Category: 2026: The Last Normal Year

  • The Year Intelligence Gets Cheap

    The Year Intelligence Gets Cheap

    The cost to use a given level of AI falls about 10x every 12 months, and lower prices lead to much more use. Moore’s law changed the world at 2x every 18 months; this is unbelievably stronger. — Sam Altman, “Three Observations,” February 2025

    What a Thought Used to Cost

    Let me put the AI cost concept into plain terms, because the whole argument of this week hides inside it. AI sells intelligence by the token, a fragment of a word. A million tokens is about 750,000 words, roughly five long novels in length, reasoned through, with questions answered. Price that cost per unit honestly, and you can watch the ground move under every organization in this series.

    As the founding president of College of Eastern Idaho, analysis of this level was a budget line, and a serious one. A program demand study cost five figures and took a semester, and by the time the report landed, the labor market it described had already shifted. Consultants billed by the day. Boards rationed their questions because answers were expensive, and presidents guessed more than any of us admitted at the time.

    This week I asked a machine for the first draft of an environmental scan of eastern Idaho aimed at a community college demographic, and it arrived before my coffee cooled and it cost less than the coffee.

    This concept of declining costs of compute is the second force of the Last Normal Year, and it is the one leaders feel last, because it arrives dressed as good news. Intelligence is getting cheap. Clarity is not. Everything still ahead in this series, the delegation of work, the drawing of lines, the premium on judgment, runs on this curve. So let us look at the curve itself.

    The Collapse of Compute, in Three Numbers

    First number. Stanford’s AI Index, the most careful annual count of this field, tracked the cost of querying a model at the performance level of the systems everyone met in late 2022. In November 2022, that work cost $20.00 per million tokens. By October 2024, it cost $0.07. The price of a fixed unit of machine intelligence fell more than 280 times in twenty-three months.

    Hold that against the computing curve our generation planned by. Moore’s law doubled computing power about every eighteen months, and it rebuilt the modern world at that pace. To fall 280-fold at Moore’s pace would have taken roughly twelve years. This took less than two.

    Second number. Epoch AI, which tracks the cheapest way to buy a fixed level of capability, found prices falling between 9 and 900 times per year depending on the task, with a median near 50 times per year. Performance at the level of GPT-4, the breakthrough system of 2023, fell from $37.50 per million tokens to $0.18 between March 2023 and February 2025. Read that as a board member. A capability that three years ago only the best-funded labs and companies could afford to use at scale now rents for pennies. Other trackers, the venture firm Andreessen Horowitz among them, put the sustained rate at roughly ten times cheaper per year for the same capability. So far, Altman’s 10x rule has been the conservative estimate, not the aggressive one.

    The third number may matter most because it helps explain the first two. In 2022, achieving a score above 60 percent on a widely used AI knowledge and reasoning test required an enormous AI model with 540 billion parameters, the internal values a model uses to process information and generate answers. The test, known as MMLU (Massive Multitask Language Understanding), measures how well AI answers questions across dozens of subjects, from history and science to mathematics and law. At the time, a model of that size required the computing power of a data center.

    By 2024, a model of 3.8 billion parameters cleared the same bar. Intelligence is not only getting cheaper. It is getting smaller, which means it is getting local: onto the laptop, the phone, and soon the device on the student’s desk and the trustee’s kitchen table, no procurement process required and no permission asked.

    Translate the seven cents into institutional terms. At that price, a college could have a machine read its entire catalog, every syllabus it offers, and ten years of board minutes, then summarize all of it, for less than what one hour of a consultant’s time used to cost. The constraint on institutional self-knowledge is no longer the money. It is whether anyone thinks to ask, and whether anyone checks what comes back.

    Every era gets one input whose price collapses: steam, electricity, computing, cellular airtime. Each time, the institutions that won were not simply the ones that bought the input. They were the ones that redesigned around it. Add these three numbers together and the conclusion is not merely that AI became affordable. It is that the economics beneath every knowledge institution have inverted. The scarce input around which, for a century, we built our budgets, our hierarchies, and an entire consulting industry is on its way to free.

    What Scarcity Becomes

    When a resource gets cheap, the value does not vanish. It moves. For a hundred years, our institutions were organized, reasonably, around the scarcity of analysis. Staff offices, research shops, and consulting firms existed because thinking at scale was expensive, and the executive’s craft was deciding which few questions deserved the cost.

    The rationing is now obsolete. The craft is not; it has been promoted. When answers are abundant and nearly free, the bottleneck of leadership moves upstream, to three things that are not getting cheaper. The first is the question. Peter Drucker warned that the important and difficult job is never to find the right answers; it is to find the right question. In a world of seven-cent answers, the quality of leadership increasingly depends on the quality of inquiry.

