Between employees giving discretionary effort and those doing the minimum. That effort cannot be mandated. Only earned.
5
Things AI cannot replicate
Empathy. Presence. Judgment. Creativity. Hope. MIT Sloan calls these EPOCH. Every one is human. Every one holds your org together when tools fail.
N/A
AI cannot distinguish good ideas from mediocre ones
Not a low score. Not a gap to close. A capability that simply does not exist in the machine.
35%
Productivity gain
When humans guide AI output vs. far less when automation replaces human oversight. Stanford HAI, 2024. The human is not overhead. The human is the multiplier.
Tap or hover each to explore
Priceless.
Every year, organizations spend more on tools and less on the people who use them — and then act surprised when nothing compounds. This isn't a theory. It's a pattern. And it has receipts.
Scroll to explore
The Foundation
People aren't the bug. They're the feature. That is not a debate. That is not a perspective. That is the wall behind the wall — the foundation that makes every other argument possible.
If you don't agree that people have inherent value that isn't contingent on their output, none of the rest of this computes. And that's useful to know quickly.
Before you read anything else
Not all investments compound the same.
Every line on this chart was approved without requiring proof of return — except one. The rope didn't get that courtesy. Look at where the rope went anyway.
"The rope is not a metaphor. It is the argument. A person is the only asset in your budget that can exceed the model used to measure them. And you spent $468 on it last year."
Average large-company training spend per employee · Training Magazine, 2025
Section One
Complexity went up. Investment went down. Nobody noticed.
Work has gotten harder every decade. Not because people got worse at it. Because the systems they navigate multiplied, the tools they master expanded, and the expectations on them kept climbing — while the investment in preparing them went in the opposite direction.
This is not a feeling. It is a documented, measurable divergence. And the gap between what we ask of people and what we give them to work with has been widening for twenty years.
1950s – 1980s
Work was bounded. Investment was human.
A job required mastery of a defined set of tools and processes. There was no SaaS stack. No digital transformation budget. Just: teach the person how to do the work. Organizations trained people because that was the only way knowledge moved.
1990s – 2007
Technology arrived. Training peaked — briefly.
As software entered the workplace, organizations briefly understood they needed to invest in adoption. Training hours climbed. The calendar of optional development classes appeared on intranets. Managers sent people.
Peak: 37 training hours per employee in 2007 — ATD
2001 – 2009
The first collapse.
Employer-paid training fell 28% in eight years. Almost no industry was immune. Large companies led the decline. The calendar disappeared. Managers stopped sending people — not because the value changed, but because the metrics made the cost visible and the return invisible.
28% decline in employer-paid training — Waddoups, 2016
2015 – 2023
The SaaS explosion. Complexity multiplied.
The average organization went from a handful of tools to hundreds. By 2023, the average company used 371 SaaS applications. The job got exponentially more complex. Training did not follow.
371 SaaS apps per org in 2023, up from ~8 in 2015 — BetterCloud/Productiv
2020 – 2024
The freefall. 61% in four years.
Formal learning hours fell from 35 in 2020 to 13.7 in 2024. At the same moment AI tools arrived and AI spend exploded. The work got harder. The tools got more powerful. The investment in the people using them reached its lowest point in a generation.
35 hrs (2020) → 13.7 hrs (2024): –61% — ATD 2025
"Cognitive overload costs organizations $322 billion annually in lost productivity. The cause isn't lazy employees or poor management — it's a fundamental misunderstanding of how human minds process complexity."
Dr. Laura Weis — Enterprise Technology Leadership Summit, 2024
The Part That Doesn't Show Up in the Data
That person exists in every organization.
Every data point in the previous section represents a person. Not a headcount. Not a resource. A person who came to work, did the job they were given, and was never given what they needed to do it better — not because they didn't want it, but because nobody decided they were worth the investment.
You have worked next to them. You may be them.
The Excel Spreadsheet Guy
Nobody asked him to build it. It wasn't in his job description. But he saw that every week someone was manually calculating the same numbers for three hours — and he knew how to make it take thirty seconds. So he built the tool. Entire departments still run on it. He was never on any high-potential list. He was just paying attention, and he had enough knowledge to do something about what he saw.
