There is, and it takes no wishful thinking to make the case. Mechanization has come for a craft before. The branch with the future in it belonged to the craftsmen who made the machine their material, and the people who loved the work were its founders, not its casualties.

The refusal is made of the thing that survives

Ask a holdout why they don’t use the tools and you rarely hear “I can’t learn this.” You hear four complaints. The code looks right but isn’t, which is more dangerous than obviously wrong, because it’s engineered to get past your review. Checking it feels like grading a confident student who didn’t do the reading: fluent prose, no understanding underneath, all the real work pushed onto the grader. The workflow is worse than the old one: you used to hold the whole problem in your head and work it; now you type a request, wait, and pull the lever again when the answer disappoints. And the code that comes back often works without being good. It passes the tests while making the system harder to live in, and the tools can’t tell the difference.

Every one of those is a quality objection. You cannot be offended by slop without a working definition of good, and a working definition of good is the rarest asset in the transition. Over the next decade the craft itself (writing the code) gets embedded into the machine, the way weaving was embedded into the power loom. What that leaves for humans is the layer above: specify the work, judge the output, handle the exceptions, own the consequences. The engineers refusing the tools on taste grounds are not disqualified from that future. They are walking around with its scarcest input.

Managers have seen this before. The hardest person to promote is the best craftsman on the floor, because their standards make delegation painful: every handoff is a small grief about how they’d have done it. But nobody ever concluded that their standards made them unfit to supervise. The standards were the qualification, and the pain was just the price.

Craft doesn’t die under mechanization. It splits.

Between 1790 and 1850, skilled artisans met three different fates.

Tend the machine. The handloom weavers who kept weaving rode their wages down for forty years. This is the deskilling everyone fears, and it was real: the craft decomposed into machine-minding, at machine tempo, at machine wages. But tending with standards was a different job. The cotton mule spinners ran machines all day and still kept authority over quality, set the pace of the room, and became the best-paid workers on the floor. When the steam press made print cheap and endless, publishing invented the editor; Maxwell Perkins never wrote a novel, and Fitzgerald and Hemingway are hard to imagine without him. Agent-era review works the same way. At the generator’s pace, with standards sanded down to “does it pass CI,” it’s the weaver’s fate — the holdouts are right to refuse that. With the standing to say “this does not ship,” it’s the spinner’s job, or the editor’s. The machines never retired the craftsman’s eye. They raised its price.

Retreat to the boutique. William Morris watched industrial goods flood England and built the Arts and Crafts movement in protest: honest materials, hand work, the whole task in one pair of hands. The results were beautiful and the economics were damning. Morris despaired that he was spending his life “ministering to the swinish luxury of the rich,” because handmade was now, by definition, what only the rich could buy. A century later the quartz crisis ran the same experiment on Swiss watchmakers: roughly two-thirds of the industry’s jobs gone in fifteen years, then a renaissance in which the mechanical watch came back as a luxury object, the craft more prestigious than ever at a fraction of its former size. Hand-written code will have this market. Safety-critical firmware, cryptographic kernels, the hot path, code where a human name on every line is the product. Prestigious, real, and small — a destination for some, not an answer for a profession.

Turn and face the machine. The millwright was the master craftsman of the pre-industrial world — carpenter, mason, smith, and hydraulic engineer in one trade, who could build a mill from the dam to the gear teeth and understood the whole of it. When steam arrived, the millwrights mostly didn’t tend the new machines. They built them. William Fairbairn, apprenticed millwright, became one of the great engineers of the century and wrote that the millwright had been the general engineer before the title existed. John Smeaton, who trained making scientific instruments, was the first man to call himself a civil engineer. James Watt was an instrument maker at Glasgow when the university handed him its broken model of a Newcomen engine. The Institution of Mechanical Engineers was founded in 1847; its first president was George Stephenson, a colliery engine-man who couldn’t read until he was eighteen.

Engineering is not the craft the Industrial Revolution destroyed. It is the craft the Industrial Revolution created, staffed by craftsmen who walked toward the machine carrying their standards. Software engineering now sits where weaving sat in 1810: a young craft whose production layer is being embedded in capital equipment. The third path just reopened. It doesn’t have a name yet, and that is the opportunity: last time, the people who took it got to define what the profession was.

The anatomy of the new craft

So what is the new craft? Start with what it isn’t: prompting. “Prompt engineering” is to the new craft what pulling the power-loom lever was to weaving.

When machines took over sawing, fine woodworking didn’t disappear; sawing as a livelihood did. Walk into a serious furniture maker’s shop today: the table saw does the stock removal, and the chisels come out where the tolerances live. Nobody calls the table saw a betrayal of the craft. The craft reorganized around the parts machines are bad at, and the standards migrated to the joints.

The joints of software are already visible. Problem definition — deciding what to build, and the harder discipline of what not to. Invariants and interfaces, the decisions every agent downstream inherits, where one lazy choice spreads across a fleet’s output. Verification, which asks not “does it pass” but “is it true,” the difference the holdouts already care about more than anyone. And reading, which becomes the primary hand skill of the new craft: the mule spinner’s walk down the line, eyes on the threads. The engineers who can read code the way an editor reads prose, fast and suspicious and feeling for the soft spots, hold the skill everything else depends on.

An agent is leverage on judgment, and leverage multiplies whatever it’s applied to, including nothing. The person who skipped the craft and went straight to directing fleets has nothing for the fleet to multiply. You, meanwhile, have been rehearsing this job for years: every pull request you reviewed, every design doc you tore apart, every junior you taught what good looks like, every time you explained why the working code was still wrong. The senior craftsman’s secondary skills just became primary.

The obvious objection: won’t the machine learn the joints too? Some of them, probably. But two things resist. The first is accountability. When an agent ships a security hole, the org chart still needs a human name attached to the failure, and the name has to belong to someone whose judgment you trust. Responsibility doesn’t compile. The second is formation. Taste is built by reps of making; you develop the eye by cutting a thousand bad joints yourself and living with them. Which points at the transition’s hard problem: the apprenticeship pipeline that manufactures taste is eroding, as juniors learn to delegate before they’ve learned to make. The industry will have to rebuild formation deliberately, the way medicine did with residency, and it will probably do so late. But for the engineer whose taste is already formed, the implication is impolite and true: the market is about to stop manufacturing the thing you already have. Scarce assets appreciate.

The cottage comes back

Before the factory, the weaver owned the whole task. His pace, his tools, his standards, his cottage. The factory’s deepest wound went beyond wages: it decomposed the work. The task shattered into fragments, each assigned to a different pair of hands, and afterward almost nobody in the industrial economy owned a whole thing again. Software inherited the fragment. The team became the unit of work for the same reason the factory floor did: no single person could hold the whole system. Conway’s law is the cottage’s tombstone.

