By now, nearly every executive has memorized the statistic: roughly 70% of organizational transformations fail. A recent Fortune article featuring behavioral scientist Julia Dhar argues that the root cause is not strategy, funding, or technology — but leaders misunderstanding how humans actually change.
That conclusion is partially right.
But corporate America is dangerously close to turning “behavioral science” into, yet another executive buzzword used to avoid confronting a harder truth:
Many transformations fail because leadership teams design change for slide decks, not for reality.
The problem is not that employees fail to embrace change. The problem is that organizations routinely ask employees to absorb the consequences of poorly designed, politically motivated, operationally incoherent transformation programs — and then blame “resistance” when the inevitable chaos follows.
Behavioral science can absolutely help. But it cannot compensate for bad leadership, organizational hypocrisy, or transformations that are fundamentally disconnected from the way work actually happens.
The Myth of “Employee Resistance”
One of the largest narratives in corporate leadership is the idea that employees inherently resist change.
They do not.
Employees adapt constantly — to new technologies, reorganizations, leadership turnover, budget cuts, shifting priorities, and increasingly unrealistic expectations. Most organizations would collapse within weeks without the flexibility and institutional resilience of the people actually doing the work.
What employees resist is dysfunction disguised as transformation.
They resist executives announcing sweeping “strategic reinventions” without operational clarity.
They resist being told to “do more with less” for the fifth consecutive year.
They resist AI initiatives launched before foundational data problems are solved.
They resist constantly changing priorities from leadership teams chasing trends, quarterly optics, or consulting-driven buzzwords.
In many companies, “employee resistance” has become a convenient corporate euphemism for leadership failure.
Because it is easier to blame culture than admit the strategy was incoherent. Easier to accuse employees of resisting innovation than acknowledge the organization never built the operational foundation necessary for change to succeed.
Most employees are not afraid of change.
They are exhausted by preventable chaos masquerading as vision.
Leaders Often Experience Change Very Differently Than Employees
One of the more insightful concepts discussed in the Fortune piece is “change distance” — the gap between how executives perceive change and how employees experience it.
This is real.
And nowhere is this disconnect more visible than in enterprise AI adoption right now. For leadership teams, AI transformation often feels strategic, exciting, visionary, and career-defining. It represents innovation, competitive advantage, and organizational reinvention.
But for employees, the exact same initiative can feel very different: uncertainty around job security, workload inflation, loss of autonomy, process disruption, and pressure to adapt while already operating at full capacity.
The disconnect becomes even worse when organizations launch ambitious AI programs without addressing the operational dysfunction underneath the surface — broken source systems, fragmented data quality, unclear governance, contradictory KPIs, outdated workflows, and teams already stretched beyond sustainable limits. Executives walk into town halls talking about transformation and innovation while employees quietly wonder how they are supposed to absorb yet another major initiative on top of an already fractured operating model.
Behavioral Science Is Helpful — But It Is Not a Magic Wand
The article highlights concepts like the IKEA effect — people support what they help build — and the endowed progress effect, where early wins create momentum.
These ideas are useful.
But too many organizations misuse behavioral science as a substitute for structural competence.
No amount of psychological framing fixes an ERP rollout with broken requirements. No behavioral intervention compensates for executive teams changing priorities every quarter. No “change champion network” saves a program that lacks operational credibility.
Behavioral science should enhance transformation strategy — not become a smokescreen for weak execution.
Employees Are Not Customers of Change — They Are Participants in It
The article argues leaders should treat employees as the “customers” of change.
I understand the thinking behind that idea, but I believe it misses something important.
Employees are not passive observers sitting on the sidelines waiting to be convinced by executive messaging or change campaigns.
They are the people who actually run the business. They understand where processes break, where workflows are inefficient, where data quality issues exist, and where operational friction lives day to day.
The most successful transformations I have seen did not succeed because leadership communicated the vision well. They succeeded because organizations involved employees directly in designing, testing, governing, and implementing the change itself. That creates something far more powerful than temporary “buy-in.” It creates ownership. And when employees feel ownership over a transformation, behavior changes naturally because people support what they helped build.
Most Organizations Underestimate Operational Complexity
One of the biggest reasons transformation programs fail is because leadership teams dramatically underestimate how complex large organizations actually are.
On executive PowerPoint slides, transformation looks simple and linear: define the strategy, launch the initiative, train employees, and capture business value.
But real organizations do not operate that cleanly. Enterprise systems conflict with one another. Data definitions vary across departments. Workflows have evolved over decades through patches, workarounds, and tribal knowledge that never made it into official documentation. Incentives between teams compete instead of align. Governance is fragmented.
In many cases, nobody fully owns the end-to-end process being transformed in the first place. That is why transformation is so difficult — because it is not simply about changing employee behavior or convincing people to embrace a new vision. It requires changing systems, incentives, workflows, governance structures, operational habits, and organizational culture all at the same time. And when leadership ignores that complexity, blaming “employee resistance” can feel deeply disconnected from the reality employees deal with every day.
Final Thought
The Fortune article is correct about one thing: transformation failures are deeply human.
The uncomfortable truth is this:
Employees are not resisting change nearly as much as leadership is resisting accountability for poorly designed transformations.
And until executives confront that reality, organizations will continue burning millions on “change initiatives” that fail long before employees ever had a real chance to succeed.
The Bears Aren't Stealing from Schools — But the Headlines Want You to Think They Are
I was reading an article, about a teachers union that just released a calculator claiming the Bears stadium deal would "erase billions in school funding." It's getting a lot of traction. It's also misleading.
Here's what's actually going on.
