to Bornet, connection is the ability to authentically relate with other humans to build relationships, empathize, collaborate, lead, and more. He says these activities will remain automation-proof because AI lacks flesh-and-blood lived experience, so it can’t relate to humans with real emotion or understanding.

(Shortform note: While Bornet describes connection as an innate human capability, developing it often requires intentional practice. One way to do this is through Authentic Relating, a practice that uses structured exercises to deepen human connection. Many of these take the form of games designed to build empathy and understanding. This creates what practitioners call “containers”—safe spaces where people can practice connecting. Bornet presents these connection skills as automation-proof, but it’s possible AI could eventually learn them. For one, some experts are working on “embodied” AI, which combines AI software with robotic hardware in the hopes of expanding AI’s real-world abilities—perhaps into connection.)

Bornet says that our ability to connect with one another involves three distinct abilities: emotional intelligence, communication, and leadership.

Emotional Intelligence

The first ability, emotional intelligence, is being able to feel and understand your own emotions and those of others. Emotional intelligence helps you form real and lasting relationships because everybody wants to feel seen and understood.

(Shortform note: In Emotional Intelligence, Daniel Goleman says emotional intelligence (EQ) has five parts: knowing your emotions, managing them, and motivating yourself, as well as recognizing others’ emotions and maintaining healthy relationships. Unlike IQ, which Goleman says remains fixed throughout life, you can learn about and improve your emotional intelligence. This matters because EQ plays a larger role in life success than IQ, meaning that you can do a lot to change your outcomes by working on your EQ. For instance, it’s involved in how successful your relationships, career, and family are—so all of these can improve if you improve your EQ.)

Communication

The second ability is communication, which means more than just talking to people. Human communication involves speech, non-verbal social cues, and awareness of cultural contexts and emotional subtexts. While people can communicate on all these levels, AI can’t.

(Shortform note: For the introverts among us, communication can seem daunting. But in How to Talk to Anyone, Leil Lowndes boils the skill down to one easy rule: Make others feel liked and appreciated. Humans have a basic need to feel valued, and this need governs all social interactions. When you make it clear that you like and are interested in someone, they’ll feel at ease and engage more freely. Lowndes suggests using body language—a relaxed, upright, open posture—to signal that you like people and are open to conversation.)

Leadership

The last ability is leadership, which Bornet defines as being able to guide, inspire, and influence others for the better. This relies on both emotional intelligence and communication skills. Leaders navigate complex interpersonal dynamics to coordinate and inspire groups of people to collaborate on large projects.

(Shortform note: When it comes to leadership, one approach that could help you better connect with others is servant leadership. In Servant Leadership, Robert Greenleaf defines this service-oriented leadership approach: He says that servant leaders (who don’t necessarily hold formal authority) prioritize the common good, uphold their social responsibility, and focus on inspiring others. In practice, this involves putting others’ needs first, helping their followers and communities to grow and thrive, and treating leadership as a spiritual calling. According to Greenleaf, this approach can empower any kind of institution (like churches, schools, and businesses) to be more socially responsible, which he says they currently fail to do.)

How can you begin building these connection skills? According to Bornet, you can learn to better connect with people by learning to empathize. He suggests practicing active listening, opening yourself to different perspectives, and becoming more sensitive to the many differences among various kinds of people. This will help you create inclusive spaces where people feel seen and included.

(Shortform note: To put Bornet’s advice into action, you might try a practice like Empathy Circling, a structured approach to building empathy. Participants take turns speaking while others engage in active listening, with each person reflecting back what they heard until the speaker feels fully understood. Then, it’s the next participant’s turn to speak. In rounds of 15 minutes or so, each person gets the chance to practice listening and empathizing as well as being heard. Empathy Circling has been effective at bridging gaps between ideological groups, like those for and against abortion as a form of medical care.)

