The Skill That Compounds Like Interest
Leaders across dozens of industries kept naming the same human capability as the one that gets more valuable every year. It's not what most people are investing in.
A surgeon in Central Texas watches a robot place a knee implant within a millimeter of perfection. The machine does it better than his hands ever could.
He’s not worried.
“Robotic systems guide implant placement within a millimeter,” says Christopher Jimenez MD, Founder of Bone Drs. “What they can’t do is weigh an 80-year-old’s life expectancy against implant durability and her wish to travel to see grandkids – then make a call and own it.”
That gap between what a machine can execute and what a human must decide is where the next era of professional value lives. But the gap keeps shifting, and most people are looking for it in the wrong place.
I gathered perspectives from founders, CEOs, and operators across dozens of industries on one question: which human skills hold their value over time, and which don’t? Not the corporate training version of this conversation. The version shaped by decades of watching real skills rise and fall in the field.
What came back was a pattern. And the pattern kept pointing to the same uncomfortable conclusion.
The skills most people spent their careers building are the ones depreciating fastest.
The Skill That Keeps Compounding
If I had to name the single capability that came up most across every response, it would be this: the ability to see what isn’t being said.
Not empathy in the abstract, feel-good sense. Something more specific.
The ability to read a room, a situation, a set of data, or a person – and understand what’s actually happening underneath the surface presentation.
Joseph Agresta, President of Benzel-Busch, runs a luxury automotive group that has survived four generations. His great-grandfather started as a blacksmith shoeing horses. The family business endured because of one skill that transferred across every era.
“A customer tells me they need a G-Wagon for ‘safety,’ but I’m listening for whether they mean physical safety, financial security, or social status – because that determines whether they’ll be happy in six months or become a service headache.”
That’s a diagnostic skill that sharpens with every interaction.
Agresta notes that his sales data shows customers now research everything online and then come in specifically to test one thing – “whether we’ll honor the promise after the sale – and they decide that in about 90 seconds of interaction.”
Justin Belmont, Founder and CEO of Prose, sees the same pattern in media strategy. “The ability to sit in five different client conversations and realize they’re all actually describing the same underlying problem. AI can generate options, but it still takes a human to say, this is the real constraint, this is the leverage point.”
The word that kept showing up across industries – roofing, law, medicine, fitness, finance – was pattern recognition.
Not the data science kind. The human kind. The kind you can only build by being in the room, on the roof, across the table, thousands of times.
Dan Keiser, Principal Architect at Keiser Design Group, has been designing homes for nearly 30 years. He describes the skill as “listening for what clients aren’t saying.”
When someone asks for a four-bedroom house, they’re not mentioning that their kid has sensory issues and needs quiet zones, or that aging parents might move in.
“I’ve learned to ask about Sunday mornings, arguments, and what makes them feel trapped in their current space. That excavation work can’t be automated because it requires reading body language, hearing hesitation, and knowing when to push back on their own stated ‘requirements.’”
Tim Cakir, Chief AI Officer and Founder of AI Operator, gave a concrete example. His team was helping a financial services firm implement AI across their operations.
The AI recommended fully automating client onboarding – the data supported it, the efficiency gains were clear. But someone on the team with 20 years of industry experience said: “Our high-value clients chose us because of the personal touch during onboarding. Automating that would save us money and lose us our best customers.”
Cakir calls this contextual judgment. “Knowing whether that strategy makes sense for this specific company, with this specific culture, at this specific moment.”
Marzena Beltek, General Manager of Doma Shipping & Travel, has been managing international logistics for over 30 years and sees this skill play out in situations most people would never consider.
She described a family relocating from Chicago to Warsaw whose container got flagged by customs over a vintage motorcycle. “The internet had already told them what forms they needed. What they couldn’t find online was someone who understood that Polish customs cares less about the bike’s value and more about proving it’s not for resale, and that the way you present that case matters as much as the paperwork itself.”
She got it cleared in two days instead of two weeks. She had the same information everyone else did. The difference was judgment about how institutions actually behave.