The CSR with the Lost Prescription
She was a customer service rep at a healthcare organization. A patient called, confused and scared, because their prescription wasn't processing. She had been trained — really trained — on how the system worked end to end. She knew three pathways to the same answer. She found the one that worked. The patient got their medication. The call took four minutes. Nobody at the top ever heard about it. It just happened, because someone had been given what they needed.
The Woman at the Hair Salon
She wasn't even an employee. She was a client sitting in the next chair, listening while her hairdresser learned how to use a new tool. When she was done, she stood up and said: "I don't want to leave without that website." Nobody had decided she was the target audience. But the room had been built wide enough to include her. That's what happens when you stop deciding in advance who is worth reaching.
These are not edge cases. They are what happens every day in organizations that invest in their people. The question is not whether these people exist in your organization. They do. The question is whether you gave them what they needed — or whether you decided they weren't on the list.
Pattern observed across 25 years · regulated healthcare · financial services · fintech
What investment actually looks like — one person, one arc
Early 2000s
Learned Excel. Took the class.
Then went home and kept going. Figured out VLOOKUP at 11pm from Microsoft's help documentation and sheer stubbornness. VLOOKUP had been in Excel since 1985 — third most used function after SUM and AVERAGE — but most people were still doing manually what it could do in seconds. He wasn't most people. He saved someone three hours the next morning. Nobody noticed. He noticed.
Mid 2000s
Discovered PivotTables.
This was the door behind the door. VLOOKUP moved data. PivotTables changed how he saw it. He stopped looking at rows and columns and started seeing questions. What if I slice it by region? By month? By person? By product? The data wasn't just organized anymore — it was answering things nobody had thought to ask. He started asking things nobody had thought to ask.
Late 2000s
Built the macro nobody asked for.
He saw that every week someone was manually calculating the same numbers for three hours. He knew how to make it take thirty seconds. So he built the tool. Not because anyone asked. Not because it was in his job description. Because he could see the problem and he had enough knowledge to solve it. Entire departments still run on that macro. He was still not on any high-potential list.
2015 Onwards
Moved to Power BI.
The data wasn't just answering questions anymore. It was telling him which questions to ask. He started connecting data across departments — things that had never been in the same room. Finance and operations. HR and productivity. Customer data and training completion. He became the person people went to when the numbers didn't make sense. Not because he was assigned to that role. Because he kept showing up with answers.
2020
XLOOKUP dropped. He already knew.
Microsoft finally built into the product what millions of people had been doing manually with workarounds for 35 years. The need had created the function — exactly as it always does. He'd been following the beta. He was teaching it to the team before IT sent the memo. When someone asked how he already knew, he shrugged. He'd been paying attention. He's always been paying attention.
Today
This is where the rubber meets the road.
One intro class. Two decades. A trajectory nobody could have predicted — and nobody could stop. This is what investment in people actually produces. Not a completion rate. Not a satisfaction score. This.
Stronger Teams
He lifts others. Knowledge spreads. What he figured out at 11pm doesn't stay with him — it becomes institutional memory. The people around him get better because he was there.
Better Systems
He fixes what breaks before it breaks badly. He improves what everyone else has learned to tolerate. The org runs better, faster, cleaner — because he was in it and cared enough to do something about what he saw.
Smarter Decisions
He connects dots others don't see because he's been building the network for twenty years. He asks the question before anyone else knows there is one. That's not talent. That's compounding knowledge doing its work.
Compounded Impact
His work creates disproportionate value over time in ways that cannot be predicted in advance and cannot be captured on a spreadsheet after the fact. One intro class. Two decades. Incalculable return.
You cannot mandate this. You cannot schedule it.
But you can create the conditions for it — and then get out of the way.
There is a cost to this that never appears in a budget. When I facilitated leadership development sessions, I would ask: do you use any of this outside of work? Every time, without fail, hands went up. Someone who used to fight constantly with their spouse discovered they simply valued different things — and knowing that changed everything. My sister and I had a moment where I realized she had never known there was an order to the glasses in the cabinet. Once I knew it wasn't intentional, everything about it changed.
People who learn to give specific, useful feedback at work take that skill home. They become better parents. Better partners. Better neighbors. The skills don't stop at the door of the office. They ripple. A workforce of people who were genuinely developed — who were taught, not just managed — produces better families, stronger communities, more functional institutions.