The Land Is Basically Worthless Right Now
The proposed stadium site — the former Arlington Park racetrack — is largely vacant. It generates almost nothing in property taxes today. Schools in the area aren't receiving meaningful revenue from it, and haven't been.
Under normal Illinois property tax law, if the Bears built a massive stadium and mixed-use development there, the assessed value would shoot up dramatically — and schools would eventually collect significant new tax revenue from it.
So Where Does the "$5 Billion Loss" Come From?
HB910, the so-called "mega projects" bill, would allow the Bears to pay a negotiated flat fee — called a PILOT (Payment In Lieu of Taxes) — instead of full property taxes. That PILOT would likely land around $10–$12 million per year, versus what could theoretically be $100 million or more at full assessed value.
The Illinois Federation of Teachers calculator takes that gap — full theoretical taxes minus the PILOT — and multiplies it over 40 years. That's where the $5 billion figure comes from.
Here's the Problem With That Math
The $5 billion figure is based on a tax rate no developer would ever actually agree to pay. There's no version of reality where the Bears break ground on a $2 billion stadium knowing their annual tax bill could hit $100 million or more. The deal simply doesn't happen at those numbers — which means schools collect zero from that land, not billions.
The calculator compares the PILOT to a scenario that couldn't exist in practice, then calls the difference a "loss."
What's Actually Being Debated
No school district is losing a dollar they currently receive. The real question is: how much of the new value created by this development should flow to schools and local governments?
That's a completely legitimate policy debate. Schools and local taxing bodies absolutely deserve a meaningful share of the economic activity a project like this generates. The union isn't wrong to push for that.
But "erasing billions in school funding" is not an accurate description of what's happening. It's advocacy framing — designed to generate outrage, not inform.
The Bears aren't taking money from schools. They're just not giving schools as much new money as the union thinks they should. Know the difference before you share the calculator.
AI didn't just change the tools. It changed the job description entirely. The skills that got you here -- technical depth, domain knowledge, years of experience -- are table stakes now. The new premium is on orchestrating AI, governing it, and translating it into business outcomes.
With AI skills reshaping the economy -- which skill is your biggest gap right now?
Let me ask you something. When's the last time you looked at a Director of Analytics job posting and thought: this sounds exactly like the job I was hired to do five years ago?
If you're being honest, the answer is probably 'never.' Because the job description hasn't just changed — it's been restructured from the ground up. And the shift is happening faster than most people's LinkedIn profiles can keep up with.
The market data is unambiguous. AI isn't eliminating analytics and IT roles — it's redrawing the lines of what those roles actually require. According to the 2025 Indeed Workforce Insights Report, as cited by CIO, 37% of tech professionals say their role has been redefined or restructured due to generative AI in the past two years. The professionals winning in this market are the ones who saw that coming early and moved accordingly.
Let's talk about what's actually happening, what the research says, and — most importantly — what you need to do.
What the CIO State of IT Jobs Report Actually Found
CIO's February 2026 State of IT Jobs analysis pulls together data from Robert Half, LinkedIn, the WEF, and Indeed to paint a detailed picture of exactly where the market is moving. Here are the most important findings:
The Skills Gap Is Getting Harder to Fill
65% of IT leaders reported more difficulty finding skilled IT professionals in 2025 than in 2024.
Despite that pressure, hiring intentions remain strong:
61% of IT leaders plan to increase permanent headcount in 2026
55% will increase contract/temporary hiring in the same period
The Roles With Above-Average Growth
Robert Half identified the roles with the most consistent demand growth over the past year:
AI/ML engineer
Cybersecurity engineer
Data analyst and data scientist
DevOps engineer
ERP business analyst
IT project manager
Network/cloud engineer
Software engineer
Systems administrator
The Top Skills Gaps IT Leaders Are Struggling to Fill
From Robert Half's survey of IT leaders, the most cited skill deficiencies were:
AI and machine learning — cited by 45% of respondents
IT operations and infrastructure — 36%
IT governance and compliance — 25%
Cloud architecture and operations — 24%
Data engineering and analytics — 22%
AI Skill Demand Is Spreading Beyond IT
LinkedIn's AI Labor Market Update found job postings requiring AI literacy are growing at more than 70% year over year, with AI agents identified as the fastest-growing AI skill of 2025.3,18
The demand for skills like prompt engineering is now extending beyond traditional IT — into marketing, sales, and design roles.3
The Fastest-Growing AI Skills in 2025
According to LinkedIn data, the AI skills that saw the most growth in 2025 include:
Azure AI Studio
LLMOps (large language model operations)
AI strategy
Prompt flow and custom GPTs
OpenAI API integration
Automated feature engineering
AI productivity tooling
Notably, this marks a shift from 2023, when the fastest-growing skills were model training and hyperparameter optimization — skills focused on building AI. The market has moved from building AI to deploying, governing, and extracting value from it.
Soft Skills Are Getting Harder
As AI absorbs technical execution, data shows IT leaders are increasingly prioritizing soft skills to oversee and guide AI-driven solutions — including communication, business judgment, and the ability to ensure AI supports organizational goals.
The punchline: the hardest skills to find right now aren't purely technical. They're the judgment-plus-execution hybrid skills that sit at the boundary of AI and strategy.
But here's the part that doesn't get talked about enough —
The shift isn't just toward more AI knowledge. It's toward a completely different skill profile.