Joseph Depena, Owner of VP Fitness, has spent 13 years training people in Providence, and he described the same capability in a completely different domain.
“When someone’s form breaks down on rep 8, that’s not a technique issue – it’s a window into how they handle discomfort everywhere else. I can now predict who’ll ghost after two weeks versus who’ll hit a two-year streak based on how they respond when a workout gets hard.”
That prediction comes from watching thousands of humans under physical stress and paying attention to what their behavior actually means.
If you’ve watched Star Trek, you know the difference between the Enterprise computer and Captain Picard. The computer can pull up any fact, run any simulation, project any outcome. But when the situation is ambiguous and the data conflicts, someone still has to sit in the chair and decide.
That’s the skill.
My Takeaway: The pattern recognition described here is compressed experience. Thousands of reps of seeing the gap between what people present and what’s actually going on. This is the skill that compounds like interest – slow at first, then unmatchable. If you’re early in your career, the single best investment you can make is getting into rooms where the stakes are high and the information is messy. Every one of those reps counts.
What Died – And What Replaced It
The second pattern was just as consistent. Across every industry represented – law, medicine, architecture, logistics, software, fitness, real estate – the same category of skill lost value: knowing things.
Memorizing statutes. Reciting product specs. Holding inventory details in your head. Running spreadsheets manually. Being the person in the room who could retrieve information fastest.
Herman Martinez, Founder of The Martinez Law Firm, has practiced criminal defense for over 25 years, including time as Chief Prosecutor for Harris County.
“Twenty years ago, knowing penalty group classifications or sentencing guidelines by heart made you valuable. Now anyone can search that in seconds.”
What replaced it surprised him. “Knowing how a former colleague-turned-prosecutor thinks when they’re deciding whether to offer a plea deal, or which judge will actually depart from sentencing guidelines based on individual circumstances. That’s relationship intelligence and systems thinking, not information retrieval.”
Griffin Sher, Partner at Sher & Volk, P.A., a maritime law firm, experienced the same shift. “I graduated Cum Laude from Tulane’s maritime program partly by memorizing the Jones Act inside-out. Now ChatGPT spits that out in seconds.”
What Sher didn’t expect was where the value migrated.
“First client interviews became more valuable. Injured seamen can Google the Jones Act now, so they don’t need me to explain it. They hire me in the first 15 minutes when they realize I understand why they didn’t report the injury immediately – because reporting gets you blackballed from future ships.”
Cultural fluency replaced legal fluency. Understanding people replaced memorizing precedent.
Kuldeep Kundal, Founder and CEO of CISIN, has spent two decades in software development and saw the same pattern play out in tech. “Having a deep understanding of the specific syntax for one programming language was enough to establish a successful career.”
That’s now commodity knowledge. What replaced it is what he calls “architectural orchestration” – the understanding of how different systems, AI agents, and data flows intertwine.
Sahil Agrawal, Founder and Head of Marketing at Qubit Capital, offered the cleanest test I encountered for whether a skill is appreciating or depreciating: “If the skill is about producing a first version of something, it is probably depreciating. If it is about knowing which version is right and getting other people to act on it, it is compounding.”
Agrawal has seen this play out on his own team. “One manager was submitting AI-generated work that looked polished on the surface but pushed all the real thinking upstream. The editing, the judgment calls, the knowing what to cut. That was the actual skill, and he was outsourcing it without realizing it.”
Niclas Schlopsna, Managing Partner at Spectup, saw the same shift from the venture advisory side. “Basic information retrieval and routine content production used to signal competence, but AI systems now perform them faster and more consistently.”
What replaced them is the ability to frame the right question before generating the answer.
He shared a moment that crystallized it: “In one fundraising advisory session, two models suggested opposite growth strategies, and the real value came from interpreting which scenario aligned better with the company’s operational psychology rather than just statistics.”
The distinction that kept emerging landed differently than the usual “soft skills vs. hard skills” framing. It’s about where in the decision chain your skill lives.
Upstream – framing the problem, choosing the direction, making the judgment call – or downstream – executing, retrieving, processing.
Downstream is getting automated. Upstream is getting more expensive.
