Getting Comfortable Feeling Uncomfortable: What Leading a School AI Policy Committee Taught Me About Agency

Getting comfortable with feeling uncomfortable. © IPKat.

Claiming Agency When the Future Arrives Faster Than the Institution Can Plan

In 2014, Horace Dediu gave a talk in Munich about technology adoption curves. His point was not simply that new technologies spread. Everyone already knew that. His more unsettling point was that the cycles themselves were compressing. Technologies that once took generations to diffuse were reaching mass adoption in years. The distance between “interesting novelty,” “consumer habit,” “infrastructure,” and “institutional dependency” was shrinking.

That insight has stayed with me because it changes the management problem. When the future arrives slowly, institutions can wait, study, benchmark, procure, train, and codify. When the future arrives quickly, waiting becomes a decision. Delay does not preserve neutrality; it leaves practice to be shaped by vendors, early adopters, anxious resisters, informal norms, and the path of least resistance.

That was the situation we faced with the Falls Church City Public Schools AI Advisory Committee.

Artificial intelligence was not waiting for a policy. Students already had access. Teachers were already experimenting. Parents were already worried. Vendors were already embedding AI into platforms. Detection tools were already being marketed. Public expectations were already diverging. The Board needed guidance, but the evidence was incomplete, the tools were changing, and the community did not begin from a shared mental model.

The easy response would have been to ask for more time, more research, more consensus, or more certainty. The better response was to get comfortable feeling uncomfortable.

That phrase can sound like a slogan, but it became a working discipline. It meant staying in the uncertainty long enough to learn from it, but not so long that uncertainty became an excuse for inaction. It meant allowing competing values to collide without letting the collision become personal. It meant resisting premature consensus without indulging endless divergence. It meant claiming agency.

Agency is not the same thing as control. Control assumes the environment will hold still long enough for the plan to work. Agency accepts that the environment will move, and then asks: What can we still shape? What commitments are worth making now? What future can we help bring into being?

That became the real lesson of the AI Committee.

The Future Did Not Ask Permission

When schools talk about AI, the debate often collapses into a false binary: allow it or ban it. That frame is emotionally satisfying because it creates the illusion of control. But it is a poor match for reality.

The Committee’s shared baseline named the situation plainly: access is inevitable; detection is weak; benefits are uncertain; risks are real. That sentence did more to focus the work than almost any other artifact we produced.

Access being inevitable does not mean all uses are acceptable. Weak detection does not mean integrity is impossible. Uncertain benefits do not mean benefits are imaginary. Real risks do not mean the only responsible response is prohibition.

Those four realities forced a different kind of conversation. We were not choosing whether AI would exist in the lives of students and staff. We were choosing whether FCCPS would have a principled, public, human-centered way to govern it.

That distinction matters because institutions often confuse discomfort with danger. Sometimes discomfort is a warning. Sometimes it is evidence that a real question has finally been named.

In our case, discomfort showed up in predictable forms. Members worried about academic integrity. They worried about students outsourcing thinking. They worried about teacher use of AI in feedback or evaluation. They worried about privacy, disability rights, family notice, linguistic bias, special education, developmental appropriateness, workload, procurement, equity, and public trust. They also saw opportunity: accessibility, personalization, staff support, future readiness, creativity, translation, differentiated materials, and new ways to make learning visible.

The uncomfortable part was that many of these concerns were true at the same time.

AI can support learning and substitute for learning. AI can expand access and widen inequity. AI can save time and create new governance burdens. AI literacy is essential, but direct AI use is not equally appropriate for every age.

The Committee’s job was not to flatten those tensions. It was to make them usable.

From Competing Values to Productive Tension

A lot of our facilitation practice at OFFSET3 draws from the Competing Values Framework, developed by Kim Cameron and Robert Quinn and later adapted in practical innovation contexts by Jeff DeGraff. The framework is valuable because it refuses the fantasy that organizations have one correct value system.

Create, compete, control, and collaborate all matter. Creative energy helps imagine what could be. Competitive energy creates urgency and performance. Control creates reliability, accountability, and safety. Collaboration creates trust, inclusion, and shared ownership. Each value is incomplete without the others. Each also creates friction with the others.

