The Agents Are Here. You're Just Not Paying Attention.

The AI moment that gave us back our humanity.

Now accepting speeches, keynotes and assignments for 2026

I work with organisations and leadership teams on SuperSkills, human capability, and AI-era decision-making through speeches, workshops, keynotes and advisory work. In person or virtual. UK, Europe, global. → Book a discovery call | → Explore: thesuperskills.com

In addition to this newsletter, I recommend a few others. All free. Check them out here.

Friends,

An Austrian developer built a personal AI agent, named it after a lobster, and released it for free. Within weeks, OpenClaw had 150,000 GitHub stars, coverage in Nature, CNBC, and IBM’s research blog, security warnings from Cisco and Palo Alto Networks, and a companion social network called Moltbook where AI agents talk to each other while humans watch from the sidelines.

That’s the headline. But it’s not the story.

The story is that OpenClaw went viral because it does something most AI tools still don’t: it acts. Not suggests. Not drafts. Acts. It manages your inbox, books your flights, schedules your calendar, browses the web, controls your smart home, and remembers everything across sessions. All from a WhatsApp message. People are texting their AI from bed and waking up to a cleared inbox and a reorganised week.

And OpenClaw isn’t alone. Meta acquired Manus, an autonomous agent that hit $100 million in revenue eight months after launch, for over $2 billion. Genspark built a “super agent” that doesn’t just research your question. It picks up the phone and makes restaurant reservations using an AI-generated voice. Anthropic, OpenAI, and a dozen others are racing to give AI access to your files, your browser, your computer. Agent skill marketplaces are appearing where you plug in new capabilities the way you’d install an app on your phone. This isn’t one product. It’s a category being born in real time.

Most people won’t call any of this “agents.” They’ll just call it AI. The same way nobody ever really distinguished between “the internet,” “the world wide web,” “email,” and “a browser.” In the mid-nineties, normal people didn’t say “I’m using TCP/IP to access a hypertext document.” They said “I’m going online.” The technology stack didn’t matter. The behaviour change did. One day you were driving to Blockbuster. The next you were streaming. One day you were queueing at the bank. The next you were tapping your phone. The labels collapsed into a single word, the internet, and life reorganised around it.

The same collapse is happening now. Chatbot, agent, copilot, super agent. The industry loves its categories, but they’re already blurring into a single experience: you talk to AI, and things happen in the real world. That’s the shift. Not a new product. A new verb.

Something shifted. AI stopped talking and started doing.

There’s an irony here that’s easy to miss. In The Matrix, agents were the villains. Relentless, suited programmes hunting Neo through a simulated world, hissing “Mis-ter Ander-son“ before trying to destroy him. Their entire purpose was to keep humans locked inside a system they didn’t understand. They were the thing you ran from. Twenty-six years later, we’re inviting agents into our inboxes, our calendars, our bank accounts, and asking them to run the show. Same name. Very different pitch. Nobody’s running. We’re texting them from bed.

One agent planned a perfect three-day trip to Tokyo (flights, hotels, concert tickets, local transport) and forgot to book the return journey home. A human travel agent might have made the same mistake after a long day, and you’d have laughed it off. We hold agents to a standard of perfection we’ve never applied to ourselves. Even with the missed return flight, the agent saved four hours of planning. These tools are clunky. Early. Sometimes laughably wrong. But they work. Not perfectly. But enough.

And “enough” is where everything changes.

Everyone is talking about this. Almost nobody is asking the right question.

If you’ve been anywhere near LinkedIn, X, or a group chat this week, you’ll have seen it. Matt Shumer, an AI startup CEO, wrote a 4,000-word essay called “Something Big Is Happening,” comparing this moment to the early days of Covid. It’s been viewed over 60 million times. The New York Times discussed it on Hard Fork. Forbes published a response. Gary Marcus called it “weaponised hype.” Reddit tore it apart. Your colleagues have been forwarding it with messages ranging from “this is terrifying” to “this is nonsense.”

Shumer’s core message: AI is moving faster than you think, it’s coming for your job, and you need to start using these tools immediately or get left behind.