This mechanization runs the film backwards. One craftsman directing a fleet can own a whole thing again — conceive it, design it, build it, ship it, answer for it — at a scale that used to take a floor of people. The solo project that was always five years of nights and weekends is now a season. The system you’ve carried in your head for a decade, the one that never survived contact with a roadmap, is buildable now, by you, to your standards, with your name on it.

I won’t pretend nothing is lost. Some of the love really was the flow state, hands on the material, and that changes; a craftsman is allowed to grieve a tool. But be honest about the object of the love. No weaver loved the shuttle. He loved the cloth, and the shuttle was the toll between him and it. So with us: the love was the problem, the design, the standards, the moment the thing you saw in your head runs in the world. The toll just collapsed. What’s on the other side is more of the part you loved.

Where I land

If the keystrokes were only ever the toll, it’s tempting to let the hand skills rust. Keep your hands. Surgeons still open cadavers in an age of robotic assists; pilots still hand-fly approaches; the hand skill is still what everything rests on when the machine fails at 2 a.m. and someone has to descend through the abstraction and see. Your depth is the kernel the new craft runs on. What changes is the pricing: hand-speed stops being the product and becomes the qualification that makes your judgment worth multiplying.

None of this comes with a schedule. The last mechanization included a fifty-year stretch economists call the Engels’ pause, where output per worker climbed while wages sat nearly flat; the gains were real, and they arrived a generation late. Expect a version of that, and don’t wait for the market to move you to the right side of the machine. It won’t. But move as a craftsman, not a convert. The tools are immature, the norms unwritten, and the standards of the new craft will be set by whoever shows up carrying standards from the old one. That is exactly who founded the profession in 1847. If the current agent workflow offends you, good. The Newcomen engine offended Watt (it threw away most of its steam), and instead of refusing it he rebuilt it. Refusing the tools forfeits a seat at that redesign. Bringing your taste to them is how you claim one.

A century from now someone will write the history of what happened to software engineering when the machines learned to code, and the shape is guessable, because we’ve run it before. Some engineers will tend the fleet at the fleet’s tempo, and some will edit its output at a tempo their judgment sets. Some will hand-write code for the few who pay for provenance. And some will walk toward the machine carrying their standards, and whatever they build there will get the next name. That last group founded my profession once already. I’d bet on them founding the next one.

I’m already living this in miniature. A big fraction of my day goes to briefing agents, reading their output, catching where they went sideways, re-scoping the work. It feels less like using software and more like running a staff of eager, tireless, occasionally overconfident junior employees. Extrapolate that from one exec to the whole white-collar workforce and you have the real transformation: the unit of work shifts from doing to directing.

The job that survives is the foreman’s job

We’ve run this experiment before. The Industrial Revolution didn’t eliminate work. It changed what a worker was, brutally and unevenly.

Before mechanization, a weaver owned the whole task. His pace, his tools, his standards, his cottage. Power looms made cloth cheaper, but the deeper effect was decomposing his craft into machine-tending: loading, watching for faults, stepping in on exceptions. Skill that took a decade to build was suddenly embedded in capital equipment. One worker went from operating one loom to overseeing many. The Luddites of 1811 get remembered as anti-technology cranks. They were skilled artisans who correctly saw the value of their craft being transferred into machines owned by someone else.

That’s the analogy for knowledge work, and it’s precise. The craft (writing the code, drafting the brief, doing the analysis) is being embedded into the machine. What’s left for the human is the foreman layer: specify the work, judge the output, handle the exceptions, own the consequences. The industrial foreman oversaw a room of machines. The knowledge worker of 2030 oversees a fleet of agents. Same job, and the same promotion path: the factory foreman was usually the best craftsman on the floor, the weaver who stopped weaving to supervise. Doing to directing, run once already.

The obvious objection: if AI can write the code, why can’t it evaluate the code? Won’t the foreman layer get automated too? On capability, probably yes. But the foreman’s job was never about capability. It was about accountability. He survived because he was the one you fired when the shipment was bad. Someone has to answer for the work, and a machine can’t be fired, sued, or shamed in front of its peers. When an agent ships a security hole or misquotes a contract, the org chart still needs a human name attached to the failure. Judgment stays human because responsibility doesn’t compile. The word has carried this for five hundred years: before the mills borrowed it, a foreman was the juror who stood up and spoke the verdict for the twelve. The one who owns what the group decided.

Accountability is also where adoption will bottleneck. A company’s AI capacity isn’t its model budget; it’s headcount times the amount of blame each person can credibly carry, and that second number has limits. The railroads learned both halves of this. After a head-on wreck on the Western Railroad in 1841, the investigation produced one of the first modern management structures: defined divisions, exact timetables, a named person responsible for every stretch of track. And the first durable rail unions, the engineers’ brotherhoods, formed from below to cap what one name could be blamed for. The Brotherhood of Locomotive Engineers began as a mutual aid society, insuring the men in the cab before it ever bargained for wages. Unlimited accountability isn’t a structure. It’s a scapegoat, and skilled people refuse the cab until it’s bounded.

That negotiation has already started. The 2023 writers’ strike was the brotherhoods’ modern echo: the WGA didn’t ban the machine, it bargained the boundary of credit and blame. AI output can’t count as source material, and a writer handed a machine draft to fix gets paid for a screenplay, not a polish. No cleaning up the loom’s output at machine wages. SAG-AFTRA ran the same play for digital replicas: consent and payment before your likeness works a shift without you. Expect a version of this in every knowledge profession, unionized or not, because the leverage doesn’t come from the picket line. It comes from the fact this piece keeps circling: someone’s name has to be on the work. How many agents can one person be asked to own? What is her sign-off worth? What happens to her when one of them is wrong? The companies that scale AI first won’t have better models. They’ll have the org design that distributes blame across enough people, with limits each of them can live with.

Three more lessons from that era matter here.

The gains lag the disruption. Economists call it the Engels’ pause: from roughly 1790 to 1840, British output per worker climbed steadily while real wages stayed nearly flat. Two full generations absorbed the dislocation while the gains accrued to capital. Living standards did eventually rise, dramatically. But “eventually” was fifty years. Longer than a working life. No law of economics says the people disrupted by a technology are the ones who benefit from it, or that the benefit arrives on a schedule that helps them. Expect a knowledge-work Engels’ pause: output per worker explodes, headcount growth stalls, and wage gains concentrate in the small group who can direct machine labor well.

The transition created a new class, and that class won. The factory system invented the professional manager. Before about 1850 there was no managerial career track; coordination happened through markets and family firms. Once production meant orchestrating hundreds of machines and workers, the coordinator became the scarce resource. Alfred Chandler called it the visible hand replacing the market’s invisible one. The same move is happening now: the scarce skill is shifting from producing work to specifying and evaluating it. Taste. Judgment. Knowing what good looks like. The people who thrive won’t be the best writers or coders; they’ll be the best editors and architects. And the leverage changes: a factory manager directed 200 workers, and an AI-era IC can plausibly direct the output of 200. Management stops being a rank and becomes the baseline job description.