That was exactly the point.

At the Committee’s formation, we did not try to suppress value conflict. We tried to make it visible. Members brought different forms of expertise and different working styles. Some were oriented toward possibility and future readiness. Some were oriented toward human development, equity, and student well-being. Some pressed for governance, definitions, data privacy, and implementation discipline. Some emphasized speed, clarity, and decision-readiness.

Those differences could have become factions. They did not have to. Properly held, they became a source of design intelligence.

When a member pushed for stronger limits on AI use with younger students, that was not anti-innovation. It was developmental judgment entering the design space.

When another member pushed for AI literacy across grade bands, that was not techno-optimism. It was recognition that students will inherit an AI-saturated world.

When a member challenged AI detectors, that was not indifference to cheating. It was concern that unreliable tools would create false certainty, confirmation bias, and unfair discipline.

When a member asked whether a recommendation could actually be implemented by teachers, that was not obstruction. It was systems thinking.

The Committee became stronger when it learned to hear those moves not as attacks on the work but as contributions to the work. In innovation terms, disagreement was not a bug. It was a sensing mechanism.

The High-Quality Target

At the start, we used a simple innovation cycle: set a high-quality target, enlist deep and diverse domain expertise, take multiple shots on goal, accelerate failure, and learn through experience and experiment.

Innovation cycle illustrating goal setting, expert engagement, iterative experimentation, and learning. Adapted from DeGraff & Lawrence (2020).

That cycle was important because the Committee could not begin by drafting final language. We did not yet know enough. More importantly, we did not yet know what we disagreed about.

The high-quality target was not “write an AI policy.” That would have been too narrow. The target was to produce recommendations useful enough for the Board to make principled tradeoffs under uncertainty. The recommendations had to be clear enough to guide practice, disciplined enough to protect students and preserve human accountability, and flexible enough to adapt as tools and evidence changed.

That target gave us a way to judge the work. A draft could be eloquent but not useful. It could be cautious but not actionable. It could be visionary but not governable. It could be technically sophisticated but unreadable to a lay Board member. It could satisfy one value orientation while ignoring another.

The target gave us something better than comfort. It gave us orientation.

Shared Baseline: Reducing Uncertainty Without Pretending to Eliminate It

The phrase “reducing uncertainty” became central to the Committee’s work. We did not pretend uncertainty could be eliminated. That would have been dishonest. AI is moving too quickly, and the evidence base is too uneven. But uncertainty can be reduced enough to support action.

That idea has deep roots in intelligence analysis and strategic foresight. Kris Wheaton’s work is useful here because intelligence is not about waiting until every fact is known. It is about helping decision-makers act under conditions of incomplete information. A rigid process can be misleading if it gives the appearance of certainty without actually improving judgment. Better methods help people clarify assumptions, test competing hypotheses, identify indicators, and decide what would change their view.

The Committee needed that kind of discipline.

The shared baseline did not settle every policy question. It created enough common reality to stop relitigating first principles at every meeting. We agreed on plain-language definitions of AI, large language models, generative AI, and AI literacy. We identified high-salience claims and more careful policy framings. We acknowledged that AI does not replace teachers, that AI use is not always cheating, that AI can undermine learning when unstructured, and that AI detection is often unreliable.

Those baseline realities did not make the work comfortable. They made the discomfort productive.

Once we had a shared baseline, disagreement became more specific. We could argue about grade bands, direct student use, disclosure, teacher feedback, consent, procurement, risk tiers, and implementation capacity. That was progress. Vague anxiety had become structured uncertainty.

Weather Watching

Innovation work requires weather watching.

Weather watching is different from polling. It is not merely asking people what they prefer. It is looking for weak signals, pressure changes, recurring concerns, edge cases, and early warnings. In the AI Committee, weather came from many directions: staff and teacher work, student input, standing advisory committees, public town halls, survey responses, member expertise, external reports, classroom realities, and Board questions.

Adapted from the innovation‑targeting framework described by DeGraff and Lawrence (2020).

The Committee’s public setting made this both harder and better.

It was harder because advisory committee work operates under open-meeting rules, records obligations, and public trust expectations. We could not simply behave like a private product team in a war room. We had to respect process, transparency, and boundaries around asynchronous collaboration.