The backlash was instant. Critics pointed out that Shumer is an AI startup CEO with a direct financial interest in stoking urgency. That his claims about “perfect” code are exaggerated. That he conveniently ignores the hallucinations and security vulnerabilities that plague every current system. A Forbes analysis by a Harvard fellow called it “crisis narrative that conveniently aligns with commercial incentives.” Reddit users noted the irony of a doom essay that ends with a call to buy more AI subscriptions. Fair points, all of them.

But the dismissals are also wrong. Not completely. But enough. The engineers in those same Reddit threads who actually use the latest models daily are confirming that something genuinely shifted this month. A managing partner at a major law firm told Shumer that AI is already performing at the level of his associates. The gap between people who tried ChatGPT in 2023 and people using Opus 4.6 or Codex today is enormous. Dismissing the pace of change because the messenger is self-interested doesn’t make the change less real.

So here’s where it lands. Shumer is right about the speed. His critics are right about the hype. And both sides are missing the same thing.

Shumer’s advice is almost entirely “use the tools.” Learn AI. Get the paid version. Spend an hour a day experimenting. His critics say “don’t believe the hype.” And the more thoughtful responses that followed in the days after are asking the right question but struggling to answer it: how do we avoid human atrophy? How do we develop the capabilities that AI can’t replace? What should leaders and educators actually do?

The question is everywhere. The framework is missing.

Neither Shumer nor his critics say anything about what you should become. Shumer treats humans as users of AI. His critics treat humans as victims of it. Both frames leave you passive. One is frantic. The other is fatalistic. Neither helps.

The question isn’t whether AI is moving fast. It is. The question is what kind of person thrives on the other side. And that’s a question about human capability, not tool adoption.

This essay is an attempt to provide the missing framework.

The lens nobody is using

I’ve spent three years talking to over 200 organisations about how they’re responding to AI. The pattern is remarkably consistent. Companies look at AI through three lenses:

Process transformation. Map existing workflows, automate the repetitive parts, digitise the rest. It becomes a transformation programme with steering committees and Gantt charts.

Cost reduction. Identify which roles agents can replace, cut headcount, redeploy savings. Boards love this one because the numbers are clean and the impact is immediate.

Revenue upside. Use AI to build new products, enter new markets, serve customers in ways that weren’t economically viable before. This is the most exciting lens, and the rarest.

These are all legitimate. But they share the same blind spot.

None of them ask the question that matters most: what happens to human capability when the machines handle execution?

When you automate the tasks, cut the roles, and chase the revenue, what are you asking the remaining humans to be? What capabilities do they need? What needs to be developed, and what needs to be unlearned? What does it mean to be valuable in an economy where the machines do the doing?

This is the fourth lens: human capability. And almost nobody is using it.

The reason is structural, not personal. In most organisations, AI lands as an IT deployment problem. It arrives through procurement, gets rolled out through technology, and gets measured in licences activated and hours saved. It almost never reaches the CEO’s desk as a question about what the company needs its people to become. And so the most important strategic question of the decade gets answered by default, in the gap between what the tools can do and what the humans were never developed to handle.

The real story of the agentic age isn’t that AI is replacing our humanity. It’s that AI is revealing how little of our working lives was genuinely human in the first place.

We’ve been here before. Twice.

In the late 1940s, automation began eating manufacturing jobs. A generation of workers valued for their physical output, for their speed on a line, endurance on a shift, precision with their hands, found the economy suddenly rewarding something different. It wanted people who could think, coordinate, communicate, and manage information. The knowledge economy was being born.

The transition was disorienting, painful, and deeply unfair. But here’s what’s easy to forget: it didn’t just destroy jobs. It created entirely new categories of work nobody in 1945 could have imagined. The person in a 1948 steel mill couldn’t have conceived of a “user experience designer” or a “growth strategist.” By the 1990s, millions were building careers in roles that didn’t exist when their grandparents were working.

Then it happened again.

Think about what a phone was twenty-five years ago. You carried your Nokia everywhere. You played Snake on it. And you were happy, because the thing in your pocket could make calls and survive being thrown across a room. Then slowly, imperceptibly, more was added. A camera. A browser. Email. Maps. An app store. Games graduated from Snake to Angry Birds, from Scrabulous flipping to Words With Friends. Each addition felt small. Each one was a Trojan horse.