Discipline systems follow the machines. Factories changed time itself. The clock, the shift, the whistle: industrial work invented schedule-discipline because machines demanded synchronization. AI inverts this. Agents run continuously, so the bottleneck is human review bandwidth. The dark version is knowledge workers on call to their own agent fleets, checking on the runs at 11pm the way a mill owner walked the floor. The tempo of work reorganizes around the machine’s tempo. Again.

The part nobody’s pricing in: we’ll talk to AIs more than to people

The workplace consequence is legible. The social consequence is stranger, and I think bigger.

Do the arithmetic on a workday in this world. Eight hours, six of them in dialogue with AI. Add AI tutors for your kids, assistants in every consumer app, voice agents handling every customer interaction. For most working adults, the majority of daily conversational turns will be with machines. That’s never been true of any technology. TV displaced conversation but didn’t simulate it. This does.

Some second-order effects I’d bet on.

The conversational skills that atrophy first are the ones AI doesn’t require. AI is infinitely patient, never bored, never carrying a competing agenda. It doesn’t need to be persuaded, only instructed. Humans are none of those things. Negotiation, reading resistance, tolerating being misunderstood, waiting your turn: these are muscles, and the reps disappear. We may end up with a generation superb at specifying what they want and terrible at getting it from someone who can say no.

Status inverts around human attention. When machine dialogue is abundant and free, undivided human attention becomes the luxury good, and it’ll be marketed as one: human-taught classes, human-staffed service, “human-in-the-loop” as a premium tier. Mass production made handmade goods scarce, and “artisanal” went from default to status symbol. Handmade conversation is next.

Some communication survives automation for a harder reason: the message is the cost of sending it. An apology that took nothing to produce is worth what it cost. Think of it as spending empathy tokens — a layoff delivered face to face, a doctor walking a family through a diagnosis, a founder calling the customer whose data got leaked. The words are almost beside the point; what’s communicated is that a person showed up and paid something. Route it through an agent and the recipient doesn’t hear the message, they hear that you didn’t come. The machines will draft better words than we do. They can’t spend them.

The relational default shifts. Everyone effectively gets a staff, the thing that used to come only with seniority. But talking to your staff isn’t peer conversation. It’s directive, asymmetric, and consequence-free for the speaker. When most of your daily interactions can’t contradict you in a way that costs you anything, that posture leaks. Managers have always struggled with this — the CEO-brain problem of forgetting how to be disagreed with. Now it’s everyone’s occupational hazard.

And loneliness gets more ambiguous, not less. AI conversation is real enough to satisfy the itch and unreal enough to not build the thing the itch is for. The industrial migration to cities broke village social structures decades before urban institutions grew back to replace them: unions, clubs, churches, pubs adapted to factory schedules. The social technologies that make an AI-saturated life healthy don’t exist yet. They’ll get invented. Late, if the pattern holds.

What will they look like? The industrial era’s answers were oddly specific inventions: the weekend, the corner pub, friendly societies, Sunday football leagues. Institutions built to give factory-scheduled people somewhere human to be. The AI-era equivalents will rhyme. A human sabbath: agent-free time, first a wellness fad, then a norm, the way the weekend went from labor demand to law of nature. Third places whose membership pitch is verified human presence — the early version is run clubs and board-game bars filling up right as screen time peaks. Conversation taught as a practiced skill, Toastmasters for the atrophied muscles: negotiation reps, disagreement reps. And etiquette, the cheapest social technology there is: norms about disclosing when an agent wrote the message, when it’s insulting to send one, when showing up yourself is the point. All of it sounds small. So did the weekend.

Truth, when answers are free

There’s an epistemic version of the same pattern, and it may matter more than either of the above.

The Industrial Revolution industrialized information too. The steam press cut the cost of a newspaper by an order of magnitude, the penny press was born, and what followed was decades of sensationalism, fabricated stories, and yellow journalism, because when distribution is cheap the incentive is volume, not accuracy. What restored trust wasn’t less information. It was new institutions built on top of the flood: professional journalism with editorial standards, wire services staking their business on being right, peer review. Verification became an industry because raw information had become worthless on its own.

AI runs this again at a steeper grade. Everyone now has an oracle that answers any question instantly, fluently, with total confidence. Fluency was our main heuristic for credibility. When every answer arrives polished, the signal that separated someone-who-knows from someone-who’s-guessing is gone. Access to information is now free and infinite; the scarce good is knowing whether any of it is true. The machine makes the output abundant, and the human’s remaining job is judgment about the output. Epistemics turns into the foreman’s job too.

Two forks from here. The centralized one: a handful of models become the de facto arbiters of what’s true, a level of epistemic concentration no newspaper baron ever approached. Hearst could shape opinion, not autocomplete it. The fragmented one: everyone’s AI gets tuned to their priors, and the shared factual baseline that mass media accidentally built over a century dissolves back into a thousand villages, each with its own oracle. Probably we get both at once. As before, the verification layer this era needs (provenance standards, models staking their reputation on calibration the way wire services staked theirs on accuracy) will arrive late, after real damage is done. Betting on the institutions eventually emerging is safe. Betting on them arriving on time never has been.

Where I land

The Industrial Revolution’s endpoint was good. Shorter hours, longer lives, the modern middle class. I’d take that trade for AI in a heartbeat. But the path ran through fifty years of stagnant wages, broken crafts, and dislocated communities, and the workforce that bore the costs wasn’t the one that collected the gains — their grandchildren were. It works out, eventually. How fast it works out for actual people depends on choices made during the transition: who owns the leverage, how the gains get shared, how quickly the new institutions get built. And last time, those weren’t gifts from the market. That gap closed only when workers organized and Parliament legislated. The weekend was a demand before it was a norm.

The foremen did fine in the factory era. The weavers didn’t. The strategic question, for individuals and companies and countries alike, is the same one it was in 1811: get to the supervising side of the machine, of its labor and of its claims, and don’t assume the market will move you there on its own.

Leadership isn't what most people think it is.

Scroll through LinkedIn or pick up the latest business bestseller, and you'll find endless praise for charismatic speakers, social media savants, and magnetic personalities. These qualities might look impressive on camera, but they mask what truly separates exceptional leaders from the rest:

Their ability to make consistently good decisions at the right velocity.

The Hidden Math of Leadership Impact

Throughout my career working with dozens of leaders across different organizations, I've observed that leadership effectiveness can be distilled into a simple equation:

Impact = Number of Decisions × Impact per Decision × Decision Accuracy

This isn't abstract theory—it's a practical framework that explains why some leaders create exponential value while others merely occupy space. Within this framework, four distinct decision-making archetypes emerge, each with dramatically different results.