My Takeaway: The “knowing things” era is over. What replaced it is specific. Relationship intelligence – knowing how a judge thinks, why a client is actually anxious, what a founder can’t articulate about their company.
Systems thinking – understanding how parts interact under stress. Problem framing – the ability to look at a failed business and see that the real issue is an unmet emotional need nobody identified.
If your professional value is built on information you can retrieve, you’re sitting on a depreciating asset. The migration path is toward interpretation, judgment, and the ability to act on incomplete information.
The Automation Line Moved – But Not Where Anyone Expected
This was the most surprising pattern. Nearly everyone expected the same things to get automated: data entry, scheduling, routine processing. That happened.
Nobody expected where the line moved next.
Ash Sobhe, CEO of R6S, put it directly: “AI is better than I expected at tasks I thought required creativity, and worse than I expected at tasks I thought were purely mechanical. It writes decent marketing copy but struggles to integrate a payment system with an undocumented API.”
The old categories – creative vs. mechanical, soft vs. hard, human vs. technical – turned out to be the wrong map entirely.
“The real divide is well-defined versus ambiguous,” Sobhe notes. “AI handles well-defined problems beautifully. Humans still own ambiguity.”
Belmont saw the same thing from the content strategy side. “I used to think creativity was mostly safe and data crunching was at risk. Turns out, a lot of surface-level creativity is pretty automatable, and deep analytical thinking about messy, real-world tradeoffs is harder than we thought.”
Jake Byrne, President of America Roofing Company, offered the most visceral example of what can’t be automated. He’s spent 20-plus years running roofing projects across Arizona.
“AI can analyze a photo. It can’t feel the micro-bounce of delaminating plywood or notice that the flashing profile doesn’t match the original install, meaning someone did a quick fix that’s now creating a new failure point.”
Byrne describes his core skill as reading failure chains – not just what broke, but why it broke there, in that sequence, and what’s about to fail next.
“A $40 pipe boot collar cracks, lets in water, rots a $200 section of decking, which sags and breaks two $8 tiles, which lets in more water, which spreads to the drywall, ruins insulation, grows mold, and you’re at $15,000.”
That cascade pattern shows up in every industry. Supply chains, project schedules, customer service workflows. The skill is seeing the chain early and interrupting it.
Pleasant Lewis, Owner of Fitness CF, has run gyms in Florida for 40 years. He shared an automation surprise that captures the new line perfectly.
His AI-powered feedback system can categorize thousands of member responses instantly. But it still flags the wrong things.
“Last month our system highlighted ‘parking complaints’ as top priority – lots of mentions. But when I read them, they were from the same three people. Meanwhile two quiet comments about childcare wait times represented 15 families about to quit.”
The machine saw volume. Lewis saw meaning. The parking complaints were three loud voices. The childcare comments were a quiet signal that real revenue was walking out the door.
That’s the new frontier. The machine counts. The human weighs.
Debra Vanderhoff, Founder of MicroLumix, has built companies across biotech, finance, and operations. She saw the same shift from the other direction – where automation revealed where human value actually lives.
“Fifteen years ago, I was valued for systematizing sales workflows and operational efficiency. Today, any decent software does that automatically.”
What replaced it was something harder to name: “Knowing which problems are worth solving before they’re obvious.”
In 2019, her team started building automated disinfection systems months before COVID hit. They didn’t predict a pandemic. They noticed 54,000 people were already dying daily from preventable infectious disease and nobody was protecting high-volume touchpoints between manual cleanings.
That’s pattern recognition in unmet need. And you can’t prompt your way into it.
My Takeaway: The real automation divide is defined vs. ambiguous. This reframes the whole conversation about future-proofing your career. Instead of asking “Is my job creative enough to survive?”, ask “Does my work require operating in ambiguity that changes depending on context?”
If you’re doing well-defined creative work – writing formulaic copy, generating standard designs, producing templated content – you’re more exposed than you think. If you’re doing something that requires reading messy situations where the answer depends on context that shifts – you’re more protected than you feel.
The machine counts. The human weighs. But you still need something worth looking at before you can make the call.