Nobody noticed the moment the phone stopped being a phone. Nobody announced that the pocket-sized device we’d adopted for calls and Snake now contained more computing power than Lex Luthor ever dreamed of. But somewhere between the ringtones and the touchscreens, the thing in your pocket became a device from which you could do multiple jobs from a park bench. The transformation didn’t announce itself. It just continued evolving.

Agents are following the same arc. A cleared inbox here. A booked flight there. A reorganised calendar. A research task you used to spend an afternoon on, done in three minutes. Each one feels minor. The compound effect is revolutionary.

We’re at the third inflection. This time it’s not manual labour being automated, and it’s not communication being mobilised. It’s knowledge work itself. The agents aren’t replacing your hands. They’re replacing your busywork. The coordination, the information gathering, the scheduling, the synthesising. The tasks that feel like thinking but are actually just processing.

Here’s the part that should stop you in your tracks: the knowledge economy took people with extraordinary human potential (curiosity, empathy, judgment, creativity) and buried them in scheduling, formatting, and status meetings. For decades, the most human parts of us were treated as nice-to-haves. The system rewarded processing speed, output volume, coordination capacity. It turned humans into middleware.

AI doesn’t threaten our humanity. It returns it.

The question is whether we’re ready.

This isn’t optional

Agent adoption will not be a strategic choice for most organisations. It will be a survival response.

Once one competitor uses agents to compress costs, accelerate response times, and operate leaner, everyone else inherits the new baseline. Boards don’t reward restraint. They punish underperformance. Leaders will adopt agents to hit targets first and rationalise the decision later.

Nobody chose email. Nobody woke up one morning and said “I’d like to spend three hours a day managing an inbox.” It spread because not using it became professionally impossible. One firm sends a proposal by email. The competitor who’s still printing, signing, and posting loses the deal before the envelope arrives. That’s not a technology preference. That’s natural selection.

Agents will follow the same path. Not adoption by enthusiasm. Adoption by economic gravity.

The unresolved questions around liability, around who’s responsible when an agent books the wrong flight, sends an inappropriate email, or commits you to terms you didn’t authorise, won’t slow adoption. They’ll be figured out after the fact, the way they always are. The economics won’t wait for the lawyers.

Part of the reason Shumer’s essay went so viral is that it captured the feeling of this inevitability. Not because his analysis was perfect, but because millions of people sensed the ground shifting and his essay put words to the vertigo. The steps were small (one model release, then another, Opus 4.5, Opus 4.6, Codex, an Austrian developer releasing a personal agent named after a lobster), but the distance covered was enormous. People weren’t reacting to a sudden shift. They were waking up to a gradual one. The fact that it felt sudden is the point. We just weren’t looking down.

What this looks like in twelve months

For early adopters, and there will be millions of them, daily work starts to feel unrecognisable.

Your admin disappears. Not gradually. In chunks. The scheduling, expense reports, invoice chasing, travel booking, vendor comparison, meeting prep. All delegated. The average knowledge worker spends 40% of their week on coordination and logistics. That time comes back. The question nobody’s asking: what do you do with 16 extra hours when you’ve built your entire identity around being busy?

You stop browsing the internet. When an agent can research, compare, and purchase on your behalf, “going online to find something” starts to feel like driving to a library to look something up. The browser becomes something your AI uses, not something you use.

Your agent talks to other agents. Moltbook isn’t a novelty. Within days of launching, it had 37,000 registered agents and over a million human observers. An entire coordination layer forming in real time that humans aren’t party to. What starts as AI agents posting memes at each other becomes your agent negotiating with a supplier’s agent, coordinating with your colleague’s agent, resolving issues with a company’s customer service agent. The infrastructure for agent-to-agent coordination already exists. The infrastructure for human understanding of it doesn’t.

Small teams become absurdly powerful. A team of three does what a team of thirty did two years ago. The economics of lean organisations become extraordinary. The economics of bloated ones become fatal.