    The second is verification. Cheap answers arrive fluent, confident, and occasionally wrong, and they arrive by the thousand. Someone has to check them against the record, and checking well is slow, disciplined human work. The third is judgment: deciding what the answers mean for this institution, these students, this mission. AI can make an answer cheap. It cannot make the Why cheap, and the Why still belongs to the board. And notice: none of the three, question, verification, or judgment, can be delegated without losing the thing itself. A board that outsources its questions has outsourced its leadership; however cheap the answers get.

    There is a budgeting trap inside the good news, and boards with collaboration of CEOs should name it before their CFOs do. When the price of a unit falls, total use explodes. Economists call it Jevons paradox. Enterprise spending on LLM services rose from $3.5 billion to $8.4 billion in six months during 2025, even as the unit price was collapsing. Do not budget for savings from cheap intelligence. Budget for reallocation, because every department will find ten new uses for a resource that costs seven cents, and they will be right about most of them.

    One caution, because clarity includes honesty about the curve itself. Epoch AI notes that the fastest price declines may not persist, and the frontier of capability keeps getting more expensive even as yesterday’s frontier gets cheap. Institutions will live on both curves at once. Knowing which curve a given decision stands on is part of the new clarity.

    Three Questions for the Boardroom

    So put three questions on the agenda, in the boardroom and in the cabinet meeting alike.

    First question. If analysis costs nothing, what work do we stop doing? Every institution carries reports nobody reads and studies commissioned to postpone a decision. Cheap intelligence exposes them, because neither defense holds anymore: the reports were excused by the cost of thinking, and the studies by the safety of delay.

    Second question. Where are we still rationing thought out of habit? Policies, approval chains, and planning cycles built in the age of expensive answers now ration a resource that costs seven cents. Some of that rationing still protects the mission. Much of it is simply the past, governing the present.

    Third question. Which of our past decisions were really purchases of scarce expertise, and what are they worth now? Contracts, positions, and partnerships bought for access to analysis deserve a fresh look, not because they were wrong, but because the scarcity they answered is ending.

    The governing question of this series still sits above all three: does this move us toward our Preferred Future? The institutions that thrive after normal will not be the ones with the most intelligence. Intelligence is about to be free, and free things differentiate no one. They will be the institutions whose leaders stayed clear: clear about purpose, clear about their questions, and clear about the decisions that remain stubbornly, valuably human.

    Next week: what happens when change arrives in every sector at the same time.

    At Aman and Associates, we work with governing boards, CEOs, and executive teams to strengthen organizational focus and align resources with mission and impact. In a focused two-hour Zoom Futuring Session for your board or executive team, we help leaders step beyond traditional planning to define their Preferred Future and examine what should continue, what should change, and what may have completed its useful life.

    Rick Aman, PhD | Aman and Associates

    rick@rickaman.com | rickaman.com/articles

  • The Year AI Stopped Answering Questions and Starts Working

    The Year AI Stopped Answering Questions and Starts Working

    Just as electricity transformed almost everything 100 years ago, today I actually have a hard time thinking of an industry that I don’t think AI will transform in the next several years. — Andrew Ng, Stanford, 2017

    As I begin the first of five themes that will make up my thesis of this series, let me state the concept plainly. 2026 is the last year artificial intelligence will be thought of as a parlor trick, a clever novelty for the early adopters down the hall. In 2027, adoption begins in earnest, and not gradually. Once a capability crosses from interesting to indispensable, organizations move all at once, and the adoption curve turns vertical.

    If 2026 is the last normal year, it is worth pausing to say what normal actually looked like. It looked stable. Artificial intelligence was in our institutions, but it lived on the fringe, in the hands of early adopters, used mostly for research. We asked it things. Which restaurants near the theater take reservations? When is the finance committee free? What do these invoices say? Then we took the answer and did the work ourselves; made the call, chased the calendars, matched the paperwork by hand. From the executive team or board table, all of that looked manageable, a policy here, a pilot there. It was a reasonable response to the technology as it then stood. The trouble is that the technology is not standing still.

    2027 is where the migration begins. The early adopters, as they usually do, are showing us the destination first: the move from research to task accomplishment, from asking AI to assigning it. “Book a table for four, Friday at seven, Italian near the theater,” and the table is held, the confirmation sent, the calendar marked. “Get the finance committee together next week for ninety minutes,” and five calendars are checked, a room booked, invitations sent, the packet assembled. “Match this month’s invoices against our contracts and pay what agrees,” and the routine is paid and filed while only the exceptions wait for a human. Notice what changed. The deliverable is no longer an answer. It is an outcome. And once AI can be handed outcomes, a new question moves to the center of institutional life: which tasks should it be assigned, and which should it never touch? That is not a technology question. It is a leadership or governance question, and it belongs to boards or CEOs.