It was better because those constraints forced discipline. Miro became a visual workspace, but contributions had to be routed carefully. Meeting records mattered. Drafts had to be traceable. AI-assisted synthesis could speed the work, but it could not become the source of truth. Members needed to be heard, but the Board needed an integrated product, not a scrapbook.

Weather watching kept us honest. It reminded us that AI policy was not one issue. It touched instruction, assessment, academic integrity, privacy, procurement, disability rights, multilingual learners, staff workload, family trust, student well-being, communications, and long-term governance.

The weather was complex. The method had to be stronger than the weather.

Workstreams as Parallel Uncertainty Reduction

The Committee organized into workstreams: research and shared baseline, policy drafting, classroom practice and examples, assessment and integrity, data privacy and security, equity and access, communications and community input, and implementation and change adoption.

That structure could sound bureaucratic. It was not. Done well, workstreams are not silos. They are lenses.

Each workstream reduced uncertainty from a different angle. Research and shared baseline work clarified what was true enough to proceed. Classroom practice tested policy ideas against real teachers, students, assignments, and grade bands. Assessment and integrity separated support from substitution, disclosure from misuse, and human judgment from automated suspicion. Data privacy and security defined what had to be true before tools could touch student data. Equity and access kept attention on who could be harmed, excluded, mislabeled, or left with weaker protections. Communications focused on public trust, while implementation translated the recommendations into capacity, training, sequencing, ownership, and review.

The workstreams mattered because a single conversation could not hold the whole problem at once. The Committee needed a way to distribute attention without fragmenting responsibility.

That is a general lesson for innovation leadership: uncertainty becomes less overwhelming when it is converted into reviewable objects. A principle. A matrix. A scenario. A risk tier. A checklist. A draft section. A decision point. A dissent note. A readiness indicator.

The moment uncertainty becomes an object, people can inspect it, improve it, challenge it, and decide what to do next.

The Glidepath Corridor

The alignment graph we used — what to do on one axis, how to do it on the other — became a way to visualize movement from chaos toward responsible action.

At the beginning, we had low alignment on both axes. Members did not fully agree on what FCCPS should recommend, and we did not yet know how the Committee would produce a Board-ready package under public-meeting constraints and a compressed timeline. That was the chaos zone.

The committee reduced uncertainty by building alignment on what to do and how to do it and tracking towards timely, not premature convergence.

The goal was not to leap instantly to the upper-right corner. That would have been fake alignment. The goal was to move through a glidepath corridor: enough divergence to surface real risks and possibilities; enough convergence to produce a usable recommendation.

This distinction matters. Some disorder is productive. Unmanaged disorder becomes drift.

Early meetings favored divergence: introductions, assumptions, misconceptions, risks, opportunities, values, scenarios, and stakeholder signals. Middle meetings forced translation: principles into constraints, constraints into risk registers, and risk registers into workstream deliverables. Later meetings forced convergence: integrated sections, committee voice, section owners, unresolved-issue triage, and Board-facing synthesis.

The glidepath was not just a metaphor. We instrumented it. Deliverables, owners, dependencies, timelines, source-of-record artifacts, public snapshots, and KPIs helped show whether we were becoming more ready or merely busier. Readiness metrics showed whether work was prepared for the next gate. Convergence metrics showed whether references were intact, owners were assigned, deliverables mapped to valid checkpoints, definitions of done were complete, and recommendations linked back to principles and risks.

That gave the chair and committee a way to manage complexity without relying on intuition alone. We could see where we were drifting: overdue deliverables, missing public links, incomplete dependencies, unresolved owners, weak traceability, or too much divergence too late in the process.

Feeling uncomfortable is useful only when there is a way to navigate.

The Pivot from Anthology to Committee Voice

By our sixth meeting, the Committee had generated an enormous amount of material. That was a success and a problem. Members had contributed thoughtful, substantive, sometimes strongly held views. The record was rich. But the Board did not need a 180-page anthology of individually attributed essays. It needed a clear architecture.

That required a pivot.