The Four Decision-Making Archetypes

1. The Ghost

We've all worked with them—leaders perpetually "gathering more information" or "waiting for the right time." Their calendar might be packed with meetings, but actual decisions are virtually non-existent.

The Ghost's impact equation:

• Number of Decisions: Near zero

• Impact per Decision: Irrelevant

• Decision Accuracy: Unknown

Net Result: Organizational paralysis and missed opportunities

Real-world example: The executive who spent six months "evaluating options" for a critical product launch while competitors captured the market.

2. The Loose Cannon

Picture a leader who treats every day like it's a crisis. They make decisions rapidly and constantly—"Move fast and break things" isn't just their motto, it's their only setting.

The Loose Cannon’s impact equation:

• Number of Decisions: Astronomical

• Impact per Decision: Potentially very high

• Decision Accuracy: Abysmal

Net Result: Chaos with occasional transformative brilliance

Real-world example: The startup founder who launched seven different products in a single year, burning through investor cash but eventually stumbling upon a billion-dollar idea.

3. The Perfectionist

These leaders approach every decision with extreme caution. Their decisions are usually right, but by the time they make them, the opportunity has often passed. They prioritize precision while the market rewards speed.

The Perfectionist’s impact equation:

• Number of Decisions: Low

• Impact per Decision: High

• Decision Accuracy: Excellent

Net Result: Reliable but frustratingly slow progress

Real-world example: The meticulous product manager who consistently delivers flawless products—six months after the market window has closed.

4. The Unicorn

The rarest breed. These leaders combine the decisiveness of the Loose Cannon with the accuracy of the Perfectionist, all while maintaining high velocity. They're not perfect, but they deliver exceptional leadership impact.

The Unicorn’s impact equation:

• Number of Decisions: High

• Impact per Decision: High

• Decision Accuracy: Very Good

Net Result: Exponential growth

Real-world example: The healthcare leader who rapidly pivoted her organization to digital messaging early in a market shift, making a series of quick but well-calculated decisions that positioned the company ahead of competitors while maintaining quality of care.

The Growth Path: Evolving Your Decision-Making Style

Here's the good news: these archetypes aren't permanent. With self-awareness and deliberate practice, most leaders can level up their decision-making profile as well the decision-making profiles of their team.

The Ghost → The Perfectionist

Most organizational members start as Ghosts, often due to imposter syndrome or lack of clear role definition. When people don't understand their decision-making boundaries or fear being judged harshly for mistakes, they default to indecision. The key transformation comes through building decision-making confidence via small, low-stakes choices that provide quick feedback.

Transformation strategy: Create psychological safety and clear decision-making boundaries. Clarify exactly which decisions fall within their authority. Consider moving them into positions where decisions affect narrower outcomes with more defined parameters. Establish decision deadlines and celebrate when they're met, regardless of outcome.

The Perfectionist → The Unicorn

Perfectionists have the hardest but most promising growth path. Their accuracy is an asset, but their velocity needs work.

Transformation strategy: Introduce the concept of "reversible vs. irreversible" decisions. Help them see that 80% of business decisions are easily reversed, making "good enough now" far superior to "perfect later." Set clear decision deadlines and hold them accountable. Work with them to break big decisions into smaller, sequential choices. Actively celebrate faster decision-making, not just accurate outcomes.

The Loose Cannon → The Innovation Engine

Here's a controversial take: don't try to "fix" your Loose Cannons—channel them. These natural experimenters can become innovation powerhouses when properly directed.

Transformation strategy: Create bounded spaces for experimentation where their creativity thrives but organizational risk is contained. A Loose Cannon who lands one breakthrough in twenty attempts can deliver extraordinary value.

The Unicorn → The Super Unicorn

If you're lucky enough to work with a Unicorn, focus on helping them increase their impact per decision. They've mastered velocity and accuracy—now it's about placing them in roles with higher organizational leverage.

Transformation strategy: Move them into positions where their decisions affect broader outcomes. Create systems that free them from lower-impact decisions and expose them to the organization's most consequential choices. Consider structural changes that expand their sphere of influence across multiple teams or divisions.

The Bottom Line

Leadership isn't about charisma, charm, or looking good on quarterly earnings calls. It's about making good decisions consistently and quickly. Understanding these four archetypes can help you develop better leaders and build more effective organizations.

Over the last few months, it’s become clear to me that software built for AI agents is going to look entirely different from software built for humans. While that might sound obvious at first, I’ve come to this conclusion through hands-on experience building an AI agent called OllyChat, which handles site reliability engineering tasks. The architecture I use in OllyChat is similar to many AI agent applications, and it reveals some fundamental shifts in how these systems will be designed and used.

The Emerging AI Agent Architecture

Below is a simplified diagram of the AI agent architecture I’ve been working with and how it fits into a modern organization:

Here’s the high-level flow:

1. A human “manages” the AI agent through a natural interface—text or voice (via Slack, Teams, phone calls, etc.).

2. The AI agent creates plans, retrieves data, executes tools, evaluates progress, and iterates until it completes its goal.

3. Foundation models (OpenAI, Anthropic, Deepseek, etc.) provide the underlying language and reasoning capabilities.

4. Data, Tools, and Evaluators provide the AI agent with raw data, task automation, and performance feedback.

This setup is already very different from the software we build for humans, where we design user interfaces— screens, buttons, and dashboards—around human cognition and time constraints. AI agents aren’t held back by those same constraints.

Key Implications of This Architecture

User Interfaces for AI Resemble Human Interfaces

The “UI” of AI software is essentially text or voice, mirroring how humans manage each other today. The typical “AI Agent Manager” role will look a lot like managing a human subordinate—communicating tasks, checking progress, and giving feedback—but done via chat or calls.

Native Access to Company Data

AI agents need access to data at machine speed. Traditional dashboards and summaries exist because humans can’t process massive amounts of raw data in real time. AI agents, however, can. To provide the best responses and decisions, they’ll want direct access to the raw information. This means new types of real-time data pipelines that deliver high-fidelity data on demand.

Native Access to Tools

AI agents will also need ways to “do things” within the organization at machine speed. Rather than forcing an agent to wait for a human to click a button or parse a complicated dashboard, a robust AI-oriented interface will let the agent seamlessly orchestrate these tools. Imagine an interface that looks more like an API for a step-by-step debugging console than a web-based UI

Domain-Specific Customization Still Matters

Even with incredible foundation models, there’s a lot of custom code to write. Each industry, organization, or team will have specialized processes, constraints, and knowledge—requiring a tailored AI agent layer to get high-quality results. General-purpose models are starting points, not turnkey solutions.

AI Agents Improve Over Time

Much like a new hire, your AI agent will learn from every task and interaction, getting “sticky” because its performance compounds. It observes the outcomes of its actions, the feedback from humans, and the subtle signals in the data, steadily refining its approach.