"This generation is the first in modern history to score lower on standardized cognitive measures than the one before it. Not because they are less capable. Because the decision was made, repeatedly, at every level, to do more with less — to remove knowledge from people in the name of efficiency."
The workplace is the most reliable place most people will ever get structured, intentional exposure to how to think, how to solve, how to build something that lasts. When organizations stop investing in that, they are not just cutting a budget line. They are cutting off a source of human development that has nowhere else to come from for most people.
This is not inevitable. It is a choice. And it is being made — right now, in budget meetings, in headcount decisions — by people who have never once had to prove that their own development was worth what it cost.
"For decades, we have made the same trade: remove the human knowledge, keep the output. Calculators replaced math reasoning. Spell check replaced writing structure. AI is the most powerful version of that pattern yet — and the most dangerous, because the stakes of not understanding it are higher than any tool that came before."
SPARKS Method Master Thesis, March 2026
Section Two
The math they don't do.
Organizations are meticulous about measuring the cost of training. They are remarkably incurious about the cost of not training. These numbers have been sitting in plain sight for years. Here they are.
Tap or click any number to see the full story
$468
Per person · per year · large companies
$85K
Per month · AI tools · average org
40%
AI spend that underperformed
$142M
Annual digital inefficiency
13.7
Formal learning hours in 2024
94%
Would stay if you invested
33–200%
Cost to replace one employee
60%
Never received formal training
Select any number to see what it actually means.
→ Start with $468
$468
Per person · per year · large companies
What large corporations invested in training per employee in 2025. Less than most monthly software subscriptions. Less than one conference ticket. The full annual bet on a compounding asset.
Average monthly AI tool spend per organization in 2025 — a 36% increase from 2024. No proof of return required upfront. No training budget included. Just the tools, sitting there waiting for someone who knows how to use them.
Of the average $54M enterprise transformation budget in 2026. The tools didn't fail. The people using them were never prepared. Deployment without enablement is a waste multiplier.
What digital inefficiency costs the average enterprise per year. $72M in employee time lost to friction. $50M compensating for technology not used effectively. $20M from projects that failed due to low adoption. All of it preventable.
Down from 35 hours in 2020. Of those 13.7 hours, 13% of the budget goes to compliance training that 49% of employees admit clicking through just to complete. The real development number is a fraction of an already embarrassing number.
Of employees say they would stay longer at a company that invested in their development. The answer to the retention problem has been available for decades. Organizations keep spending money on replacing people instead of keeping them.
Of annual salary. For a manager, up to 200%. For C-suite, up to 213%. The savings from not investing in people don't disappear — they move to a different column and get called something else.
Of employees. Skills are self-taught. The organization runs on capability it never built and does not own — until the person leaves and takes it with them. And then pays 33–200% to replace it.
Average monthly AI tool spend — approved on the promise of future efficiency, no proof of return required
$4,830 / yr
Average SaaS spend per employee — renewed automatically, 53% of licenses unused within 30 days
$142M / yr
Cost of digital inefficiency — mostly because people were never trained to use what was bought for them
33–200%
Salary cost to replace one person who left because nobody invested in their development
Training the person who stays?
$468 a year. And you almost didn't approve it.
Section Three
It's not that I mind you having standards. It's that I mind you having two sets.
There is one standard for tools, travel, executive ritual, and AI spend. And a different standard — stricter, narrower, more suspicious — for anything that invests in the humans doing the work.
This is not a policy. It is a habit. And it has compounded for decades.
What gets waved through
AI tooling at $85K/month — approved on the promise of future efficiency. ROI proof: optional.
Software at $4,830/employee/year — renewed automatically. 53% of licenses go unused within 30 days.
Business travel — two-thirds of the $100B annual spend may be waste. Nobody demanded a measurement framework.
Executive offsites — tens of thousands to reenact consensus that could have been an email. Filed under leadership.
What gets interrogated
Training for the people doing the work — requires a business case, a measurement plan, a Kirkpatrick level, and a champion willing to fight for it.
Development programs — "too soft to measure." Meanwhile the cost of not developing people shows up in turnover and rework. Nobody tracks that column.
Letting people go to class — managers block it because it shows on productivity metrics. The cost of keeping them at their desks: invisible.