    Let me own the nuance before going further. AI has not stopped answering questions, and it will go on answering them. What has shifted is the center of gravity. I know the difference from my own practice. For years I have taught leaders to prompt with a role, with information, and with a mission, and to make the machine interview them before it advises them. That was still a conversation, a better one, but a conversation. What is arriving now is that same discipline put into action: a task handed over with an outcome attached, and the work done while you do something else. An answer sits there waiting for you. An agent acts.

    The Evidence Is Already Here

    This is not a forecast. Anthropic, the company behind Claude, tracks how its systems are actually used in its Economic Index, and its September 2025 report found that in business use through its API, the programming doorway through which a company’s own software assigns work to AI directly, 77 percent of the use was automation: work delegated, not questions discussed. The honest caveat: those customers are early adopters, technology-forward firms who wired AI in first, and 77 percent is not your college or your nonprofit today. But early adopters are not an exception to the future. They are its preview. What they are doing with software now, institutions will do with operations next, because the tools are being built for delegation and priced for everyone. None of that is happening at scale in colleges and nonprofits yet, and the timeline matters for this series; in our institutions, 2026 is still the year of questions, and assigned work arrives in 2027.

    Look for the early signs in work every leader will recognize. Advising outreach that assembles itself. Invoices processed end to end except for the exceptions. Vendor contracts gathered and cross-checked before legal reads them. Donor research compiled overnight. Board packets assembled from five sources into one. Environmental scanning that used to wait for a consultant’s visit. None of this requires imagining a new institution. It requires noticing that assigned work, quietly, has begun.

    What Changes Inside the Institution

    Workflows change before job descriptions do. That sequence is worth the leadership attention, because it means the transformation will be well underway before it appears in any organizational chart or personnel policy. A process gets quietly rewired. Then another. Only later does anyone rewrite the job that used to contain those steps.

    What moves first is the repetitive, the multi-step, and the rules-based: the work with clear inputs, clear steps, and a checkable outcome. In a college or a nonprofit, that is a larger share of the operating budget than most leadership teams realize, spread across finance, human resources, facilities, and reporting. What remains for people is the work at either end of the process: setting the task well, and reviewing what comes back. Staff time shifts from producing first drafts and chasing routine steps toward verifying, correcting, and judging machine output.

    That shift sounds simple and is not. Reviewing well is a higher skill than producing routinely, and it is a different skill. The employee who once learned the institution by doing its paperwork now meets the institution through exceptions and judgments. Presidents should expect a training and supervision question here long before they face a staffing question, and boards should expect to hear about it in those terms.

    Delegation Is a Leadership Act

    Here is the governance core. The most important decision leaders make about AI is not what it can do, but what it must never decide.

    Boards and executives already know how to delegate. Boards delegate to presidents, to committees, to auditors; CEOs delegate to cabinets and staff; in every case with authority limits, approval thresholds, and reporting lines, and they review the results. Delegating to an agent requires the same discipline, set just as deliberately; what it may do alone, what requires a human signature, and what is off limits entirely. The clean rule is this. AI proposes. Humans choose. AI executes.

    The discipline has two levels, and they follow the institution’s own logic. The boundary around purpose belongs to the board alone; why this institution exists and what it stands for is the board’s to hold. Inside that boundary, the working rules of delegation are set by the CEO and the leadership team, under limits the board approves. What makes both levels necessary is a property of the machine itself. Ask AI a factual question and it answers. Ask it for options and scenarios, the What of strategy, and it answers, and that is legitimate work for it. Ask it a Why question, why this institution exists, what we stand for, whether we should do this at all, and it will answer that too, fluently, confidently, and with no marker anywhere that a boundary has been crossed. Nothing in the machine distinguishes the question it should answer from the question it must never be asked to settle. If those boundaries are not marked by the humans charged with them, they will not exist.

    One more refinement, because it closes a loophole. It is tempting to say leadership keeps the Why and the What and simply delegates the How. But when AI produces three execution paths, complete with costs, timelines, and a confident recommendation, and the institution takes the recommended one, a decision has been made, about money, risk, people, and sometimes mission. Where there are multiple Hows, choosing among them is deciding. Humans must do the choosing, openly, and then let the machine execute what was chosen. Delegation that hides a decision inside a recommendation is not delegation. It is abdication with better formatting.