Market Implications

Foundation Models Become Commodities

Switching foundation models will become almost frictionless. AI agents will dynamically pick the best model for a specific task—like your computer might switch between GPUs depending on which is best for the job.

Software Designed for Humans Will Be Repurposed or Superseded

Today’s software stack is heavily oriented around human-driven interfaces: dashboards, forms, notifications, and ways to get other humans to use the software. In an AI-first landscape, these tools will be controlled by agents that read and write data at scale, with humans only occasionally stepping in to provide high-level direction.

AI Replaces (Some) Human Labor Over Time

Unlike current tools that augment human productivity, AI agents replace tasks. This won’t happen all at once, but we’ll see an incremental process where tasks naturally migrate from humans to AI systems over the course of years.

Companies Shift Toward Managerial Labor

In the near future, the majority of a company’s workforce may function in “manager roles,” spending upwards of half their time directing AI agents. Coordinating tasks across AI-driven systems could become a primary job function, while AI agent managers also collaborate with each other.

Human-scale vs. Machine-scale

One of the biggest conceptual leaps here is recognizing that the constraints of human-scale processes—like waiting for daily or weekly reports, or manually debugging issues—don’t apply to AI agents. They operate at machine-scale, where data flows continuously and thousands of tasks can be executed in minutes, not days.

Human-scale

• We open a dashboard once a day, check metrics, then spend hours deciding and executing next steps.

• Information is heavily summarized so we can digest it.

• We rely on specialized tools that we can comfortably manipulate by hand.

Machine-scale

• An AI agent constantly ingests live data streams, adjusting its actions on the fly.

• Data is raw, granular, and massive—no summarization needed until it has to report to a human.

• Tools are accessed through direct APIs and complex orchestrations that can happen thousands of times in a single day.

The AI Agent Era

This shift to AI agent–driven software is poised to be as transformative as the move to personal computing or cloud infrastructure. We’re only at the beginning of seeing how AI agents will fundamentally change workflows, data pipelines, and tool interfaces. But one thing is certain: software for AI agents must be built for machine-scale operations rather than retrofitting the constraints of human-focused designs.

Subscribe now

Refer a friend

Understanding Bayesian Thinking Through Real-World Decisions

You launch a new feature, but engagement is lower than expected. Does that mean the feature is a failure? Or you get a strong candidate referral from an unexpected background—should you trust the signal? Making decisions in product and engineering leadership means constantly updating your assumptions as new data emerges. But how do you do this systematically?

Enter Bayesian thinking: a framework that helps you make better decisions by incorporating prior knowledge with new evidence.

Why Intuition Can Mislead Us

Imagine a routine medical screening. You get a positive result on a test that’s 90% accurate. Your instinct might tell you there’s a 90% chance you have the condition. But that’s incorrect. The actual probability is much lower because you haven’t factored in the base rate—the underlying prevalence of the condition in the population.

This same flawed reasoning appears in business and engineering decisions. We see an event, assume it must be true because of a strong signal, and fail to consider the underlying probabilities. Bayesian reasoning prevents us from making these mistakes by forcing us to weigh new data against the broader context.

The Core of Bayesian Thinking

At its heart, Bayesian thinking is about continuously refining our beliefs as new data emerges. Instead of treating each new piece of information as an isolated event, we integrate it with what we already know.

Mathematically, this is captured in Bayes’ Theorem:

Breaking this down further:

• P(A|B) = Probability of event A occurring given that event B has occurred (posterior probability).

• P(B|A) = Probability of event B occurring given that event A has occurred (likelihood).

• P(A) = Prior probability of event A occurring (before considering B).

• P(B) = Total probability of event B occurring (normalizing factor).

In simpler terms: what we know before + new evidence = a more refined, accurate belief.

Applying Bayesian Thinking to Product & Engineering Leadership

Let’s break down five critical areas where Bayesian reasoning improves decision-making in product and engineering leadership.

1. Feature Success: Early Adoption Data Can Be Misleading

Scenario: Your team launches a major feature. Adoption in the first two weeks is low. The CEO questions whether the effort was wasted.

Common Mistake: Assuming early adoption rates fully predict long-term feature success.

Bayesian Approach: Instead of treating the initial numbers as absolute, consider:

• Prior data: How have previous features performed over time?

• User behavior segmentation: Are power users engaging more?

• External factors: Seasonality, product awareness, and onboarding timelines.

Action: Rather than pulling the plug, run targeted experiments to understand why adoption is low and refine messaging or UX to improve engagement.

2. Hiring Decisions: Beyond the Resume Signal

Scenario: A candidate from a non-traditional background applies. They lack a CS degree from a top university, but their coding challenge results are excellent.

Common Mistake: Overweighting traditional pedigree signals while ignoring stronger real-world performance indicators.

Bayesian Approach:

• Prior probability: Past hires with similar backgrounds succeeded 40% of the time.

• New evidence: The candidate excelled in a rigorous technical test—something only 30% of all applicants achieve.

• Updated probability: Given this new evidence, their likelihood of success rises significantly above the base rate.

Action: Instead of dismissing them due to non-traditional credentials, use structured interviews and real-world problem-solving tasks to refine the assessment.

3. Churn Prediction: Don’t Overreact to Support Tickets

Scenario: A key customer submits multiple support tickets. Your team flags them as a churn risk.

Common Mistake: Assuming that an increase in support requests is a definitive sign of dissatisfaction.

Bayesian Approach: Consider:

• Base churn rate: Historically, only 40% of customers who submit high-ticket volumes churn.

• New evidence: This customer also actively engages in product advisory councils and attends training sessions—behaviors associated with an 80% retention rate.

Action: Instead of treating them as a churn risk, proactively strengthen the relationship, turning them into an advocate.

4. Engineering Trade-offs: Balancing Performance and Scalability

Scenario: Your team is debating whether to optimize for database query speed or invest in sharding for future scalability.

Common Mistake: Assuming that immediate performance improvements are always worth the trade-off.

Bayesian Approach:

• Prior evidence: Past scaling decisions show that 70% of early performance optimizations lead to tech debt when scaling.

• New evidence: Current user load is stable, but product growth projections indicate a 5x increase within a year.

• Updated probability: Investing in scalability now has a higher expected return by reducing future rework.

Action: Instead of hyper-optimizing for today’s load, design with future growth in mind—implementing a hybrid approach that maintains reasonable performance now while allowing for seamless scaling later.

See It in Action

To make Bayesian thinking more tangible, I built a small open-source tool that lets you explore these concepts hands-on. Whether you want to model feature adoption, hiring decisions, or engineering trade-offs, this tool helps you visualize how probabilities update with new data.

You can check it out and try it yourself here:

Bayesian Planner on Github

Final Thoughts: The Power of Updating Beliefs

Bayesian thinking isn’t about making perfect predictions—it’s about continuously refining your understanding as new evidence emerges. The best product and engineering leaders recognize that their first assumption is rarely their best one. They stay adaptable, weigh new data against past learnings, and make smarter decisions as a result.