Onboarding done properly — "we don't have time." 40% of employees quit within a year when training is poor. That cost lands in a different column.
"If leaders can approve tens of thousands of dollars for a meeting that mostly affirms what everyone already knows, they have no standing to demand perfect, immediate ROI from training that actually changes what people can do."
— This document
The Pushback — And What the Data Says Back
These are the three arguments most often used to justify underinvesting in people. Each one sounds reasonable. Each one has a receipt. Click any to see both sides.
What they say
"We invest where we can see a measurable return. Training is frequently too vague, too generic, and too hard to connect to business outcomes. If you can't show me the number, I can't justify the spend."
What the data says back
That same CFO cannot demand immediate linear ROI from a process whose payoff is often delayed and indirect — without also admitting that the costs of not training are hidden rather than absent.
→
"What is the cost of not training people to use the $85,000/month in tools you already approved?"
What they say
"Not every problem is a training problem. You have to diagnose the gap first. Target specific roles with measurable performance gaps. Be strategic. Be surgical. Broad training is waste."
What the data says back
Diagnosing before defaulting is real work and correct. But when the answer is always "don't train everyone," that is not nuanced analysis — it is a preference dressed up as methodology.
→
"The choice was never train strategically or train broadly. It was train everyone — or let everyone keep using it untrained in the dark."
What they say
"The real problem isn't training — it's process design, system complexity, and management quality. Training alone can't fix a broken workflow. You're treating a symptom."
What the data says back
Yes — and that is exactly the point. Training is one lever. But organizations keep tightening the metric on the worker while loosening scrutiny on the system.
→
"You cannot keep adding complexity, tools, and expectations while starving the one investment that helps people navigate all of it — and then act surprised when performance suffers."
Section Four
A dollar is always worth a dollar.
A person is not.
Currency is fixed. One to one. No exceptions. A dollar will never be worth more than a dollar through its own effort. You can invest it and get a little more, but it is predictable, bounded, linear.
A person is not currency. A person compounds. What they know, what they learn, what they do with it, what they teach someone else, what they solve three years from now with something they learned today — none of that is fixed. It multiplies in ways you cannot predict and cannot audit in advance.
So when an organization treats a person like a dollar — fixed value, known output, spend only what you can justify immediately — they are not being disciplined. They are misunderstanding the asset class entirely.
💵
A Dollar
$1 = $1
Fixed. Predictable. Auditable in advance. Will never be worth more than a dollar through its own effort. Cannot solve a problem you didn't see coming.
🧠
A Person
$1 → ∞
Exponential. Unpredictable. Cannot be audited in advance. Solves problems you haven't met yet using knowledge from years ago. Teaches what they know to everyone around them. Compounds.
📊
What You Invested
$468
Per person per year in a large company. The annual bet on the asset with no ceiling — against $85,000 a month on tools that only work when someone knows how to use them.
Last night, a problem appeared that was about to cost an hour. It was solved in under a minute using something learned years ago in a completely different context. Nobody assigned that learning. Nobody knew it would be needed. It just lived in the network until the gap appeared that required exactly that node.
That is the return on the investment you refused to make.
It does not show up on a spreadsheet until the problem it solves is already solved — or until the person who could have solved it is gone.
"Raising formal learning hours by just 10 hours increases output by 4%. Trained employees complete 10% more work in the 12 weeks following a program — and send fewer requests to their managers, freeing up nearly half the manager's time as a secondary return."
Institute for the Study of Labor · Harvard Business School, Baker Library
Before you go
So if people are the asset that compounds, the judgment AI cannot replicate, the knowledge that travels home and changes families — who built the thing everyone is now racing to replace them with?
Who builds what's next?
Not a tool. Not a platform. Not a model. People. People with curiosity, accumulated knowledge, and the ability to think about a problem no one had fully named yet. People who were taught things. People who were given resources. People who were invested in.
Every tool that ever changed the world was built by a person who refused to stop asking what comes next. That has never changed. It will not change now.
"The money is going to get wasted anyway. I would rather waste it on people. Except — and this is the point — it isn't waste at all. A dollar spent on a person who compounds is not a cost. It is the highest return available to you. You just have to believe it before you can measure it."