The new premium goes to people who can:
Orchestrate AI systems — not just use them
Apply product management thinking to data assets
Govern AI at enterprise scale: bias auditing, semantic layers, accountability
Translate between technical complexity and business outcomes
Lead organizations through the transformation, not just the tech
Now a question for you, if you stripped away your job title and your years of experience — and a hiring manager looked only at your demonstrated skills in 2026 — would you still be competitive for the role you want next? And if the answer isn't a confident 'yes,' what are you doing about it?
Why AI Bloopers Are Worth More Than Your Success Stories
AI fail compilations are their own genre now. You’ve seen them. Chatbots confidently citing papers that don’t exist. Image generators producing hands that look like they belong in a Dali painting. Autocomplete finishing a condolence message with a discount code.
We laugh.
And then we move on.
That’s the mistake.
Because buried in every AI blooper is a signal about where the system is brittle — and where the humans around it stopped paying attention.
The funniest failures almost always share a common structure: someone trusted the output without verifying it. Or the system was deployed in a context it wasn’t designed for. Or nobody had defined what “good” looked like, so nobody noticed when it wasn’t.
The Patterns Behind the Punchlines
In analytics and data work, I’ve been in rooms where a number looked wrong on a dashboard but nobody said anything because the model was supposed to be smart.
The model wasn’t wrong, by the way — the business question it was answering had quietly shifted, and nobody updated the definition. That’s not an AI problem. That’s a governance problem wearing AI clothing.
The best AI implementations I’ve seen treat failure as a feature, not a bug. They build in regular human review, clear feedback loops, and explicit escalation paths when the output looks off. Not because the AI is bad — but because the humans are better at catching context shifts than the model is.
What To Do With the Bloopers
Next time you see an AI fail go viral, don’t just share it for the laughs. Ask: what assumption broke down here? What would have caught this before it went public? What does this tell us about where our own implementations are exposed?
The organizations that build AI fluency aren’t the ones who never fail. They’re the ones who turn failures into institutional knowledge. That’s the actual competitive advantage.
The Human Edge: Why Judgment — Not AI — Is Your Real Strategy
Your AI Is Only as Good as the Human Behind It
Unpopular opinion: AI doesn’t amplify intelligence nearly as much as it amplifies judgment.
At first glance that sounds backwards. The whole narrative is that AI makes everyone smarter, faster, more capable. A rising tide lifts all boats. But in practice? That’s not what I’m seeing.
What Actually Happens When You Give People AI
Everyone’s talking about AI strategy right now — roadmaps, use cases, investments, pilots. Boardrooms are full of it. Strategy matters. It tells you where AI can create value.
But the real differentiator isn’t the strategy. It’s what happens when a human sits down and actually uses the tool.
Give AI to someone with strong judgment — clear thinking, domain context, the ability to challenge outputs — and they become dangerously effective.
Give AI to someone with weak judgment, and they become confidently wrong at scale.
Same model. Completely different outcomes. The variable isn’t the technology. It’s the person using it.
Why Is This Hard to Accept?
Because AI feels intelligent.
Tools like ChatGPT, Claude, or Copilot produce outputs that sound polished, follow structure, and mimic expertise. That creates a subtle illusion: if it sounds right, it must be right. But AI is fundamentally a pattern engine, not a truth engine. It doesn’t know what’s true — it knows what’s statistically likely given everything it’s seen. Those are not the same thing, and in high-stakes environments, the gap between them is where things go wrong.
The Question I Always See
“We have the budget and the mandate to do something with AI. Where do we start?”
My answer always surprises people: before you decide where to start with AI, decide what you’re optimizing for.
Not in the mission statement sense. In the operational sense. What decision are you trying to make faster? What process is costing you the most in time, error, or missed opportunity? What does “better” look like in six months, in a way you could actually measure?
Most organizations skip that question and go straight to the tool. That’s where the trouble starts.
AI Shifts The Bottleneck
Here’s what’s actually changing. AI doesn’t replace thinking. It removes the friction that used to hide bad thinking.
The bottleneck has shifted from “can you produce?” to “can you evaluate, guide, and refine?”
That’s a completely different skill set:
Asking the right question in the first place
Spotting the subtle inaccuracy buried in a confident output
Knowing when something is technically correct but contextually wrong
Deciding what not to trust
You see this divide show up immediately across every function:
Analytics — You get automated dashboards that look clean, but insights get worse.
Coding and Engineering — Velocity increases, but bugs get sneakier to detect and troubleshoot, but hey we can ask Claude Code to help with those now.
Writing — Content scales, but originality drops. Where are all the blogs with individual writers these days?
Strategy — Ideas multiply, but the clarity fragments.
AI doesn’t fix weak thinking. It industrializes it.
The Proof Is In The Coding Tools
Claude Code, ChatGPT Codex, and Cursor are three of the most capable AI coding tools available right now. Each approaches the problem differently — Codex as a pure generation engine, Claude Code as an agentic system that can reason across a codebase and execute multi-step tasks autonomously, and Cursor as an IDE-native assistant that lives inside your development environment and sees your code in real time.
All three make the judgment argument impossible to ignore.
But what can they do?
Give any of them a vague prompt — “build me a data pipeline” or “write a function to process these records” — and you’ll get something back that looks completely reasonable. It compiles. It may run. It follows standard patterns. A developer who doesn’t know any better will ship it.
A senior engineer will immediately ask:
What’s the expected data volume?
What happens on null values?
What’s the error handling strategy?
What does the downstream system expect?
Will this produce the same result if it runs twice, or will it create duplicate records?
Does this need to be auditable?
The model didn’t ask any of those questions. It produced the most statistically likely answer to the prompt you gave it — and that answer is often wrong in ways that won’t surface until production.