Entire industries face existential questions. I remember vividly walking into a Lunn Poly on the high street to book a holiday. You’d sit across from someone with a laminated brochure and a desktop screen, they’s type away with secret codes and keyboard shortcuts until they’d find you a package deal to Majorca on dates you didn’t want. Lunn Poly had 800 shops across the UK. It was the country’s largest travel agency. The brand doesn’t exist anymore. Absorbed into Thomson, then TUI, then quietly dissolved. The internet didn’t kill Lunn Poly overnight. But it collapsed the gap between what people wanted and the effort required to get it. Agents are about to do the same thing to a far wider set of industries. Recruiters. Insurance brokers. Estate agents. Procurement departments. Any business whose value lives in that gap should be paying very close attention.

I remember something else. Sitting across from an insurance broker late at night in his garage. He had a desk and a heater in there. He was filling in a form in blue ink, slowly, asking us to spell our names letter by letter so he could write them in capitals. We paid in cash. He gave us change from a biscuit tin full of penny coins, because every product ended in 99p. The form was for travel insurance, which almost certainly wouldn’t reach the insurance company before we got back from Tunisia. If he actually posted it. That was the system. That was normal. Now you pull up a comparison site, pick a brand you recognise, check the reviews, and you’re done in three minutes. By next year, an agent will do even that for you. And you won’t need to think about it at all.

And in years to come, we will laugh. We will laugh about the hours we spent moving square objects on screens by millipoints to align them so the information looked presentable. We will laugh about the formatting, the fiddling, the pixel-pushing, the way an entire generation of knowledge workers spent a meaningful fraction of their careers making slides look right.

We will laugh about knowing where our files lived. We used to carry the mental map of every folder, every subfolder, every version-final-FINAL-v3. Was it .doc or .docx or .txt? Could you copy and paste the table from the spreadsheet into Word without it destroying the formatting? You learned the answer the hard way, every time. We were proud of navigating this. It felt like competence. Then came shared drives, and we all spent a year learning to let go. Now agents find what you need without being told where it is. The skill that mattered was never knowing where the file was. It was knowing what you were trying to achieve with it.

We will laugh about the day we had to sign a document. You printed it, assuming the printer would cooperate, which it wouldn’t, because your Wi-Fi-enabled printer had decided it no longer recognised your laptop. You found a pen. You squiggled something that could have been Picasso on a bad day. Then you scanned it back in, hunted for free software to merge the PDF pages together, and emailed it as an attachment with “signed copy attached” in the subject line. In a race, you’d have lost. One day we will look back at this and realise that all we ever needed was biometric authorisation and a tap.

All those hours of friction and pain, solved. Not to make us lazier. To make us finally use our brains. To free us up to do the real work.

Less meetings. More maxing.

Most meetings exist because humans are slow at sharing context.

The Monday stand-up exists because nobody knows what anyone else is doing. The project status meeting exists because information lives in twelve different places. The pre-meeting meeting exists because someone needs to be briefed before the real meeting. Agents share context instantly. The information problem that meetings exist to solve simply vanishes.

The meetings that survive are the ones that were always valuable: the difficult conversation, the creative session, the strategic debate where experienced people disagree productively. The ones that die were always secretly about coordination, status updates, and CYA. Which is most of them.

Here’s the uncomfortable corollary. A large fraction of modern management exists to compensate for the limits of human coordination. When those limits disappear, some roles will be exposed as structurally empty. Not because the people in them are bad. Because the problem those roles were solving no longer exists. Real leadership, the kind that sets direction, builds culture, makes hard calls under uncertainty, and develops people, will be more important than ever. But leadership was never what filled most of those calendars.

When your calendar empties, you can finally operate at your maximum. Not maximum busyness. Maximum capacity. Maximum depth.

The productivity trap

Every technology that promised to save us time filled that time with noise.

You know the sounds. The Outlook ding. The Slack ping. The Teams chime that means someone has typed your name into a channel you muted three weeks ago. The WhatsApp vibration pattern you’ve learned to distinguish from the calendar reminder. The Samsung ringtone across the train carriage that tells you someone over sixty is savvy enough to get a decent phone but hasn’t figured out how to turn on silent. We live in a symphony of interruption, and we’ve mistaken it for productivity.

Busyness is not flow. Busyness is reacting. It’s living in other people’s inboxes, other people’s priorities, other people’s pings. It’s the endless scroll, the notification badge, the reflex of checking your phone because you thought it buzzed when it didn’t. It’s setting Outlook rules to reach inbox zero as though that were an achievement and not the digital equivalent of the recurring 9:30am team meeting that nobody wants to attend but nobody has the courage to cancel.