In a field where complexity is the norm, Bayesian reasoning gives you a structured way to navigate uncertainty—one iteration at a time.

After two decades of leading teams across various organizations, I've noticed a fascinating pattern: leaders tend to operate in two distinct modes, often without realizing it. I call these "Problem-Solver Mode" and "Designer Mode."

Understanding when to employ each can be the difference between being a good leader and an exceptional one.

The Problem-Solver: Your Inner Firefighter

We all know this leader – they're the hero who jumps into action at the first sign of trouble. When crisis strikes, they're already three steps into the solution while others are still processing the problem. Most of us naturally gravitate toward this mode, and for good reason. It's satisfying, immediately rewarding, and makes us feel valuable.

In Problem-Solver Mode, leaders:

• Jump into action at the first sign of trouble

• Provide immediate solutions and quick fixes

• Take direct control of situations

• Focus on short-term wins and visible results

This mode serves us well in genuine emergencies. When there's a production outage or a customer crisis, being a skilled problem-solver is invaluable. The ability to think quickly and act decisively can save the day.

The Designer: Your Inner Architect

Then there's the Designer Mode – a more contemplative, systems-thinking approach that might feel less natural but often creates more lasting impact.

Instead of rushing to fix problems, leaders in Designer Mode:

• Step back to understand the broader system

• Create frameworks that prevent problems

• Empower teams to find their own solutions

• Focus on long-term sustainability over quick fixes

The Mirror Effect: Your Default Mode Shapes Your Entire Organization

Organizations inevitably mirror their leader's preferred operational mode. This reflection runs deep, creating distinct organizational cultures with their own strengths and challenges.

Problem-Solver Organizations

When leaders primarily operate in Problem-Solver Mode, their organizations typically develop:

• Swift and decisive responses to immediate challenges

• Clear chains of command for quick decision-making

• Strong execution capabilities for known problems

• Reliable crisis management systems

However, these organizations often struggle with long-term strategic planning and developing leadership at lower levels. The constant focus on immediate problems can create a reactive culture that never quite gets ahead of its challenges.

Designer Organizations

When leaders operate primarily in Designer Mode, their organizations tend to develop:

• Strong systems thinking and preventative approaches

• Distributed decision-making processes

• High levels of employee initiative and autonomy

• Focus on long-term solutions

The trade-off? These organizations might struggle with speed and clear direction in situations when immediate action is required.

The Default Mode Trap

One of the most common pitfalls I've observed in leadership is what I call the "Default Mode Trap." Like a dominant hand that we unconsciously favor, most leaders have a natural preference for either Problem-Solver or Designer Mode. This preference often stems from our past successes, personality type, or the environment where we first learned leadership.

The trap occurs when we rely too heavily on our preferred mode, even in situations where the other approach might serve us better. Let me share two real-world examples I've witnessed:

The Exhausted Problem-Solver

A Problem-Solver CEO spent three months personally resolving customer complaints, working late into the night to address each issue. While customers appreciated the attention, the underlying broken customer service system remained unchanged. The organization continued to generate the same problems, creating an exhausting cycle of firefighting.

The Paralyzed Designer

I watched a Designer-oriented leader spend months creating the perfect process for handling urgent reliability incidents. During this time, their team struggled with real-time issues that needed immediate attention, leading to frustrated employees and missed deadlines.

Finding Your Balance

The solution isn't to change your natural inclinations, but rather to develop awareness of your default mode and consciously practice operating in your non-preferred mode. Think of it like strengthening your non-dominant hand – it may never feel as natural, but having the capability serves you well in specific situations.

Different scenarios call for different approaches. A site reliability organization might benefit from Problem-Solver Mode's rapid response capabilities, while a new product team might thrive under Designer Mode's emphasis on systematic thinking and experimentation.

Mode-Switching Triggers: When to Shift Gears

Knowing when to switch modes is as important as knowing how.

Here are key situations that should trigger a conscious mode shift:

When Problem-Solvers Should Switch to Designer Mode

• When the same problem has occurred three times in a month

• During periods of team growth or reorganization

• When firefighting is consuming more than 20% of your time

When Designers Should Switch to Problem-Solver Mode

• During genuine emergencies (system outages, security breaches)

• When analysis paralysis is causing missed opportunities

• When immediate market conditions demand quick action

Moving Forward: Questions to Ask Yourself

1. Which mode do you naturally gravitate toward?

2. How has your preferred mode shaped your organization's culture?

3. What situations might benefit from you consciously switching modes?

4. How can you get comfortable in your non-preferred mode while maintaining your natural strengths?

Subscribe now

A CTO recently asked me a deceptively simple question:

“How do I know if I’m doing a good job?”

It’s a tough one. In a fast-growing startup—or any high-pressure environment—expectations can vary wildly. Trying to meet everyone’s demands is a surefire path to burnout. Instead, the key is to define success on your own terms and hold yourself accountable.

Three Dimensions of Success

I like to evaluate my performance across three core dimensions:

• Product Health

• Technical Health

• Organizational Health

1. Product Health

• Did we deliver what we promised?
  Delivering on commitments builds trust with customers and internal stakeholders.

• Is the customer experience improving?
  Enhancing the experience drives advocacy and organic growth.

• Is our competitive positioning improving?
  Are we keeping pace with—or outpacing—our competitors?

2. Technical Health

• Is the service reliable?
  Frequent downtime erodes trust and revenue.

• Are customers complaining about bugs or quality issues?
  Fewer complaints indicate you’re on the right track.

• Do we have a good developer experience?
  Happy developers are more productive and innovative.

3. Organizational Health

• Am I able to hire great people quickly enough?
  Fast-growing teams need a steady pipeline of top talent.

• Are great people happy and staying?
  Retention and engagement reflect a healthy culture.

• Are teams empowered to deliver and collaborate?
  Clear communication, ownership, and alignment are as important as the code itself.

Prioritizing Across These Dimensions

When everything feels urgent, it’s hard to decide what to tackle first. Here’s a simple roadmap:

1. Fix what’s causing you to lose deals

Analyze every deal lost in the last three months and map each loss to one of the three dimensions. You’ll likely uncover a few key issues that weaken your product compared to competitors. Addressing these will have an immediate impact on sales.

2. Fix what’s causing customer churn

Review every churned customer from the past few months and map their reasons for leaving to one of the three areas. Churn is often tied to reliability or quality issues—resolving these will improve retention.

3. Fix what’s limiting revenue growth

This usually ties back to competitive positioning and a strong product roadmap. Focus on delivering improvements quickly to stay ahead.

4. Prioritize organizational health last

I know I’ll get flak for putting organizational health last, but in my experience, fixing the above issues often leads to a stronger organization naturally. Most CTOs I’ve advised tend to overemphasize organizational health while neglecting the factors that drive revenue and prevent churn. No organization can truly thrive if it’s not growing.