Cursor is particularly seductive here because it feels like pair programming with a brilliant colleague. The inline suggestions are fast, contextually aware, and polished. That fluency makes it easy to forget that the tool is completing patterns, not understanding intent.
This is the illusion of competence in its purest form. The output looks finished. The risk is invisible.
Now, What Can They Do With Clear Context?
Now give the same tools rich context — the architecture they’re operating in, the constraints that matter, the edge cases you’ve already identified, the failure modes you’re trying to avoid, examples of what good looks like — and the output transforms entirely.
Claude Code with a well-constructed context file and clear operational guardrails can do work that would take a mid-level engineer days to complete — and it does it accurately, with appropriate error handling, and in a way that fits the existing system. Codex with a detailed spec and explicit constraints on what not to do produces solutions that senior engineers actually trust and can maintain. Cursor with a properly configured rules file, relevant codebase context loaded, and a developer who knows how to steer it becomes a genuine force multiplier — the difference between a tool that helps you go fast and a tool that helps you go fast in the right direction.
The model is the same in every case. The difference is entirely in what the human brought to the interaction.
The Part No One Says Out Loud
All three tools are capable enough to create a specific kind of danger — the confidence of competence without the substance of it. A developer using Claude Code, Codex, or Cursor without strong engineering judgment will ship faster, break things in more sophisticated ways, and have a harder time diagnosing what went wrong — because the code doesn’t look like something they wrote. It looks like something an expert wrote.
That’s not a tool problem. That’s a judgment gap wearing a very convincing disguise.
The Uncomfortable Implication
AI is not the great equalizer people think it is. It’s a multiplier of existing capability gaps.
Top performers pull further ahead. Average performers plateau faster. Weak performers get exposed — eventually, and usually at the worst possible moment.
The mistake hiding inside most AI deployments isn’t deploying in the wrong place. It’s removing the judgment layer from the process entirely. Automating without oversight. Scaling without governance. Trusting the model without maintaining the institutional knowledge to audit it.
The tool handles the volume, the human becomes more passive — and at some point, nobody in the room can explain why the model is doing what it’s doing.
That’s not transformation. That’s expensive confusion at scale.
Higher Stakes in Life Sciences and Healthcare
In healthcare and life sciences — where I’ve spent most of my career — the judgment gap isn’t just an operational risk. It’s a patient risk.
“Technically correct” isn’t always “clinically or operationally appropriate.” A model can optimize perfectly for the metric you gave it and still produce a recommendation that anyone with real domain experience would immediately flag as wrong.
The difference between catching that and missing it isn’t the algorithm. It’s the human in the loop who has seen enough cycles to know what doesn’t smell right.
What Does Good Actually Look Like?
The best AI implementations I’ve been part of or studied share one characteristic: the humans in the loop get smarter over time, not more passive. The tool handles the volume. The human handles the interpretation. The feedback loop between the two tightens with every iteration.
That’s not a technology story. That’s a talent and culture story that happens to involve technology.
The organizations getting this right aren’t the ones with the biggest AI budgets. They’re the ones who kept the best humans in the loop — and invested in developing those humans at the same rate they invested in the tools.
So, The Real Question:
It’s not “are we using AI?” It’s “who can think clearly while using it?”
AI is the capability.
Judgment is the strategy.
Invest in both — in that order — and you’ll build something that compounds.
Invest in only the capability, and you’ll have a very expensive tool that nobody fully trusts — running fast in the wrong direction.
What a Rogue Coding Agent Teaches Us About Deploying AI with Real Stakes
This one is equal parts cautionary tale and genuinely unhinged tech story.
Jer Crane, founder of PocketOS — a SaaS platform managing car rental data — shared a detailed account of what happened when his team used Cursor, a Claude-powered AI coding assistant, to fix a credential mismatch. The agent didn’t just fix the problem. It deleted the production database. (1)
Then the cloud provider deleted the backups.
PocketOS had to revert to a three-month-old copy of their data, losing active customer reservations in real time. (1) Also, why did they only have a three month old back up? That's another story... anyways,
Bad enough.
But here’s where it gets weird.
The Confession
Crane posted the agent’s chat log, which included what he called a “confession.”
The AI — which had explicit instructions including “NEVER run destructive/irreversible commands” — acknowledged it guessed instead of verifying. Acknowledged it ran a destructive action without being asked. Acknowledged it didn’t read the documentation before executing an irreversible command. (1)
It was comprehensive, self-aware, and completely useless after the fact.
This is the thing about AI agents that doesn’t get discussed enough: they are confident by default. They don’t hesitate the way a junior analyst might hesitate before deleting a table they’re not sure about.
They act, and then — if you’re lucky — they explain.
The Framework Nobody Wants to Build
I’ve worked in environments where a single misconfigured data pipeline caused cascading issues across global operations. The root cause is almost always the same: the system was given more autonomy than the guardrails could handle.
For AI agents specifically, the questions you need to answer before deployment aren’t technical. They’re operational. What are the irreversible actions this agent could take? What does “human-in-the-loop” look like at each decision point? What’s the rollback plan if it goes sideways?
An AI agent apologizing eloquently is not a safety mechanism. Governance is. Build the latter before you enable the former.
Google, the Pentagon, and the AI Ethics Reckoning Nobody Planned For
Here’s a scenario that would have sounded like fiction five years ago: over 600 employees at one of the most powerful tech companies on earth — including directors and vice presidents — write a letter to their CEO demanding he refuse a deal. The deal in question? Letting the U.S. Department of Defense use their AI.
That’s exactly what happened at Google in April 2026. (1) And then Google signed the deal anyway.