Databox pulls metrics from the tools you already use into dashboards and scheduled reports – so you can spend less time stitching together spreadsheets and more time doing the kind of judgment work that actually compounds. Start a free trial here.
What to Tell a 25-Year-Old
Almost every respondent answered the question about what they’d tell a younger person. The answers clustered around five capabilities.
Raj Baruah, Co-Founder of VoiceAIWrapper, captured the first one. “Stop optimizing for being the person with the best answer. Practice being the person who notices when everyone is answering the wrong question.”
He describes a team that spent three weeks debating between two vendors until someone asked whether they needed an external vendor at all. “That question hadn’t been on the table. Within two days we’d identified an internal solution that was cheaper, faster to implement, and already had organizational buy-in.”
Kundal calls this “problem reframing.” “Most people are trained to provide a solution for a defined problem. In the next 20 years, AI will provide solutions for almost every clearly defined problem. The true value in people will be to look at a failed business and see that the cause of the failure is not with particularly weak software, but is due to an undefined process and/or an unmet emotional need of the customer.”
Kelly Cassaro, Chief of Learning at Generation, pointed to leading through uncertainty. “Tools will change and tasks will be automated, but the need for humans who can create clarity, regulate emotion, and help others move forward when the path is unclear will only grow.”
Sobhe advocated for cross-domain translation. “Work in multiple industries. Learn how different types of businesses think about problems. The person who can sit with a healthcare founder in the morning and a real estate investor in the afternoon, and transfer insights between those conversations, will be irreplaceable for decades.”
Joshua McAfee, Founder and CEO of McAfee Institute, put it in operational terms from his experience building training systems for Amazon and the U.S. military.
“The analysts I certify who get promoted fastest aren’t the best at OSINT or geopolitical analysis alone – they’re the ones who can connect a DNS enumeration finding to a sanctions evasion pattern to a corporate shell game.”
Tim Johnson, CEO of BIZROK, offered the underlying principle behind all of them. “Learn to diagnose systems, not just fix tasks. My dad was a solopreneur who could fix any immediate problem but couldn’t figure out why he was always trapped in his business. He had task skills, not system skills.”
The five capabilities, distilled: ask better questions, lead through ambiguity, translate across domains, connect unrelated patterns, and see the system behind the symptom.
None of them can be learned from a course. They sharpen only with use – and they require contact with messy reality to develop at all.
My Takeaway: If I were mapping these five capabilities to a single career strategy, it would be this: spend your twenties collecting diverse experiences across different domains, industries, and types of problems.
The goal is to build a cross-referencing system in your head.
Every industry you touch gives you a new pattern library. Every messy situation you navigate adds a case file to your judgment database.
The people who become irreplaceable in their forties didn’t specialize early. They synthesized broadly and then went deep when they found where their pattern library was most valuable.
In The Hitchhiker’s Guide to the Galaxy, there’s a book that contains the collected knowledge of an entire civilization. Every weird planet, every strange species, every improbable situation – all indexed and cross-referenced. Your career can work the same way. The more domains you collect, the more connections you can make that nobody else sees.
Final Thoughts
The through-line across every perspective – from surgeons to roofers, from maritime lawyers to gym owners, from AI founders to painting contractors – was this: the skills that last are the ones that require you to have been there.
Actually been there. Watched the deal fall apart. Felt the plywood give. Noticed the witness story was too perfect. Saw the client’s shoulders drop when you named the thing they couldn’t articulate.
AI is getting better at everything that can be described clearly. The parts that depend on having lived through the consequences of your own past decisions remain stubbornly human.
The skills that depreciate are the ones that can be flattened into a prompt. The skills that appreciate are the ones that require the full weight of a human life behind them.
That’s not a comforting framework for anyone looking for a quick answer. But it is, I think, an honest one.
The future will keep asking what you know. The answer that matters is what you’ve learned to see.
The skill that compounds fastest is judgment in ambiguous situations – the ability to weigh what matters when the data is incomplete, the incentives are messy, and the stakes are high.
This article contains affiliate links.