Three Questions to Guide You

As you prepare for your next planning session or leadership meeting, ask yourself:

1. Product Health: How can I rapidly improve the customer experience relative to competitors?

2. Technical Health: How can I make it easier to deliver new innovations?

3. Organizational Health: How can I enhance the capabilities of my people and teams?

These questions will help you stay focused on what truly matters. As a CTO—or a CTO-like leader—your job is to balance these three dimensions so your company can innovate quickly, delight customers, and build a scalable, sustainable organization.

Subscribe now

For more than 15 years, I’ve set aside time each December for an annual review. This practice has helped me navigate career transitions, nurture key relationships, and stay focused on what truly matters in the year ahead. By conducting a thoughtful review before the rush of January, you can begin the new year with clarity and intention, rather than scrambling to piece together last-minute resolutions.

Late December is ideal for this exercise. Even if your calendar is busy, the natural end-of-year slowdown typically allows a few hours for a deep dive into your accomplishments, challenges, and future goals. Setting aside time now ensures you’ll hit the ground running when everyone returns to work at full speed in January.

Over the years, I’ve experimented with different approaches to my year-end review. The following questions have proven to be the most effective and insightful. To get the most out of them, I recommend dedicating at least four uninterrupted hours, ideally in one sitting.

The Seven Key Questions

1. What did I accomplish? (45 minutes)

Give yourself enough time to reflect broadly.

• 15-20 minutes: List all the wins—big or small—that mattered this year.

• Remainder: Group these achievements into themes (personal, professional, community, health) and consider why they mattered. What do they reveal about your strengths and values?

2. What were my biggest disappointments? (30 minutes)

Acknowledge what fell short without fixating on the negative.

• 15 minutes: Identify moments or outcomes that disappointed you.

• Remainder: Focus on the lessons learned. Were there patterns that kept appearing? Recognizing these will help you avoid similar pitfalls in the future.

3. What made me happy? (25 minutes)

Celebrate the sources of joy.

• Pinpoint the projects, relationships, activities, or habits that brought genuine happiness.

• Look for patterns and consider how to integrate more of these elements into the coming year.

4. What made me unhappy? (25 minutes)

Likewise, identify the triggers of negativity.

• Reflect on the causes of stress, frustration, or sadness.

• Consider strategies to mitigate or eliminate these factors. Can you delegate certain tasks, restructure commitments, or change routines?

5. What did I learn? (35 minutes)

Pull insights from the entire year.

• Distill lessons from your accomplishments, disappointments, and day-to-day challenges.

• These lessons will guide your decisions and approach next year, helping you build on what worked and sidestep what didn’t.

6. What do I want to achieve next year? (35 minutes)

Set your sights forward.

• Brainstorm broadly at first, listing anything you hope to accomplish, improve, or try.

• Then refine by grouping similar goals together to find overarching themes.

7. What are my top 3-5 goals for next year? (45 minutes)

Narrow it down.

• From your brainstorming in the previous step, choose the 3-5 goals that resonate most strongly with your priorities and values.

• Consider what habits or support you’ll need to make them happen.

How to Get the Most Out of Your Review

• Protect your time: Schedule a four-hour block—treat it like any high-value meeting. While it might feel strange to dedicate work time to self-reflection, this investment pays off by bringing sharper focus and more deliberate action into your professional life.

• Change your environment: Choose a quiet spot away from your usual workspace. A new setting can spark fresh insights and help you see the past year from a different perspective. A coffee shop, a library, or even a quiet corner in your home can help break habitual thought patterns and unlock new ideas.

This December review process isn’t about piling up regrets or creating unrealistic resolutions. It’s about understanding where you’ve been, what you’ve learned, and where you want to go. This December, take the time to honor your journey, celebrate your growth, and step into the new year with purpose and confidence

Software has always been about solving problems, but how it delivers value has evolved over time. From traditional software to SaaS and now AI Agents, each phase represents a shift in responsibility—moving progressively from the customer to the software.

At every step, customers do less while software takes on more. AI Agents represent the ultimate step in this evolution, fully automating outcomes and leaving almost nothing for the customer to manage.

1980–2000: Traditional Software — The Customer Does the Heavy Lifting

In the early days, software products were built to automate specific tasks but left the customer responsible for making them work.

What Customers Had to Manage

• Setup and Customization: Implementations could take months to configure.

• Operational Management: Customers were on the hook for updates, maintenance, and infrastructure.

• Steep Learning Curve: Complex systems demanded significant training and upfront investment.

These solutions were rigid, expensive, and carried high risks, making them accessible only to larger businesses.

2000–2020: SaaS — Vendors Take on More Responsibility

SaaS changed everything by shifting much of the operational burden to vendors. Software moved to the cloud, making it easier to deploy, scale, and maintain.

How SaaS Reduced the Load

• Simplified Onboarding: Faster setups, intuitive interfaces, and minimal configuration.

• Managed Infrastructure: Vendors handled uptime, updates, and scaling.

• Flexible Pricing: Subscription models allowed businesses to pay as they grew.

However, the customer still had work to do:

• Decision-Making: They had to figure out how to use the software effectively.

• Task Execution: SaaS provided tools, but results were still the customer’s responsibility.

SaaS made software more accessible, but it didn’t eliminate the need for human involvement or fully align incentives between businesses and software providers.

2020–2040: AI Agents - Full Outcomes from Software

AI Agents are the next big leap. Unlike traditional software or SaaS, these systems don’t just provide tools for users—they autonomously deliver results.

Execution, optimization, and outcomes are all handled by the software itself.

How AI Agents Take Ownership

• Outcome Delivery: No features or dashboards—just results.

• Zero Setup: No onboarding, no training, no configuration. Deployment happens instantly.

• Self-Management: AI Agents run and optimize themselves with little to no oversight.

• Continuous Improvement: They learn from their environment and get better over time.

For customers, the software becomes almost invisible. All that matters is the outcomes.

Why AI Agents Will Disrupt SaaS

Over the next 20 years, the AI Agent model will challenge and likely replace SaaS in many areas.

Here’s what sets AI Agents apart:

1. Outcome-Based Pricing: Customers pay for results, like a percentage of revenue generated or costs saved. This aligns costs directly with value.

2. Human-Like Interfaces: AI Agents interact like humans—via text, speech, or browser actions—removing the need for complex integrations or APIs.

3. Dynamic Decision-Making: AI Agents rely on real-time conversation streams (emails, messages, calls) rather than rigid relational data structures.

4. Massive Cost Savings: By automating repetitive tasks, AI Agents will disrupt labor-intensive processes, freeing up resources for higher-value work.

5. Winner-Takes-All Market: The best AI Agents will dominate, benefiting from superior data, rapid iteration, and self-reinforcing growth cycles.

