Before You Commit to Technology, Ask What It Will Make of You

Summary

Technology is never only an instrument. Every product also acts as a tutor: it rehearses behaviors, redistributes attention, rewards certain values, and makes some futures easier to imagine than others. A navigation app may save time while weakening spatial memory; a generative model may accelerate ideation while encouraging aesthetic convergence; a social platform may connect communities while training creators to anticipate an algorithm’s appetite. For founders and creative leaders, this is not an argument against innovation. It is a case for evaluating products at three levels: immediate utility, repeated behavioral effects, and long-term cultural consequences. The best technologies expand human capability without making judgment, agency, craft, or social trust expendable. Before adopting or building a system, ask not only whether it works, but what kind of user, organization, and world its repeated use is likely to produce.

Explain like I'm 5

Imagine you get a calculator. It helps you answer sums quickly, which is useful. But if you use it before learning what numbers mean, you may become fast at getting answers and weak at noticing when an answer makes no sense. The calculator does two things at once: it solves a problem and changes what you practice. All technology works a little like that. A drawing app can help you create, but its templates may make everyone’s pictures look similar. A video feed can entertain you, but it may teach you to switch attention every few seconds. So before choosing a tool, ask: What does it help me do? What does it stop me from practicing? Who chooses what it shows me? And after using it for a year, what might I become better—or worse—at?

Deep dive

Glossary

Affordance

A possibility for action suggested or enabled by an object or interface, such as a handle inviting pulling or a button inviting clicking.

Behavioral curriculum

The habits, expectations, and values a technology teaches through repeated use, regardless of its stated purpose.

Default effect

The tendency for people to accept preselected options, giving product teams significant influence through initial settings.

Deskilling

The erosion of human expertise when a system performs tasks that people previously had to understand or practice.

Human-in-the-loop

A system design in which people review, guide, correct, or authorize automated decisions rather than being fully removed.

Local-first software

Software that keeps primary data and functionality on a user’s device while supporting synchronization and collaboration when connected.

Path dependence

The way early technical choices constrain later options because of accumulated data, habits, integrations, and switching costs.

Productive friction

Intentional effort or delay that improves reflection, safety, learning, consent, or decision quality.

Second-order effect

An indirect consequence that appears after people, markets, or institutions adapt to a technology’s immediate effects.

Technological lock-in

Dependence on a product or vendor that becomes difficult to reverse because of proprietary formats, workflows, costs, or lost skills.

FAQs

Predictions

{"items":["By 2030, leading product teams will maintain ‘capability balance sheets’ that track which human skills a system strengthens, displaces, or makes strategically scarce.","AI procurement will expand beyond accuracy and security to include reversibility, provenance, model-behavior monitoring, and measurable effects on workforce learning.","Premium creative tools will differentiate through taste-preserving constraints, controllable generation, private datasets, and workflows that make authorship legible.","Local-first and interoperable products will gain appeal as founders and institutions treat resilience and exit rights as features rather than technical housekeeping.","Deliberate friction will become a mark of quality in high-stakes domains such as health, finance, education, and civic information.","As synthetic content becomes abundant, scarce human signals—lived experience, accountable judgment, provenance, and coherent point of view—will command greater cultural and commercial value.","Organizations will increasingly separate AI-assisted production metrics from outcome metrics to avoid mistaking higher output volume for better work."}

Risks

{"items":["Deskilling: teams may lose the tacit expertise needed to detect plausible but incorrect automated output.","Aesthetic convergence: creators working from the same models, presets, and platform incentives may produce polished but interchangeable work.","Vendor lock-in: proprietary data formats, embedded agents, and workflow integrations can make switching expensive or operationally dangerous.","Metric capture: once a proxy becomes a target—engagement, velocity, tickets closed—it can displace the real purpose of the product or institution.","Accountability gaps: automation can blur who is responsible when a recommendation, ranking, or generated artifact causes harm.","Unequal exposure: efficiency gains may accrue to owners and customers while surveillance, precarity, and error costs fall on workers or marginalized groups.","Cognitive monoculture: broad reliance on a small number of models or platforms can propagate the same blind spots across industries.","Fragility: organizations that eliminate manual knowledge and fallback processes may fail dramatically during outages, attacks, or vendor changes."}

Opportunities

{"items":["Build AI interfaces that reveal sources, uncertainty, alternatives, and revision history rather than presenting seamless authority.","Create tools that measure learning and retained competence alongside speed, enabling organizations to automate without hollowing out expertise.","Design local-first creative software with open formats, user-owned archives, and optional cloud intelligence.","Offer ‘exit-readiness’ infrastructure: automated exports, vendor-neutral knowledge stores, model portability, and continuity drills.","Develop recommendation systems with adjustable goals such as surprise, depth, locality, chronology, or viewpoint diversity—not only predicted engagement.","Create provenance products for artists, brands, and publishers that record consent, source materials, transformations, and human contribution.","Design slower, high-trust modes for consequential work: pre-mortems, peer review, cooling-off periods, and structured dissent.","Build professional education around AI-era craft, teaching practitioners how to verify, direct, critique, and meaningfully transform machine output."}

For professionals

For product leaders, the principle can become a compact adoption protocol. First, write a two-column impact thesis: the capability gained and the human capacity potentially weakened. Second, map stakeholders beyond the buyer, including junior employees, contractors, subjects of data, communities, and future maintainers. Third, define a baseline before implementation—quality, cycle time, error severity, skill retention, autonomy, and concentration. Fourth, conduct a 30- to 90-day bounded pilot with audit logs and explicit human review. Fifth, hold a red-team session focused on misuse, incentive shifts, lock-in, and what happens when the system is wrong at scale. Sixth, establish non-negotiables: exportable data, named accountability, appeal mechanisms, accessibility, and a tested fallback process. Seventh, review outcomes after novelty fades. A sound decision memo should answer: What gets easier? What becomes harder to notice? Which skills atrophy? Who gains discretion? Who absorbs errors? Can we leave? Finally, assign an owner to monitor cultural effects after launch. Governance is not a one-time gate; it is the ongoing practice of keeping the tool subordinate to the purpose.

Sources & references

• Understanding Media: The Extensions of Man — Marshall McLuhan
• Do Artifacts Have Politics? — Langdon Winner
• NIST AI Risk Management Framework
• Recommendation on the Ethics of Artificial Intelligence — UNESCO
• OECD Principles on Artificial Intelligence
• EU Artificial Intelligence Act — European Commission
• Data and Privacy — Stanford Encyclopedia of Philosophy