    Wrong at Machine Speed

    The risks of the agent era are specific, and they begin with scale. A wrong answer sits in one place, where one person can catch it. A wrong agent, acting at machine speed inside connected systems, can be wrong in a hundred places before anyone notices. The error is no longer an event. It is a pattern, filed, sent, and paid.

    The defense is not suspicion, and it is not slowness. It is the discipline of verification, applied before trust. In my own work I use a simple three-part check I call the Wisdom Triad. Revelation: what are we actually seeing, in the output and in the record of what the agent did? Instruction: what does it mean for us? Application: what should we do next? With a six-step sanity check underneath it: verify the facts, examine the sources, challenge the assumptions, look for what is missing, seek an alternative view, and apply human judgment. Note the order. Human judgment is the final step of the process, not its first casualty.

    One deeper question I will raise here and deliberately not resolve. If agents do the doing, what happens to the people who used to learn by doing it? Judgment is trained on routine work long before it is exercised on hard cases. That is Week 6 ground, when we take up the premium that cheap answers place on human judgment, and it is planted here because boards and leadership teams should start watching for it now.

    So, begin Monday morning with one question on the agenda, in the boardroom and in the cabinet meeting alike: “What work in this institution is ready to hand to an agent, and what work must we never hand over?” Ask it in that order, and do not let the meeting leave with only the first half answered. Then set the answer where this series began, under the governing question of Week 1: “Does this move us toward our Preferred Future?” The institutions that thrive in the years after normal will not be the ones that delegated the most. They will be the ones that delegated deliberately, and kept the deciding for themselves.

    Next week: what happens when intelligence itself gets cheap.

    —–

    At Aman and Associates, we work with governing boards, CEOs, and executive teams to strengthen organizational focus and align resources with mission and impact. Through board retreats, strategic conversations, or a focused two-hour Zoom Futuring session, we help leaders step beyond traditional planning to define their Preferred Future and examine what should continue, what should change, and what may have completed its useful life.

    Rick Aman, PhD | Aman and Associates

    rick@rickaman.com | rickaman.com/articles

  • The Last Normal Year – When AI Ends Normal

    The Last Normal Year – When AI Ends Normal

    When AI Ends Normal

    “The best way to predict the future is to invent it.” — Alan Kay

    Two months ago I stood in an innovation and technology building in Idaho Falls and watched a ribbon cut. The Future Tech Building is the kind of facility a community dreams about for a generation. What stayed with me was not the building. It was the timeline. The voters who approved a new community college in our region in 2017 could not have described the technology that this building now houses. They were never asked to. They were asked to choose a direction, and the future filled in the details.

    Institutions are always living inside decisions made years earlier. The only real question is whether the next set of decisions gets made as deliberately as that one was.

    A Phrase Worth Borrowing

    James Mullarney of InvestAnswers, in a recent conversation with Jo Bhakdi, called 2026 “the last normal year.” I have borrowed his phrase for this series because it names something I have been watching from boardrooms and accreditation visits across the Northwest: the foundations under our institutions are beginning to move at a different pace.

    Let me be clear about what I am claiming and what I am not. I am not predicting. Prediction is a claim about what will happen, and honest people should loosely hold such claims. I am interpreting. There is a difference. A pilot does not predict the weather over the North Pacific; they interpret it, respect it, and plan the route accordingly. Leaders owe their boards the same discipline with emerging technology: not prophecy, but careful reading of what is already in motion.

    What I am seeing is this. In 2026, a governing board can still treat artificial intelligence as a side experiment, a trial in one department, a line item that can wait for next year’s budget. Beginning in earnest in 2027, AI becomes an operating condition of the institution. We will begin to assume it in budgets, staffing plans, advising, programming, and the questions employers and reviewers ask of us. The AI experimental phase is ending, whether we believe in it or not, and whether we scheduled its end or not.

    Over the next six weeks I will take what I believe to be five AI themes, one at a time, and interpret them for the future of nonprofit organizations. The five are these: 1. AI that stops answering questions and starts doing work, 2. intelligence that gets cheap, 3. change that arrives in every organizational sector at once, 4. the leadership and governance question of where the line between human and machine gets drawn, and 5. the premium that cheap answers place on human judgment. But first, boards and leaders need to understand the shape of the change itself, because the shape is what makes this moment different.

    The S-Curves Are Colliding

    Every technology follows an S-curve: a slow, flat beginning where change is easy to dismiss, a steep middle where improvement compounds quickly, and a plateau of maturity. The telephone took some seventy-five years to run that arc; the smartphone ran it in about a decade. Most of us who lead institutions built our instincts on the flat parts of those curves, where next year reliably resembled last year, with minor adjustments. Psychologists call the habit normalcy bias, the assumption that tomorrow will be like yesterday. It served us well on the flat parts. It misleads us on the bend.