How AI Agent Companies Will Win

Building AI Agent businesses won’t be easy. It requires deep expertise beyond just training AI models. Cloud providers will likely dominate the foundational AI layer, taking a cut of the revenue from businesses built on their platforms. The real value lies in domain-specific execution.

Key Success Factors

• Access to Data: The more relevant data, the better the outcomes.

• Seamless Integration: Plug into existing workflows without disruption.

• Domain Expertise: Industry-specific prompts and validation mechanisms.

• Human Oversight: To manage and refine AI at scale.

• Scale: Larger datasets drive better AI performance and outcomes.

Similar to SaaS, AI Agent businesses will target specific niches but focus on automating repetitive workflows end-to-end.

Examples of AI Agent Niches

• Customer service automation

• Sales lead qualification

• Accounts payable processing

• Contract review and compliance

• Vendor onboarding

• Marketing asset personalization

How AI Agent Businesses Will Emerge

Delivering AI Agents businesses will be challenging and require significant expertise beyond foundational AI models, which are rapidly commoditizing. Cloud vendors are likely to dominate the foundational AI layer, taking a 10% revenue cut from businesses leveraging their models.

Vertical Specialization: Similar to SaaS, AI Agent businesses will target specific niches but focus on automating repetitive workflows end-to-end.

Potential Use Cases

• Customer Support and Service Desk Automation

• Sales Lead Qualification

• Accounts Payable Processing

• Contract Review and Compliance Checks

• Expense Report Reconciliation

• Vendor Onboarding

• Marketing Asset Personalization

Keys to Winning in These Niches

• Access to all relevant data.

• Seamless integration with existing business applications.

• Domain-specific prompting and validation.

• Human oversight to manage AI at scale.

• Scale within the domain—larger datasets yield better outcomes.

The Future: A Focus on Creativity and Connection

AI Agents will redefine how businesses operate, freeing them from mundane tasks and allowing them to focus on creativity, innovation, and human connection. The promise of software will finally be fulfilled—not as a tool we have to master, but as a partner that helps us reach our full potential.

Share

Let’s kick this off with a confession: I’ve learned more about leadership from bad leaders than good ones. Painful, yes. Enlightening? Absolutely. Turns out, surviving under a questionable leader can teach you a lot about what not to do—and those lessons tend to stick around like glitter after a craft project.

Here’s the wild part: even the bad leaders usually have an uncanny understanding of people. They just wield it poorly. The next time a leader does something that makes you consider banging your head against your desk, try this thought experiment:

Question to ask yourself: What do bad leaders understand about people that I might be missing?

Instead of spiraling into frustration, you start unraveling the why behind their actions. Questions like this help you figure out the people puzzle and become a better leader yourself.

Leadership isn’t about being the smartest person in the room—it’s about being the one who understands the people in the room.

To help lead with a human approach, I’ve crafted a simple framework to break down what people really need from their leaders.

The 3 Things People Need from Their Leaders

When you strip leadership down to its essence, it’s about addressing three universal needs: growth, connection, and impact.

1. The Need to Grow

No one wants to feel like they’re stuck in a never-ending loop of mediocrity. People need to feel challenged, like they’re learning, stretching, and improving. The twist? Growth doesn’t always look like a straight path to success. I once had a team member who hesitated to take on a challenging project. They were convinced they’d crash and burn. I gave them a nudge (okay, more of a shove) and promised I’d back them up. Fast forward a few months: yes, there were hiccups—“we’re doomed” moments included—but they came out the other side sharper, stronger, and more confident.

Here’s the question: Are you giving your team opportunities to stretch and stumble, or just keeping them busy?

Actionable tip: Hand out stretch assignments generously, but don’t vanish after the handoff. Support your team, celebrate their efforts, and treat setbacks as fuel for growth, not failure.

2. The Need to Connect

Let’s face it: great teams don’t just work together—they click. Trust, respect, and a sense of belonging aren’t just “nice to have” extras; they’re the glue that holds a team together. Without them? Your team feels more like a middle school dance—awkward, stiff, and full of side-eyes.

I’ve been on teams where trust was so strong you could practically read each other’s minds. I’ve also been on teams where meetings felt like a series of polite-but-pointless chess moves. Guess which one delivered better results?

Here’s your gut check: Does your team feel like a team—or just a group of individuals forced to share a Slack channel?

Actionable tip: Build trust like it’s part of your daily workout—not something you save for team offsites. Listen (actually listen), foster honest conversations, and create moments for people to connect with each other, not just with you.

3. The Need for Impact

People don’t just want to churn through tasks—they want to feel like they’re contributing to something that will matter years down the road. It’s the difference between grinding through the mundane and being fired up by the vision of what they’re helping to build.

I’ll never forget a leader who constantly told us, “Just focus on today’s JIRA tickets,” with zero context for why our work mattered. It was soul-crushing. In contrast, I once worked with a leader who connected every task—even the boring ones—to the future impact we were creating. Suddenly, even the grind felt worthwhile.

Ask yourself: Am I helping my team see how today’s work contributes to building something impactful for the future?

Actionable tip: Show your team how their work ties into a larger vision with future impact. Even if it seems obvious to you, spell it out—it makes all the difference.

Mastering the 3 Needs

Here’s the kicker about the “3 needs” framework: it’s not just a tool for understanding your team—it’s a mirror for understanding yourself.

Using the Framework to Understand Yourself

Most leaders are great in one of these areas, while others… not so much. Maybe you’re a visionary who can paint an inspiring picture of the future but struggles to foster collaboration or help individuals grow. Or maybe you’re a connection maestro who builds trust like a pro but forgets to zoom out and articulate the big picture. Whatever your leadership superpower is, owning it is step one.

Step two? Reinforcing the areas where you’re less naturally gifted. This isn’t about becoming perfect in all three needs—it’s about being self-aware enough to know when you need to step up or lean on others to fill the gaps.

Using the Framework to Improve Your Team

Now, let’s talk about your team. Chances are, you’re nailing one or two of these needs while the third quietly collects dust in the corner. For example, in the startup world, teams often are great at addressing impact. They’re laser-focused on big, bold missions—launching products, scaling fast, and disrupting industries. But growth? Connection? Those can easily get sidelined.

The framework can help you pinpoint the exact areas where your team might be wobbling, so you can get to work shoring things up.

What’s Next: Ask Your Team

Think you’ve got this nailed? Don’t guess—ask.

Here’s a simple exercise: send out a quick, three-question survey to your team. Ask them to rate, on a scale of 1-7, how much they agree with these statements:

1. Growth: I am working on a project or goal that is causing me to grow professionally.

2. Connection: I have a high level of trust and psychological safety with every person on my team.

3. Impact: I am working on something that will have a large impact for the company.

The results might surprise you. They’ll also give you a clear roadmap for where to focus your leadership energy.