Flow is the opposite. Flow is blocking two hours in your diary to think. Not to do. To think. It’s plotting things out on paper. It’s articulating why the growth targets are unrealistic before someone wastes a quarter chasing them. It’s making time to write something, properly, and discovering what you actually believe in the process. It’s being selective and intentional, not doomscrolling through an endless feed of someone else’s content. It’s stepping away from the screen because the best ideas don’t arrive while you’re staring at one.

This is what authorship requires. Not more tools. More thinking. And thinking requires something we’ve been systematically trained out of: the ability to sit with silence and not reach for a device.

There is no law of physics that says agents will give us this. The most likely outcome isn’t that everyone goes home at 3pm. It’s that the bar for output rises, the targets grow, and the freed-up hours fill with higher-order demands that are more cognitively exhausting than the admin they replaced.

But here’s what agents can do. They can take the Outlook ding away. They can handle the pings, the scheduling, the inbox triage, the noise. They can give you back the two hours you never had. What they can’t do is make you use those hours to think. That’s on you.

The skills that got you here are depreciating in real time. And replacing them doesn’t just mean learning new things. It means unlearning old ones. Letting go of the instinct that your value comes from being busy. Releasing the belief that being the fastest executor in the room makes you the most valuable. These reflexes served you well for decades. They’re becoming the thing that holds you back. The hardest part of any transition isn’t acquiring new skills. It’s grieving the old ones. The habits, the identities, the sources of pride that no longer map to where the world is going.

Where agents fail in ways that matter

Agents fail quietly. Without ego, emotion, or hesitation. They produce plausible outcomes that pass surface inspection but collapse under moral, relational, or second-order scrutiny.

An agent will draft a restructuring plan that’s financially optimal and culturally catastrophic. It will send a follow-up to a bereaved client on schedule because nobody flagged the bereavement. It will recommend a supplier based on data that doesn’t capture the fact your team has been burned by them twice. It will optimise your calendar for maximum deep work and delete the fifteen-minute coffee chat that would have stopped a colleague from quitting. It will streamline your team’s workflow by eliminating the informal check-ins that were quietly preventing burnout.

In none of these cases did the agent make an error. In every case, it made a perfect decision using incomplete values.

Agent failure often looks like success until it’s too late. The output is clean. The logic is consistent. And the thing it missed, the human context, the unspoken history, the ethical nuance, doesn’t show up in a quality check.

Then there are the failures that aren’t about nuance at all. Within weeks of OpenClaw’s release, security researchers detected hundreds of malicious skills, disguised as productivity helpers, actually delivering malware. Over 135,000 users had already given root access to agents they didn’t fully understand. The supply-chain attack surface isn’t theoretical. It’s active. Agents are powerful precisely because they have access to your data, your accounts, your decisions. That same access makes them an extraordinarily attractive target. Trust should be earned incrementally, not granted by default.

This isn’t a reason to avoid agents. It’s a reason to understand what they can’t see, and what you need to keep seeing for them.

The skills that stop mattering

In the 1950s, the steelworker’s physical endurance stopped determining his earning power. Not because endurance became worthless. Because the economy started rewarding something else more.

The same repricing is happening now. This time it’s coming for the knowledge worker’s equivalent of physical endurance: the ability to process, coordinate, and produce at volume.

Information gathering. You used to be valuable because you knew where to find things. Agents find everything, instantly, and don’t need a coffee break halfway through. Coordination as an identity. If your entire value is being the person who connects other people’s work, you are a human API, and APIs get automated. Polished documents. Those millipoints we laughed about? Agents handle professional formatting now, not eventually. The person who spent Friday afternoon making the board deck look perfect just lost their superpower. Routine analysis. If it follows a template, an agent will do it before you’ve opened the spreadsheet. Being the person who “gets things done.” When agents execute faster and cheaper, raw throughput becomes table stakes. This is the 2026 equivalent of being the fastest person on the assembly line the year after the machines arrived.

None of these become worthless overnight. But their market value is in free fall.

The Authorship Economy

Here’s the shift that matters most, and it needs a name.