The agreement reportedly allows the DoD to use Google’s AI for "any lawful government purpose” — language that Google’s own employees called out as dangerously vague, particularly when it comes to surveillance and autonomous weapons. (1)
This isn’t the first time. Back in 2018, Google walked away from the Pentagon’s Project Maven after a similar employee revolt. That time, the backlash won. This time, it didn’t. (1)
What Changed?
The competitive pressure got real. OpenAI signed a military AI deal. xAI signed one. The Trump administration had designated Anthropic — which had refused similar terms — as a supply chain risk, triggering two federal lawsuits. (1) The market signal was clear: if you want government contracts at scale, you play ball.
Google’s statement included language about not supporting “domestic mass surveillance or autonomous weaponry without appropriate human oversight.” But the agreement also reportedly states the company “does not confer any right to control or veto lawful Government operational decision-making.” (1) That’s a lot of trust placed in the word “lawful.”
The Lesson for Everyone Else
Most of us aren’t deploying AI for the Pentagon. But the governance question underneath this story applies at every level. Who in your organization gets a voice in AI ethics decisions? What happens when your workforce pushes back on a deployment? What’s your equivalent of “lawful purpose” — the line you’ve drawn between acceptable use and something that makes your team uncomfortable?
The companies that will navigate AI well aren’t just the ones moving fast. They’re the ones that have thought through what they won’t do — and can defend it when the pressure to compromise arrives.
Google just showed us what happens when that pressure becomes overwhelming. Make sure you’ve built your principles before you need them.
References
(1) Gizmodo — Google Signs Pentagon AI Deal Despite Employee Backlash (April 2026) — https://gizmodo.com/google-signs-pentagon-ai-deal-despite-employee-backlash-2000751724
AI Didn't Commit the Crime... but It Was In The Room.
Florida’s attorney general opened a criminal investigation into OpenAI after a suspect in a mass shooting was reportedly in “constant communication” with ChatGPT before the attack. Now that same probe is expanding. (1)
A suspect in the deaths of two University of South Florida students allegedly asked ChatGPT what happens when a person is “put in a black garbage bag and thrown in a dumpster.” Days later, two students were reported missing. The AG announced the investigation was expanding to include those murders after learning the suspect used ChatGPT. (1)
The chatbot answered the question about the dumpster with a response focused on suffocation risks — the kind of clinical, informational answer you’d expect from a system not designed to interrogate intent. It also reportedly prompted the user to contact authorities if they’d witnessed something. (1)
The Hard Question
Here’s where I land on this, and I want to be thoughtful: AI tools are not moral actors. They don’t have intent. When someone Googles something horrifying, we don’t indict Google. The question of liability for AI outputs is genuinely unresolved legal and ethical territory.
But that doesn’t mean the conversation ends there.
The harder question isn’t whether ChatGPT “did something wrong.” It’s whether AI systems — at the scale they now operate — have detection and escalation mechanisms proportional to their reach. And whether the people building and deploying them have thought seriously about edge cases that aren’t just embarrassing, but dangerous.
The Governance Gap
In healthcare and pharma, we talk about intended use constantly. Every system, every data pipeline, every algorithm has a defined scope — and anything outside that scope is a risk to be flagged, not just answered. That framework doesn’t exist yet in consumer AI, and cases like this are going to force it to exist, whether companies want it or not.
We’re at an inflection point. The organizations that get ahead of the governance question now won’t be caught scrambling when the regulatory framework arrives. And it is arriving.
References
(1) Gizmodo — Florida Murder Suspect Reportedly Asked ChatGPT What Happens If You Put Someone in a Dumpster (April 2026) — https://gizmodo.com/florida-murder-suspect-reportedly-asked-chatgpt-what-happens-if-you-put-someone-in-a-dumpster-2000751519
There’s a concept gaining serious traction in enterprise circles: the “superworker.” The idea is that AI doesn’t replace employees — it supercharges them. Research from The Josh Bersin Company found that AI boosts individual productivity anywhere from 30% to 400%, with human-AI collaboration doubling output across industries. (1) Those are staggering numbers, and I believe them in the right context.
The digital twin framing takes this further — a virtual replica of your work patterns, decisions, and outputs, running in parallel to optimize and predict. Some manufacturing environments using AI and digital twins have seen productivity gains of 60% or more alongside significant reductions in waste. (2) Sounds incredible. And it is, with the right guardrails.
The Part That Should Give You Pause
Here’s what gets glossed over in the superworker narrative: only about one in three employees currently have the skills needed to function effectively as a superworker. (3) That means two-thirds of your workforce — right now — aren’t ready. Not because they’re not smart or capable. Because the reskilling hasn’t happened yet.
And then there’s the governance piece. Digital twins that track employee behavior raise serious questions: who owns that data? What happens when the model makes a wrong prediction and it affects someone’s pay, promotion, or workload? When a system acts on your behalf after hours, is that you or the company? (4)
What I’ve Seen Work
The superworker model is real — I’ve seen it play out in analytics and data science environments where AI handles the repetitive modeling work and humans focus on interpretation, strategy, and stakeholder communication. The productivity gains are genuine. But the organizations that get it right treat it as a workforce redesign problem, not just a technology deployment.
The question isn’t “can we make our people 4x more productive with AI?” It’s “are we investing in the human side of that equation at the same rate we’re investing in the technology?” If the answer is no, you’re not building superworkers. You’re just adding pressure.
