Dr. Wallace's Favorite Mental Models

Amara's Law

Why a new technology feels like a letdown in year one and a revolution by year ten — and why both reactions follow the exact same predictable curve.

Technology & innovation Strategic foresight Decision making ~9 min read

1Fast Overview

The idea in one line
We consistently overestimate how much a new technology will change things in the next year or two, and underestimate how much it will change everything over the next decade or two.
Where it comes from
Attributed to Roy Amara, a systems engineer and longtime president of the Institute for the Future, a California think tank, who is said to have made the observation around the late 1970s.
The twist most people miss
It isn't one mistake — it's two opposite mistakes stacked on the same timeline: hype now, complacency later. Most people only ever notice and correct for one half of it.
Why it matters at work
Anyone who plans budgets, roadmaps, hiring, or strategy around a new technology — AI tools very much included — is implicitly betting on a timeline. This model names exactly where that bet tends to go wrong.

2Mental Model Compass

Field of originTechnology forecasting & futures studies
Core mechanismLinear-feeling expectations colliding with the S-shaped, compounding reality of how technology actually diffuses
Best used forTechnology roadmaps, AI strategy, innovation planning, product timelines, personal upskilling decisions
Watch out forUsing it to excuse away short-term accountability, or to dismiss real near-term limitations as "just early days"

3Learn It: Step by Step

Step 1 — Two timelines, one curve+
New technology adoption tends to follow an S-curve: a slow start, a steep middle, then a plateau. Our gut expectation is closer to a straight line that keeps rising right from day one. The gap between that gut-line and the real S-curve is where Amara's Law lives.
Step 2 — The short run: hype outruns reality+
In year one or two, excitement, press coverage, and vendor promises outpace what the technology can actually deliver reliably at scale. The near-term effect is almost always smaller, messier, and slower than the demo suggested.
Step 3 — The long run: comfort outruns imagination+
Once the hype fades, most people correct too far the other way — and quietly stop re-checking their assumptions. By year ten or fifteen the technology has often rewired far more of daily work and life than anyone's two-years-ago mental model allowed for, because second- and third-order effects (new habits, new infrastructure, new business models built on top of the first ones) are hard to picture in advance.
Step 4 — Why both halves share one cause+
Both mistakes come from the same habit: picturing a new technology doing today's tasks a bit faster, rather than picturing the new tasks, habits, and systems it eventually makes possible. That framing caps the short-term estimate too high, because reliability and trust aren't there yet, and caps the long-term estimate too low, because the second-order uses aren't visible yet either.
Step 5 — Reading your own position on the curve+
The practical use isn't predicting exact dates — it's noticing which mistake you're currently at risk of making. Early in a hype cycle, lean skeptical on near-term promises. Once a technology feels "settled" or "already priced in," lean skeptical on your own confidence that nothing more will change.

4Examples Across Different Worlds

AI adoption
Generative AI tools in 2023–2024 were pitched as reinventing entire professions within a year. Most organizations instead got modest, uneven productivity wins at first — then kept finding more foundational changes to workflows, hiring, and job design each year after.
The early internet
1990s predictions that the internet would transform commerce within a few years fed the dot-com bust of 2000. Two decades later, the internet had reorganized retail, media, banking, dating, and work far beyond even the most extravagant original predictions — just much later than promised.
Smartphones
Early smartphone pitches focused narrowly on "phone plus email on the go." It took the better part of a decade for app stores, ride-hailing, mobile payments, and social feeds to fully emerge from the same device.
Electric vehicles
EV adoption felt disappointingly slow through much of the 2010s compared to breathless early projections, which cooled a lot of investor and public enthusiasm — right before charging infrastructure, battery costs, and model variety compounded into a much faster uptake later.
Remote work tools
Video calling and cloud collaboration tools existed for years with modest adoption and modest short-term impact. A forced global shift in 2020 then revealed how much more was already quietly possible, permanently changing norms for knowledge work almost overnight.
Personal upskilling
An employee who learns a new tool often expects to be dramatically faster within weeks, gets discouraged by a clunky learning curve, and drops the tool right before the compounding payoff would have started showing up months later.

5Worked Application

Situation: A mid-size company's leadership team is split over generative AI. One camp wants to go all-in immediately, promising a transformed org chart within two quarters. The other wants to wait and see, pointing to current pilots that are unreliable and visibly overhyped.

Applying the model: Under Amara's Law, both camps are half right. The skeptics are correctly diagnosing a short-run overestimate problem — the live pilots genuinely are less reliable than the pitch decks promised. The evangelists are correctly diagnosing a long-run underestimate risk — waiting for "better" tools later means missing the compounding organizational learning (data, workflows, trust norms) that is hard to backfill once competitors have had several more years of practice. The resolution is a split timeline: fund modest, well-scoped near-term pilots with honest reliability expectations to satisfy the short-run skepticism, while building the underlying capability track — data, training, workflow redesign — on a multi-year roadmap sized for the long-run payoff, and re-estimated upward on a fixed review schedule instead of locked in once.

ApproachYear 1 moveYear 5 result (illustrative)
Hype-onlyPromise transformation in two quartersCredibility damage when reality lands late and smaller
Dismissal-onlyRefuse to fund any pilotMissed early-mover data and workflow advantage
Amara's-Law-awareSmall scoped pilots, honest caveats, funded long-term trackWorkflows genuinely reorganized ahead of slower-moving competitors

This is a composite, illustrative scenario built to show the mechanism — not a specific published case study. The general pattern (short-run overestimate, long-run underestimate) is well documented across multiple technology waves; exact magnitude and timing vary widely by technology, industry, and execution.

6Common Mistakes

Using it as a permanent excuse. "It just hasn't paid off yet" can turn into a standing alibi for any underperforming initiative instead of an honest timeline recalibration.
Assuming every hyped technology follows the curve. Not everything overestimated short-term turns out transformative long-term — some technologies really were overestimated the whole way through. This model describes a common pattern, not a guarantee.
Picking a "long run" that never arrives. Without a concrete re-check date, "underestimate the long run" becomes an unfalsifiable promise that something big is always just around the corner.
Forgetting the curve also applies to risk. Early safety, labor, or social harms can be underestimated too, not just upside. The model is about the size of the effect being misjudged, not just the positive side of it.
Confusing hype-skepticism with technical judgment. Doubting a vendor's one-year promise is not the same as the underlying technology being bad; conflating the two leads to dismissing genuinely useful early tools.
Assuming past experience makes you immune. "We already learned this lesson with the internet" doesn't protect against making the identical short-run/long-run misjudgment with the next genuinely new capability jump.

7When Not to Use It

Forecasting a specific date. Amara's Law names a direction of bias, not a number. Treat it as a planning caution, not a scheduling tool.
Defending something that has genuinely failed. If an initiative has had a fair, multi-year runway and simply hasn't worked, don't use "the long run is still coming" to dodge an honest post-mortem.
Comparing unrelated technologies. A curve shape seen with the internet or smartphones doesn't guarantee the same shape for the next technology — each has its own adoption dynamics, costs, and failure modes.

9Try It Yourself: Prompt Builder

Answer a few questions and build a ready-to-use prompt for pressure-testing a technology timeline you're debating.

Your generated prompt will appear here.
Copied!

10Save as Skill

Preview
Apply to My Problem

This is the portable skill version of this model — written so you can hand it to any AI assistant and get the same thinking applied to your own timeline decisions.


      
      
      Copied!
    

    

      
      
      
      
Your generated prompt will appear here.

      
      Copied!