For decades, professional value has lived in execution. Who can research it, build it, format it, ship it, coordinate it. The knowledge economy rewarded people who could process information and produce output reliably and at speed.

Agents handle execution now. What they can’t do is decide what’s worth executing. They can’t set direction. They can’t weigh competing values. They can’t author intent.

We are moving from an execution economy to an authorship economy. The valuable person is no longer the one who does the work. It’s the one who defines what the work should be, why it matters, and where the machine should stop.

What does that look like in practice?

The manager stops compiling the Thursday status deck and starts making the decisions the deck was supposed to inform, like whether to shut down a product line that three of her best people are emotionally invested in, knowing no spreadsheet can tell her how that conversation should go.

The lawyer stops drafting contracts and starts counselling clients on risk, judgment, and strategy, the parts of legal work that require understanding not just what the law says, but what the client actually needs and can live with.

The marketer stops producing assets and starts defining positioning, designing experiments, and making bets about what will resonate with real humans, work that requires taste, cultural fluency, and the willingness to be wrong.

In each case, the execution layer gets automated. The authorship layer, the judgment, the intent, the values, becomes the job.

But what most people are getting wrong about authorship is this. They think it means writing better prompts. It doesn’t. Authorship means doing the human work before you touch AI.

It means sitting in a room and actually discussing the problem before opening a copilot. It means picking up the phone and having a real conversation instead of blurting half-formed thoughts into a chatbot and passing off the regurgitated output as your own thinking. We’ve all seen it. The document that looks like it took three days because that’s how long it used to take, but everyone knows it took three minutes and a plastic copilot. Nobody’s impressed. Everyone can tell.

Authorship means sketching on a piece of paper first. Knowing your own mind. Understanding what you think before you ask the machine to think for you. The work is the thinking. The human skills come first. AI comes second.

Last year, I spoke at a keynote for a small media company that was starting to adopt AI in a big way. The energy in the room was electric. Even I was excited. People could see where they needed to go. They were genuinely excited about the possibilities. But when I spoke to leadership, there was no budget for a capability programme, only for the tools. I said: don’t wait for a transformation programme. Start changing now, before you hit the iceberg. They didn’t love that.

I spoke to one of the delegates recently. They’re using Copilot now. I asked what else had changed and how they use it. There was a long pause. They didn’t have anything else to say.

That’s the pattern everywhere. Tools deployed. Humans unchanged. And a growing gap between what the technology makes possible and what the people have been developed to do with it.

This isn’t a criticism of where most people are. Almost all of us have been tools-only for the past three years. Each new release, each shiny new capability, lands, gets adopted, and then we wait for the next one. It’s become a cycle of chasing tools without ever pausing to ask what the tools are for. The issue isn’t the people. It’s the frame. When AI is treated as a technology deployment, you get technology adoption. When it’s treated as a human capability question, you get transformation.

One group writes the brief. The other gets optimised by it.

This isn’t a technology gap. It’s a clarity gap.

You don’t need to be first. You need to not be last.

You’re not too late. Most of these tools are genuinely early. Rough edges, security risks, reliability gaps. The people diving in headfirst are doing valuable exploration, but they’re also doing a lot of troubleshooting.

What you can’t afford is to be the person still doing everything manually in eighteen months.

In the 1950s, you didn’t need to be the first person to use a telephone. But by the 1960s, if you were still sending telegrams, you were in trouble. Not because telegrams stopped working. Because the entire infrastructure of business had reorganised around the phone. The speed of everything around you assumed you were connected.

That’s where agents are heading. Not a nice-to-have. The infrastructure that everything else starts to assume.

What can you do about this? Tomorrow morning, delegate one task to an agent. Not your most important work. Something boring. A research task, a comparison, a scheduling headache. Feel what it’s like to describe what you want instead of doing it yourself. Then look at your last two weeks honestly. It might be uncomfortable. Print your calendar if you have to. Highlight every block that was genuinely you thinking, deciding, or creating. Now look at everything else. That’s the territory agents are coming for. Finally, block two hours next week with nothing in them. No agenda, no prep, no screen. Sit with a notebook and answer one question: if agents handled all the logistics of your job tomorrow, what would you actually do with the time? If you don’t have an answer, that’s the answer. Start there.