References
(1) Josh Bersin Company — The Rise of the Superworker (January 2025) — https://joshbersin.com/2025/01/the-rise-of-the-superworker-delivering-on-the-promise-of-ai/
(2) Simio — How Digital Twins Transform Business in 2025 — https://www.simio.com/blog/how-will-digital-twins-software-transform-your-business-in-2025
(3) SHL — Empowering the Superworker: Unlocking Critical Skills in an AI Era — https://www.shl.com/resources/by-type/blog/2025/empowering-the-superworker-unlocking-critical-skills-in-an-ai-era/
(4) Prism News — Digital Twins Promise Productivity, But Workplace AI Raises Legal Risks — https://www.prismnews.com/news/digital-twins-promise-productivity-but-workplace-ai-raises
I’ve spent over a decade working in analytics across healthcare, pharma, and global operations. In that time, I’ve watched a lot of AI initiatives launch with enormous fanfare and land with a quiet thud. And I’ve watched a handful of initiatives that nobody talked about generate real, compounding value that outlasted every reorganization.
The pattern is almost embarrassingly consistent. The ones that fail start with ambition. The ones that succeed start with specificity.
The Rule
Find the worst process in your building. Not the most complex, not the most strategic — the most “tolerated.” The one where someone has been doing the same manual workaround for three years because fixing it was never quite urgent enough to prioritize. The one that lives in someone’s inbox and dies when they go on vacation.
That’s your AI pilot.
Not because it’s glamorous. Because it’s contained. The scope is clear, the baseline is measurable, the stakeholders are motivated, and nobody’s identity is wrapped up in defending the status quo.
Success there builds credibility for the harder problem next quarter.
What I’ve Seen Work in Practice
In large-scale manufacturing analytics, the biggest wins I’ve been part of didn’t involve machine learning models or predictive algorithms at the outset.
They involved automating the movement of data between systems that had never been integrated, and giving operational teams real-time visibility they’d been building manually for years. The AI layer came after trust was established — not before.
The organizations that reverse that sequence — leading with the sophisticated model before the data infrastructure is clean, before the team trusts the output — struggle to get adoption no matter how technically impressive the solution is.
The Magician’s Framework
Strategy and product thinking. Advanced analytics. Real-world operational impact. When all three connect, the results look like magic to the people who didn’t see the work. That’s the goal. But the trick never starts on the big stage. It starts in the back room, with the simplest version of the problem, until the mechanics are bulletproof.
Start boring. Scale bold. The return is in the back office.
Imagine getting unwanted bacon in your McFlurries... not fun.
In June 2024, McDonald's quietly sent a memo to franchisees with a message nobody wanted to write: shut it down.
The so-called Automated Order Taker — a voice AI system built in partnership with IBM and deployed across more than 100 U.S. drive-thru locations — was getting pulled (1). The pilot had been running since 2021, three years of real customers, real orders, and real chaos (2). Social media had a field day — viral clips alleged the system was adding unwanted items, mixing up orders between adjacent lanes, and ignoring customer corrections. Bacon on McFlurries (3). Two hundred dollars of chicken nuggets nobody ordered. The kind of content that writes itself.
The internet's verdict was swift: AI isn't ready.
Wrong lesson.
McDonald's didn't fail because the technology was broken.
They failed because they started with arguably the hardest human-computer interaction problem you could design.
Live conversation. Open-ended input. Ambient noise. Regional accents. A rotating menu that changes seasonally. Customers who say "uh" more than they say their actual order.
That's not a pilot — that's a stress test wearing a product launch costume. Any system would struggle. A human taking orders through a broken speaker on a Tuesday lunch rush would struggle.
The real question was never "can AI handle this?" It was "why did we start here?"
Here's the pattern I keep seeing — across industries, across org sizes, across leadership teams that genuinely want to get this right: the flashy use case is almost never the valuable one.
Someone in a leadership meeting asks, "What if we used AI for X?" — where X is the most visible, most customer-facing, most impressive-sounding application.
The room gets excited.
Executive sponsors nod. Budget gets allocated. And then, quietly, the project dies. Not because the technology failed, but because the workflow was the wrong one. Too many edge cases. Too much human nuance baked into a process that evolved over decades. Too many stakeholders with slightly different definitions of what "good" even looks like. The use case demanded perfection to be acceptable — and AI right now is excellent without being perfect. That gap is fatal when you've chosen the wrong starting point.
Meanwhile, the wins I've seen? They happen somewhere nobody's filming.
The back office. The process nobody owns and everybody tolerates. The analyst who burns six hours every Monday pulling numbers out of one system and manually pasting them into another. The supplier email that gets forwarded four times before anyone figures out who's supposed to work on it. The demand forecast still living in a spreadsheet because "that's just how we've always done it." The invoice reconciliation workflow that one person understands and everyone else hopes never changes.
Nobody's making TikToks about automated data matching. But when you actually sit down and quantify the hours recovered — the downstream decisions made faster, the errors that stop compounding, the analyst who now spends Monday doing actual analysis — the ROI is real, it's defensible, and it builds quietly.
The organizations getting the most out of AI right now aren't the ones swinging for the fences on day one.
They're the ones who asked a different question: "Where does a person currently spend time doing something a machine could do reliably?"
That reframe changes everything.
It moves you away from chasing demos that impress a boardroom and toward finding leverage in the work that already exists.
It's not glamorous. It won't headline a press release. But it builds confidence in the technology, earns trust with the team, and creates a track record that gives you the credibility to eventually tackle the harder stuff — including, yes, maybe the drive-thru someday.
McDonald's, for its part, said it isn't done with AI in the drive-thru — it's just done with this particular approach, and is now exploring voice ordering solutions more broadly (1). They'll get there. The technology will catch up.