We shifted from “everyone writes their piece” to integrated committee-voice sections. We assigned section leads to condense, integrate, and preserve essential points. We reviewed prefaces for neutrality and accuracy. We distinguished real disagreements from wording differences. We allowed dissent where necessary but resisted turning the report into a collection of personal position papers.

That pivot was uncomfortable because it asked members to trust integration. It also asked the chair not to overreach. Strong minority views had to be preserved where they were meaningful. Broad agreement had to be named without pretending unanimity. AI-generated or legacy summaries had to be checked so they did not inflate disagreement or smooth over real differences.

This was one of the most important acts of agency in the whole process.

Agency sometimes looks like generating options. Sometimes it looks like editing. Sometimes it looks like saying: we have enough divergence; now we must converge.

Effectuation: The Future We Can Shape

Effectual entrepreneurship offers another useful lens. Saras Sarasvathy’s research on expert entrepreneurs showed that they often do not begin with prediction. They begin with means: who they are, what they know, whom they know. They ask what they can do next. They limit downside through affordable loss. They form partnerships. They use surprises. They focus less on predicting the future and more on shaping it.

Comparison of causal (managerial) reasoning and effectual (entrepreneurial) reasoning, illustrating how predetermined goals contrast with means‑driven opportunity creation. Adapted from Sarasvathy, Dew, Read, and Wiltbank (2011).

The “pilot in the plane” principle captures the spirit: the future is not something that only happens to us. Under uncertainty, we still have levers.

The AI Committee operated in that spirit. We did not know what AI tools would look like in two years. We did not know which vendor features would become standard. We did not know which educational benefits would prove durable. We did not know how quickly students, teachers, and families would adapt. We did not know how the law, procurement practices, or state guidance would evolve.

But we knew our means.

We had a Board charge. We had parents with unusual expertise. We had staff liaisons. We had student, teacher, advisory committee, and public input streams. We had Miro. We had transcripts. We had surveys. We had workstreams. We had AI tools that could process volume. We had human judgment. We had a public process. We had a deadline.

So we acted.

We did not claim to settle the future. We claimed the next responsible move.

That is the essence of agency.

The Committee Modeled the Policy It Recommended

One of the most important things about the Committee’s work is that we used AI in a way that reflected the policy posture we ultimately recommended.

AI helped us process transcripts, surveys, Miro captures, drafts, references, meeting notes, revision histories, and dense policy material. It helped identify themes, redundancy, gaps, possible mischaracterizations, and candidate syntheses. It accelerated the work.

But AI did not replace judgment. It did not become the source of truth. The Committee maintained source records, public deliberation, human review, chair-managed integration, and traceability back to the actual record.

That distinction — AI as accelerator, not authority — became one of the central lessons for schools.

Students can use AI to support learning without surrendering authorship. Teachers can use AI to support preparation without outsourcing professional judgment. Administrators can use AI to organize information without allowing tools to make consequential decisions. Committees can use AI to process complexity without letting it fabricate consensus.

The tool was powerful. The accountability remained human.

What the Board Recognized

The Board’s response mattered because it showed that the method translated.

Board members did not treat the report as a generic technology memo. They engaged it as a governance architecture. Their questions moved quickly to the right thresholds: student use by grade band, educator use in feedback and evaluation, privacy and data security, academic integrity, detectors, consent and opt-out, analog or AI-free learning, and ongoing governance capacity.

That is a sign of success. The Committee did not eliminate uncertainty for the Board. It organized uncertainty so the Board could act on it.

The presentation was received as clear, useful, and Board-actionable. The structure helped Board members understand what needed to be considered and where boundaries might be needed. The process itself was also recognized: productive friction, deep expertise made understandable, and a model of collaboration that may have value beyond AI policy.

That last point is worth lingering on. The public value was not only the report. The public value was the method.

A community facing rapid technological change needs more than a one-time policy. It needs civic muscles: shared language, constructive disagreement, transparent records, principled tradeoffs, adaptive review, and the confidence to act before certainty arrives.

Claiming Agency

Getting comfortable feeling uncomfortable is not about learning to enjoy chaos. It is about refusing to surrender agency to chaos.

Agency begins when we stop asking uncertainty to disappear before we act.

