I have now watched five waves of technology come through, and I have watched a market get one of them completely right and completely wrong at the same moment. I am the person at the table who has seen this before, which is useful about sixty percent of the time and insufferable the rest.
A computer only cares about what you say. It never cares what you mean. People are the exact opposite.
For most of computing history that gap was obvious. You typed the command wrong and the machine said bad command or file name and then sat there, unbothered, radiating patience, waiting for you to become a better person. The failure was immediate, visible, and unmistakably yours.
Language models closed that gap from the wrong side. They answer what you appear to mean, fluently and confidently, and they never mention that they guessed. This is why so many capable people try these tools, get a mediocre result, and quietly decide the whole thing is overhyped. They said one thing and meant another. The system answered the thing they said. Nobody sent a memo.
The tool did not fail. The handoff did. Almost every practical AI skill worth teaching is a skill in closing that gap: saying what you mean precisely, grounding the request in real source material, and checking what comes back. None of it is technical. It is a communication discipline, which is why the people who get good at it are frequently not the engineers, and why my ten years as a youth minister turned out to be better preparation than I had any right to expect.
The World Wide Web. The first one I caught, and still the biggest. I was inside a Fortune 500 company installing Mosaic on corporate machines the year everything tipped, and I went straight from there into the dot-com world that a few years later became dot-bomb for a lot of good people. I own T-shirts from companies that no longer exist. Several, actually.
Collaborative computing. The promise was that if we could get everyone into the same digital room, the work would improve. We got everyone into the room. Some of the work improved. We also invented reply-all, so the ledger is complicated.
Knowledge management. The promise was to capture what the experts knew before they retired. The experts, sensibly, retired anyway. What survived was the insight that most of what an organization knows is not written down anywhere, which is still true and still expensive.
Business intelligence. The one people forget, and it was enormous. Through the nineties and 2000s organizations built data warehouses, dimensional models, OLAP cubes, ETL pipelines, and reporting stacks, and it worked. It worked so well that in a single year, 2007, Oracle paid $3.3 billion for Hyperion, SAP paid roughly $7 billion for Business Objects, and IBM paid $5 billion for Cognos. More than fifteen billion dollars in twelve months to absorb the last three independent BI vendors, each doing around a billion in annual revenue. The BI software market is worth roughly thirty billion today. That was not a fad. It permanently changed how large organizations decide things.
Artificial intelligence. This one.
Two of the first four promised more than they delivered, and all four left something durable behind once the noise cleared. But the web is the one that taught me the most, because its lesson is counterintuitive. The internet was not overhyped. If anything it was underestimated. What was overhyped was the timeline and the business models. The technology won completely and most of the companies betting on it still went under. Both of those were true at the same time, and the people who could hold only one of them made bad decisions in both directions. The believers bought everything. The skeptics missed the entire thing.
So what I look for now is the part that will still be here in ten years, and I try to build on that instead of on the press release.
Fifteen billion dollars changed hands in 2007 for tools that a state education agency could not have licensed, staffed, or maintained. The gap between what a telecom could do with its data and what a school system could do with its data was not a technology gap. It was a budget gap wearing a technology costume.
In 2001 I heard someone use the phrase Education Intelligence for the first time.
That someone was me.
Not long after, I helped bootstrap Otis Educational Systems. The idea was simple and slightly audacious: take the things Eddie and I had implemented inside a Fortune 500 telecom and bring them to education. Multidimensional data warehouses. Visual ETL. Real reporting. Forecasting and statistical analysis. The whole stack a bank or a phone company took for granted, rebuilt and priced for agencies that could not.
We were early enough that a fair amount of my job was explaining what a data warehouse was, usually twice. But that sentence is still the entire job, twenty-five years later, and I have never found anything I would rather be doing.
Nobody was ever afraid of a spreadsheet. Nobody worried that using a database made them look lazy, or replaceable, or foolish for trusting it. AI arrives carrying all of that. An individual wonders whether using it counts as cheating. A team waits for someone else to go first, because the risk is local and the credit usually is not. A leader is asked to sponsor something they cannot measure and would have to defend if it failed in front of a client. None of that is irrational. It is what thoughtful people do around a system that gives confident answers and shows no work. It does not respond to a mandate, and it has never once responded to an all-staff email. It dissolves when someone they trust shows them.
Human in the loop has become a checkbox, and in practice it produces a person clicking approve on output they could not evaluate if they tried. What holds up is a named subject matter expert with the standing to overrule the system, working from answers grounded in real, governed sources rather than a model’s general recollection of the internet. Grounding and expertise together. Either one alone is theater. This matters most where I work, because our output can end up in a due process hearing, an eligibility appeal, or an audit. A commercial AI error costs money. In the public sector it can be subpoenaed.
You cannot explain an output if you cannot explain its input. Data lineage gets treated as the boring prerequisite to the interesting work, and I will admit it does not photograph well. But it is the entire basis on which anyone will believe the answer, and in public institutions it is the difference between a decision that survives review and one that does not. Repeatability comes from the same discipline. The first implementation of anything is a project. The tenth is a product. The acceleration everyone wants comes from that, not from the model.
This is the part I care about most and it is the reason I never left education. The technology that runs a bank’s analytics operation is not fundamentally different from what a state agency or a rural district needs. What is different is the budget, the staffing, and whether anyone ever bothered to build a version that fits. Usually nobody did, because that customer cannot pay for it. I have spent twenty-five years dragging enterprise-grade capability across that line and sizing it so that a district with one data person can actually run it. A kid’s outcome should not depend on their county’s procurement budget.
Nearly all public conversation about AI in schools is about students cheating. We have now spent several years on it with tremendous energy, which is impressive considering we never fully resolved the calculator.
The larger opportunity sits with the adults. School systems run on people buried in reporting: the data coordinator reconciling submissions against three incompatible systems, the special education administrator assembling compliance documentation, the counselor who cannot get a straight answer about which students are off track because the answer lives in four places and two of them disagree.
Educators should not have to fight their data in order to help their students. I have spent twenty-five years on that one sentence and I have not gotten tired of it yet, which either says something about the sentence or something about me.
It is with the analyst reconciling a report, the consultant drafting a deliverable, the program staffer answering the same question for the ninetieth time. If a meaningful share of an organization recovers even a few hours a week, that shows up in capacity and in margin. It also shows up in whether people stay, because the hours this gives back tend to be the worst hours of the job.
I use these tools every day in real work and recover somewhere between eight and twelve hours a week doing it. That is not a demonstration figure. It is what happens when someone becomes deliberate about how they ask, what they ground the request in, and how they check the result. I would like to report that I reinvest all of it in high-value strategic thinking. Some of it goes to the guitar.
I call it my Artificial Intern. That is not a joke about capability. It is a description of the working relationship. It is fast, tireless, occasionally brilliant, extremely confident about things it should not be confident about, and it does excellent work when someone competent tells it precisely what is needed and then checks the result. Anyone who has ever supervised a bright intern already knows how to run this. You would not hand one an ambiguous assignment and publish the output unread. Same rule.
I am still getting better at it. Being a good operator is a craft worth taking seriously and worth teaching plainly, and I would always rather show someone than tell them.
Every serious problem I have watched an organization have with AI was a relationship problem before it was a technology problem.