Strategy #3: Become AI-Augmented

So far, we’ve covered how adaptability and humanity are your two key assets in the AI age. With that groundwork laid, let’s turn to Bornet’s advice on how to partner effectively with AI. As we explained earlier, human intelligence and artificial intelligence have distinct strengths and weaknesses. Because of this, humans and AI can do more together than we can alone. For instance, in a chess tournament, human-AI teams beat both human professionals and solo AI players. In this section, we’ll explain how to leverage this effect at work and how to avoid becoming too reliant on modern technology (including AI).

(Shortform note: In biology and physics, scientists call this phenomenon emergence. Emergence happens when multiple parts of a system (like a flock of birds) develop unexpected collective behaviors that individual parts can’t achieve alone. For instance, a flock of starlings flies in complex, visually stunning patterns that no one bird can produce alone. From this point of view, Bornet’s argument is that a human plus an AI can exhibit emergence. Some have already begun experimenting along these lines, with professional Go players realizing that AI plays in ways no human would—leading the human players to evolve new strategies in response. In other words, new ways of playing are emerging between humans and AI.)

Exploit AI’s Upsides at Work

Bornet writes that to use AI well, you need to think differently about work. Specifically, he recommends focusing on efficiency and the creation of value instead of long hours, busyness, and constant effort.

Work culture has traditionally valorized tireless productivity. But with AI, this will no longer be necessary because we can automate a lot of our busywork, like drafting marketing emails or internal memos. This leaves us with more time to slow down, collaborate, reflect, and live balanced lives. In the long run, this will be better for us and our work.

(Shortform note: Philosopher Peter Limberg writes that there are three groups with distinct outlooks on AI: people who are optimistic about it, those who feel pessimistic about it, and those who take a balanced view. Sam Altman (CEO of OpenAI), who believes AI will bring abundance and prosperity, exemplifies the first view, while Yuval Noah Harari sounds the alarm for those worried about AI and its existential risks. For his part, Limberg proposes recognizing that AI is powerful and holds potential, yet approaching it with caution. He says that if you imagine how bad things could get with AI gone wrong—a sort of Stoic negative visualization, like we explained earlier—you’ll become better able to see AI clearly and use it (or not) responsibly.)

Given the potential for AI to improve our work, the next step is to become AI literate. To do this, spend a portion of your work hours reading and learning about AI. Bornet says you don’t need to have world-class technical knowledge of AI, but you should know the basics of how it works. You should also stay up-to-date with recent developments, like new generative AI models or training techniques. This will help you bring useful new ideas into your workplace.

(Shortform note: If you want to become more AI literate right away, IBM offers free web-based resources for learning about AI. They cover the basics of AI design, machine learning, and big data, as well as many ethical questions related to AI, like the alignment problem (figuring out how to make sure AI serves human values). To keep up with new developments, you can sign up for newsletters like AlphaSignal or The Neuron, which deliver weekly or daily compilations of recent events, trends, and new tech releases.)

Alongside your efforts to learn about AI, start using AI to automate some of your activities at work. Bornet says that the best activities to automate are rote, rule-based, formulaic processes that have few exceptions to how they operate. For instance, a business owner might implement an AI tool to automate email responses to a set of common, repetitive customer questions that don’t need tailored replies. However, be sure you understand these processes yourself before you automate them. That way, you can fix them if something goes wrong with the AI.

(Shortform note: While AI is best at easily defined processes, tools like LLMs can also be fine-tuned to perform more specialized work. Fine-tuning involves adjusting a pre-trained AI model by training it on selected data. This teaches the model to handle more specific tasks without needing massive datasets, and it can help mitigate some of the limitations of current AI.)

Bornet offers a four-step strategy for implementing AI automation at work:

• Step 1: List and sort your work activities to identify those that consume the most time and yield the least value, as well as those that consume the least time and yield the most value. You’ll automate the former to free up time for the latter.
• Step 2: Standardize the work you’ll automate by breaking it down into a detailed set of steps. For instance, thoroughly define the steps AI would follow to successfully run monthly payroll. Refine the steps to ensure they work before implementing them.
• Step 3: Find an AI tool that fits your needs and run a small-scale trial with it. For instance, you might implement payroll automation on just a small portion of your company. Make sure it works on a small scale first.
• Step 4: Slowly scale up your AI automation, adjusting as you go. Say the above payroll automation worked for the C-suite—next, implement it for one additional department or team at a time. Monitor how it works and fix problems as they arise.