The SuperSkills

This brings us to the question that Shumer’s essay opened and never answered. That his critics raised and couldn’t resolve. That every thoughtful response since has been circling without landing on.

If the execution layer is being automated, what are the human capabilities that matter on the other side?

Not the tool skills. Not the prompt engineering. The human skills. The ones that make the tools worth using in the first place.

In studying over 200 organisations, I found that the people and teams who thrived through disruption shared seven capabilities, what I call SuperSkills: Curiosity. Change Readiness. Big Picture Thinking. Empathy. Global Adaptability. Principled Innovation. Augmented Mindset.

You already know what these look like. You’ve seen them throughout this essay. Curiosity is what makes the brief worth writing. Empathy is what catches the bereavement the agent missed. Big picture thinking is what tells you the system itself needs changing. Principled innovation is what stops you shipping something just because you can.

These aren’t soft skills. Calling them soft has always been the mistake. They’re the hardest skills in the building. And in an economy where AI handles execution and humans provide direction, they’re the whole game.

The choice

When agents act on your behalf, spending money, making commitments, managing your time, you face a question most people have never had to answer:

What do you actually want?

We muddle through because friction forces constant micro-decisions. Remove that friction and you need something most people have never developed: a clear operating philosophy for your own time, attention, and priorities.

Agents force clarity.

Lack of clarity gets automated too.

An agent without clear human direction isn’t your assistant. It’s your autopilot. And autopilots are fine, right up until the moment you need to land the plane.

Shumer told you to use the tools. His critics told you to be sceptical. Both are right. Neither is sufficient. The real question was always deeper than either side was willing to go: what kind of human does this moment require you to become?

It’s not AI versus humans. It was never AI versus humans. It’s AI plus humans. And the plus sign is where the real work lives. The curiosity, the empathy, the judgment, the courage to sit with a hard problem before outsourcing it to a machine.

The people who develop those capabilities will write their own brief. The rest will spend their careers executing someone else’s.

The agents are here. The only question left is what you’ll do with the humanity they give back.

Stay Curious - and don’t forget to be amazing,

Rahim

PS. If your organisation is grappling with how AI is reshaping work and leadership, I speak and advise on this.

Rahim Hirji Author, SuperSkills (2026) | Keynote Speaker | Advisor

Building human capability for the AI era.

→ thesuperskills.com

→ Book a discovery call

→ Connect on LinkedIn

Tools I Use:

• Jamie: AI Note taker without a bot

• Refind: AI-curated Brain Food delivered daily

Recommend Reading:

Now

1. Something big is happening. The essay I referenced above. You have to read it. Also worth reading the retort from Gary Marcus. In addition to the video above, see: Mustafa Suleyman sets out Microsoft AI’s goal of ‘humanist superintelligence’

2. The Best 30 books on AI - You need to read 4 or or more of them to keep up. Or just pick from this list.

3. 12 Sleepmaxxing Tips To Steal From Olympians - Athletes share the products, apps, and tips they use to maximize their Zs.

4. How The Times Is Digging Into Millions of Pages of Epstein Files - Two dozen journalists. A pile of pages that would reach the top of the Empire State Building. And an effort to find the next revelation in a sprawling case. I’m fascinated by how horrific this is. And I’m hoping the NYT are figuring out a way to get justice.

5. ‘Bigorexia’ Is On The Rise. Here’s What Parents Should Know. Although it can affect anyone, this lesser-known disorder commonly affects boys and young men.

Next

1. Ai Doesn’t reduce work, it intensifies it - I agree, in certain professions. Great HBR article

2. Scent, In Silico - smell is about to go digital. Youve been warned

3. A.I. Is Giving You a Personalized Internet, but You Have No Say in It

   The relentless addition of artificial intelligence in popular apps raises questions about what’s at stake. The answer: the future of the internet and its lifeblood, digital advertising.

4. “What’s Wrong With This Idea?” Simples

5. Moats in the Age of AI - If software becomes nearly free to build and AI models become commoditized, where does economic value actually get captured?

If you enjoyed this edition of Box of Amazing, please forward it to a friend or share the link on WhatsApp or LinkedIn.