    Artificial intelligence is not riding one S-curve. It is riding several at once. Capability is still climbing the steep middle of its curve. Cost is falling just as steeply down its own: Stanford’s 2025 AI Index reports that the cost of a given level of AI performance fell more than 280-fold between November 2022 and October 2024. Adoption is now entering its steep middle as these tools move from experiment to infrastructure. When S-curves collide, their effects do not add. They compound.

    This is why the plateau of the old curve is the most dangerous place an institution can stand. On the plateau, enrollments look stable, budgets balance, and the strategic plan appears to be working. Everything in the board packet says normal. But the plateau of one curve is often just the flat bottom of the next one viewed from above, and institutions that mistake a plateau for permanence are usually the last to know the bend has already happened beneath them.

    We have watched this movie before, at a slower speed. I spent part of my career in colleges that treated the internet as a side experiment: a computer lab here, a cautious online course there, a website maintained by whoever had time. Within a decade it was an operating condition. Registration, instruction, library services, advising, and the expectations of every student who walked in the door all assumed it. The colleges that read that curve early did not merely adopt technology sooner. They redesigned themselves while redesign was still cheap, and they set the terms their slower peers later had to meet. The difference this time is compression. That transition took the better part of fifteen years. The one now beginning will not grant us fifteen.

    Winner Takes Most

    The steep middle of an adoption curve rewards early adopters in a way the flat parts never did. Capability attracts talent, talent attracts resources, resources buy more capability. The technology world calls this dynamic winner takes most, and it is visible everywhere in the current build-out of computing power, data centers, and energy.

    Colleges and nonprofits sometimes assume this dynamic belongs to someone else’s sector. It does not. Students choose. Donors choose. Employers choose. The institution that learns fastest begins to set the expectations every other institution is measured against, in responsiveness, in personalization, in cost. Winner takes most does not mean one college or one organization takes everything. It means the rewards concentrate; the distinctive and the efficient pull away, applications and gifts and talent flow disproportionately toward them, and the comfortable middle, neither the least expensive nor the most distinctive, gets squeezed first. I do not offer that as a threat. I offer it as a planning fact. Capabilities built during the steep part of the curve are worth considerably more than the same capabilities built after the new plateau has reset everyone’s expectations, because by then they are no longer an advantage. They are merely the price of staying open.

    More Than AI Is Moving

    It would be a mistake, though, to plan for AI alone, because AI is not the only force in motion. The demographics of college-going are shifting, with fewer traditional-age students in many of the regions we serve. Public funding and public trust are both under pressure. Workforce expectations are being rewritten in real time. Planning for AI alone is planning too narrowly, because these forces do not arrive one at a time. Cheap intelligence is landing on institutions already strained by demography and economics, and the interaction of the forces matters more than any one of them in isolation.

    The Question Changes

    So the board’s question has to change. “What will happen to us?” is a spectator’s question, and boards were not seated to spectate. The governing question of the next several years is “What future will we choose?”

    That is not a slogan. It is the actual work. A Preferred Future is chosen deliberately or it is inherited by default; there is no third option in which the future politely waits for our planning cycle. The discipline I have used with boards for years still holds, and it matters more now: direction provides continuity while strategy provides adaptability. Mission stays steady. Plans flex. And one question belongs on every agenda, asked of every significant proposal that crosses the board table: Does this move us toward our Preferred Future?

    Ask it of the small things as well as the large ones. A contract renewal, a new position, a program that has run the same way for a decade: each either carries the institution toward the future it says it wants or spends its scarcest resources, time and attention, on the future it is leaving. Leaders that ask the question consistently, in season and out, will find that it quietly reorganizes everything else on the agenda.

    2026 is the last normal year. What we choose in it will be anything but normal in its consequences. Over the coming weeks we will look at each development in turn, and at what governing boards, presidents, and executive teams should do about them while choosing is still ours to do.

    Communities don’t drift toward greatness. They choose it.

    —–

    At Aman and Associates, we work with governing boards, CEOs, and executive teams to strengthen organizational focus and align resources with mission and impact. Through board retreats, strategic conversations, or a focused two-hour Zoom Futuring session, we help leaders step beyond traditional planning to define their Preferred Future and examine what should continue, what should change, and what may have completed its useful life.

     

    Rick Aman, PhD | Aman and Associates

    rick@rickaman.com | rickaman.com/articles