(Shortform note: Automation isn’t always as easy as following four steps—a fact illustrated by the peanut butter sandwich challenge from robotics. In this challenge, you follow steps like Bornet’s above, and the first is easy enough: Identify the task (make a peanut butter sandwich). The second is where things get rough. A robot (or AI) needs exhaustive, explicit instructions for things obvious and intuitive to humans, like how to pick up and move a knife, open a jar, and scoop peanut butter. Getting there requires lots of testing (cycling through Bornet’s steps 2 and 3), until each small part of the process works. As for step 4? Scaling AI poses a number of challenges, like creating datasets to fine-tune models for specialized tasks.)

Avoid the Downsides of Modern Tech

Last, Bornet stresses that you should avoid overusing modern technologies, including AI. Many modern digital technologies, like social media, encourage us to overuse them. For all its potential for good, AI also carries this risk. This is because its powerful responses provide a sort of instant gratification—send a few prompts, then get an image created or an article written in a snap. If we use this power unthinkingly, we risk becoming dependent on it and forgetting how to do things ourselves (like write, code, paint, or make music). To avoid falling prey to tech and AI overuse, follow these guidelines:

• Don’t use AI for anything that would give you joy. Instead, treat it as a tool for reducing arduous, rote work—not something to make your whole life more convenient. For instance, do your own drawing, creative writing, or music making.
• Set boundaries around your tech use. Schedule times for using your phone, whether for email, social media, or browsing. Also schedule times to put your tech away and focus on being present where you are, in the real world.
• Clean up your digital presence. Share only essential data and make sure you trust the organizations you give it to. Review your privacy settings regularly, and don’t be afraid to request that your information be deleted from accounts you no longer use.

Bornet clarifies that moderation doesn’t mean rejection. Don’t avoid using AI or technology entirely. Rather, be intentional and strategic about how you use it, and always make space and time for genuine human connection, real-world experiences, and enjoyment that doesn’t depend on technology.

Get Tech Savvy: Minimalism and Privacy

Bornet isn’t the only expert concerned about the downsides of technology. In Digital Minimalism, Cal Newport argues that if we don’t use it intentionally, technology can ruin our lives. This is because many devices and apps, like smartphones and Facebook, are designed to be addictive. They use intermittent positive reinforcement—such as likes on your post—to form your usage habits. This is the same mechanism that slot machines use, a fact that tech designers like Nir Eyal (Hooked) acknowledge. AI algorithms also play a part: They serve up personalized content for our individual feeds, increasing the chances that we’ll keep scrolling.

Newport argues that to escape this trap, we should be highly selective about which technologies we use. Tips and tricks (like setting tech boundaries) aren’t enough. We each need to overhaul our relationships with our devices. Newport shares three steps for doing this:

• Eliminate all optional tech (like social media or video games) from your life.

• Spend 30 days without that tech to break the habit of using it.

• Selectively reintroduce certain technologies, but only those you decide add more value than they remove.

Unlike Bornet (who advocates moderation), Newport suggests keeping tech away from what brings us joy. He writes we should reclaim moments of solitude instead of filling them with screen time. Solitude is good for your mental health, and reclaiming it is also a first step toward improving your relationships and normalizing tech-free leisure time.

Beyond providing health benefits, digital minimalism may also make it harder for tech companies to harvest your data—which allows them to track you across the web and serve you targeted ads. In The Age of Surveillance Capitalism, Shoshana Zuboff explains that tech companies gather data about what we do online, then they use that data to predict and influence our actions. For instance, Facebook might know that you’re more likely to engage with politically polarizing posts, so they feed you even more to keep you scrolling. To stop this, you can ask companies to delete your data, and to take control moving forward, you can practice Newport’s and Bornet’s principles for healthy, effective engagement with tech.