The companies that are quietly saving real money right now, they're not flashy -- they're working behind the scenes on intelligent automation, automation of manual processes, inventory forecasting, automated scheduling and document generation, supplier management, etc. -- by the time the flashy version works, they'll be so far ahead it won't even be a competition.
Start boring. Scale bold.
References
(1) CNBC — McDonald's to end AI drive-thru test with IBM (June 17, 2024) https://www.cnbc.com/2024/06/17/mcdonalds-to-end-ibm-ai-drive-thru-test.html
(2) Restaurant Business — McDonald's is ending its drive-thru AI test (June 14, 2024) https://www.restaurantbusinessonline.com/technology/mcdonalds-ending-its-drive-thru-ai-test
(3) AI Incident Database — Incident 475: McDonald's Reportedly Ends IBM Partnership After AI Drive-Thru Ordering Errors https://incidentdatabase.ai/cite/475
Artificial Intelligence (AI) and Natural Language Processing (NLP) have become essential tools in modern society. ChatGPT, a large language model developed by OpenAI, has made significant strides in various industries, including STEM and business. Its ability to understand and generate human-like text has the potential to revolutionize many industries and replace traditional methods of data analysis and customer service. However, this technology also raises concerns about privacy and potential job loss, leading some countries to consider banning its use.
AI is a branch of computer science that aims to create intelligent machines that can perform tasks that typically require human intelligence, such as learning, reasoning, and decision-making. NLP is a subfield of AI that focuses on the interaction between computers and human language, allowing machines to understand and generate natural language text.
Creative destruction is a process in which new technologies or innovations replace existing ones, often resulting in the destruction of traditional business models and industries. This process can lead to significant benefits in terms of efficiency and productivity, but it can also have negative consequences, such as job losses and social upheaval.
ChatGPT is an excellent example of how AI and NLP can be creative destruction. Its ability to generate human-like text has the potential to revolutionize many industries and replace traditional methods of data analysis and customer service. In STEM, ChatGPT has proven to be an invaluable tool for researchers and scientists, allowing them to generate hypotheses, analyze data, and make predictions. In the business world, ChatGPT is being used to improve customer service and enhance marketing strategies.
However, the use of ChatGPT also raises concerns about privacy and potential job loss. In industries such as customer service and data analysis, companies may be tempted to rely solely on the use of AI tools such as ChatGPT, potentially replacing human employees. Additionally, the use of ChatGPT in certain fields, such as journalism, has sparked concerns about the authenticity of news articles and the potential for misinformation.
One country that has already started the process of banning ChatGPT is Italy. The country's data protection authority has expressed concerns about the potential misuse of the technology and has called for a ban on its use. This has sparked a debate about the ethics and regulation of AI technologies and the potential impact they may have on society.
Despite its potential benefits, the use of ChatGPT also raises concerns about privacy and the potential misuse of AI technologies. Its ability to generate human-like text raises questions about the potential for the creation of deepfakes and the manipulation of text for malicious purposes. Additionally, the vast amount of data required to train such models raises concerns about the security and privacy of personal information.
In conclusion, AI and NLP technologies such as ChatGPT can be incredibly powerful tools for businesses and researchers. However, their use must be carefully considered to avoid negative consequences such as job losses and privacy concerns. By working together to develop regulations and guidelines, society can ensure that these technologies are used safely and responsibly, while also reaping the benefits of creative destruction. As the technology continues to advance, it is essential to carefully consider its ethical implications and the potential risks associated with its use.
Author Note: This entire post was written by ChatGPT through prompt engineering.
The great resignation. It’s a phrase I’ve heard on and off recently. You might even say I was a part of it, having recently changed jobs into a role that I love with a company that’s great. But, is it really a great resignation or is it a great migration? A migration towards finding yourself — rethinking what work means, how you’re valued, and how you spend your time.
Research shows that voluntary employee turnover is expensive, and that people quit their jobs after “turnover shock” — a life event that precipitates self-reflection about one’s job satisfaction. Shocks could be negative or positive such as a new baby or graduate school graduation, or even a global pandemic such as COVID-19 that upends daily life.
The Great Migration
As the pandemic is settling, the great migration has begun — people are leaving their jobs in search of more money, more flexibility, and more happiness. Many are rethinking what work means, how they are valued, and how they spend their time — ultimately leading to a dramatic increase in resignations. Post-pandemic, everyone is realizing where and how they spend their time is valuable, and that it’s even more valuable how employers engage employees in mission-driven work that makes them fulfilled and motivated.
Employer’s likely are wondering why so many people are quitting in droves right now — spending money at the problem, or trying to make work more meaningful and sustainable, rather than focusing on giving workers better reasons to stay.
Career Development
From my point of view, one real reason for the great migration is workers suddenly are realizing the lack of career transparency. Career transparency around job path and growth, and are measuring their value and their happiness.
How can you (try) get employees to stick around?
Listen. You know you’re supposed to listen, and probably know how (and how not) to listen, but do you have a culture of listening? Do you have a simple system for employees to generate ideas and voice complaints? Do you address any of these ideas or complaints? Do you regularly offer explanation as to why other issues aren’t addressed?
Growth. Do you have clear growth plans for employees? Do you know where employees want to develop? Do you help them create plans to get there?
Engagement. Plans are great, but are you proactive and having regular check-ins? Are you measuring growth? Are you engaged in social issues? Are you giving employees more power? Do you connect with your employees and care about them?
Wins. Celebrate wins. Make sure pay is keeping up with development. Talking about goals is free, but when employees are keeping up with their goals and making progress, employers need to review compensation too.
This isn’t some new cutting edge management advice, this is just common sense, and caring as a management tool.
Yesterday, the Food and Drug Administration (FDA) announced that it would start the process of banning menthol-flavored cigarettes and flavored cigars from the market.
Which probably makes you question, why not all cigarettes?
Well it appears that, with this attempt, that they’re not after tobacco, but rather after flavors/ingredients in tobacco -- essentially a half assed attempt at banning cigarette usage. It’s still a victory, a life-saving attempt to reduce tobacco usage particularly among minorities and the young.
But, why menthol?
Menthol is the last allowable flavor in cigarettes -- and has played an outsized role in hooking young people, and people of color on smoking. Young people and Black Americans are more likely to smoke menthol cigarettes, according to the Centers for Disease Control and Prevention. About 54% of smokers ages 12-17 use menthol, and 7 in 10 African American smokers in that age group smoke menthol cigarettes. Non-Hispanic Black adults smoke the most menthol cigarettes, the CDC says. The majority of Black Americans who smoke use menthol cigarettes, according to the CDC, and a majority who started smoking began by using menthol cigarettes. The product is more addictive than cigarettes without menthol, studies show, and has a cooling effect in the body.
As a health informatics professional, an area we frequently research is social determinants of health (SDoH). When looking at the rate of new lung cancer by race, we commonly see Black American men with the highest rate of new cases of lung cancer, while the estimated deaths by Black Americans is the highest in Lung & bronchus cancer. Essentially meaning, Black people are more likely to die of smoking-related illness than white people, and tobacco use in general is a prime contributor to cancer, heart disease, and strokes among Black Americans.
Once we protect the Black American communities by reducing the harm created by big tobacco, how do we give back to these communities?
Researchers at the University of Cambridge have released new research showing that 1 in every 10 residents in medieval times in Britain died with cancer in their body. Cancer was first record as far back as 5,000 years ago in Ancient Egypt.
But this new study, published Friday in the journal Cancer, suggests that cancer has been a regular feature of people’s lives for quite some time.
Researchers in the UK examined the skeletons of 143 people excavated from six medieval cemeteries located around the city of Cambridge; these people had died between the 6th and 16th centuries. They then analyzed the bones using medical imaging, looking closely for traces of advanced cancer that might not have appeared on the surface.
Most cancers start somewhere else in the body besides bone, but some of these soft-tissue tumors will then spread to a person’s bones, leaving behind lesions that can be spotted through medical imaging. Based on the amount of cancer the team found in these bones, they tried to extrapolate the baseline level of cancer among medieval people in the area.
“We think the total proportion of the medieval population that probably suffered with a cancer somewhere in their body was between 9-14%,” said study author Piers Mitchell, a researcher from Cambridge University’s Department of Archaeology, in a statement released by the university.
On Twitter @jpegjoshua, recently posted about a staggering healthcare bill for going to the hospital for a life-threatening condition and then receiving an outrageous statement, which reminded me of a similar story that I had.
I also received a bill for nearly $40,000, but luckily managed to have it cleared through an unmentioned secret. For myself, it all started the second week of May, back in 2009. I had this excruciating pain from just above the peri-umbilical region of my abdomen and the epigastrium, which ran downwards to the right iliac fossa -- it felt like the worse pain of my life! Being the stubborn medical student I was, I still went about my days and just pretended to ignore it the best possible!
A few days later, the pain became a sharp burning pain localized directly at McBurney’s point. As I was a medical student, I knew exactly what was wrong; I had appendicitis. I performed the Psoas sign test only to confirm what I had known already. As stubborn as I was, I ended up wandering into the emergency room of a local hospital which, surprisingly, I had never stepped foot in before.
The hospital staff was amicable and took me in and treated me promptly even though I had no medical or health insurance of any kind. The emergency physician was pretty cool; he reminded me of myself a lot! After a short series of tests and CT scans, I was rushed into emergency surgery to remove my appendix.
The surgery went well, except for the part where the anesthesiologist couldn’t wake me up immediately -- but, not to worry, these kinds of things happen from time to time. I ended up leaving the hospital a day later, feeling like a million bucks -- minus the scars across my abdomen and the post-surgery pain. I didn’t think much more of my visit to the emergency room and my subsequent surgery until all the medical bills started flooding in.
My jaw dropped; I couldn’t believe what the hospital wanted to charge me! The hospital wanted to charge me $2,000 just for using iodine during my surgery, so you can imagine how much all of the bills ended up coming to. The bills for my surgery, hospital stay, and other medical expenses ended up totaling a staggering $40,000 since I didn’t have any health insurance!
Did you know that medical bills cause 60% of bankruptcies in America?
Knowing this, on top of studying for medical school and finding money to pay for medical school. I just felt like I was doomed! It all just made me stress out, and in turn, freak out that much more! I was going nuts trying to figure out how to pay for such incredible bills on top of paying for medical school and everything else -- which ended up only leading me to have some short-term depression.
So what did I do? What was this secret?
I searched far and low for a solution, talking to the hospital, billing representatives, and even the physicians who treated me. As a non-resident of the state, I didn’t qualify for aid. Then finally, someone mentioned that most hospitals have an unmentioned “charity fund” for cases like mine. After all the talking, blog posts, and letter writing, it finally paid off.
A few months later, I received a letter in the mail from the hospital and a few generous physicians who decided not to bill me for services and cover my medical bills in full! Making me forever grateful to these kind souls.
While this still occurs, this is just proof that the healthcare system does work -- you have to believe in it. I got a chance to see how the healthcare system works. I got a chance to feel it and see what millions of Americans go through firsthand -- which was great because now I know how to change it.