That does not mean acting recklessly. It means acting with structure. It means naming assumptions. It means inviting competing values into the room. It means distinguishing what we know, what we believe, what we fear, what we need to test, and what would change our minds. It means using pilots, off-ramps, feedback loops, and review cycles. It means preserving human accountability. It means making the record visible enough that others can learn from it later.

For schools, claiming agency in an AI world means refusing both panic and passivity.

It means not telling students that AI is forbidden magic while they use it everywhere else. It means not telling teachers that AI will save them while ignoring workload, trust, and professional judgment. It means not telling parents to trust the district while hiding tool use, data flows, or unresolved risks. It means not letting vendors define educational values by default.

It means saying: we will decide what must remain human. We will decide where AI supports learning and where it substitutes for learning. We will decide what evidence is good enough before scaling. We will decide what transparency families deserve. We will decide how to protect students who are most vulnerable to harm. We will decide how to revisit policy as the facts change.

That is not control of the future. It is responsibility for the present.

Horace Dediu’s lesson about accelerating adoption cycles leads here. If the cycles are getting faster, prediction alone will not save us. Static policy will not save us. Waiting for best practices to stabilize will not save us. Best practices may arrive too late, after bad defaults have already hardened.

What helps is an adaptive culture: one that can learn in public, tolerate discomfort without becoming defensive, and use disagreement as design input. Such a culture can move from chaos to alignment without pretending the path is smooth. It can use tools without being used by them. Most importantly, it can claim agency.

The FCCPS AI Advisory Committee was one local example. It was imperfect, compressed, and uncomfortable. It had too much information, too little time, too many live issues, and no possibility of perfect certainty. In other words, it was exactly the kind of situation institutions will face again and again as technology cycles compress.

The lesson is not that every committee should copy every artifact we used. The lesson is that communities can build a way of working that makes uncertainty actionable.

The practical discipline is straightforward, but it is not easy. Communities facing rapid technological change need to gather the right people, name the values in tension, set a high-quality target, and build a shared baseline before rushing to solutions. They need to watch the weather around them, turn ambiguity into workstreams, and instrument the glidepath so leaders can see whether the work is moving toward readiness or drifting into motion without progress. They need to use AI to process volume without allowing it to replace judgment, converge when it is time, preserve dissent where it matters, and deliver something decision-makers can actually use.

Then they need to repeat the process as the world changes.

That is what it means to get comfortable feeling uncomfortable. It is not resignation, paralysis, or performative courage. It is agency.

FCCPS Committee Record and Project Artifacts

Falls Church City Public Schools AI Advisory Committee. (2026, February 7). FCCPS AI Advisory Committee — Workstreams registry (system of record) [Governance registry]. GitHub. https://github.com/JohnWBlack/fccps-ai-deliverables-governance/blob/main/governance_docs/FCCPS_AIAC_Workstreams_Registry.md

Falls Church City Public Schools AI Advisory Committee. (2026, February 18). FCCPS AI Advisory Committee workstream development summary and lead assignments [Governance record]. GitHub. https://github.com/JohnWBlack/fccps-ai-deliverables-governance/blob/main/governance_docs/FCCPS-AI-Advisory-Committee_Workstream-Development-Summary-and-Assignments.md

FCCPS Ad Hoc AI Advisory Committee. (2026, April 28). Facing AI with Clarity, Care, and Purpose. [Policy recommendations report]. https://drive.google.com/file/d/1My3t6eP_FJbM7t8AslFH634hacIG5xDi

Falls Church City Public Schools AI Advisory Committee. (2026, April 29). Board feedback and impact memo: Reception of April 28, 2026 School Board work session presentation and next-step implications https://drive.google.com/file/d/1jZe69YCTENvCAHBRIZvySRyPEg0zPre1.

Black, J. W. (2026). FCCPS AI deliverables governance [GitHub repository]. GitHub. https://github.com/JohnWBlack/fccps-ai-deliverables-governance

Black, J. W. (2026). FCCPS AI governance: Deliverables planning + tracking dashboard [Web application]. Replit. https://governance-data-viewer.replit.app/dashboard

Black, J. W. (2026). FCCPS AI governance: WHAT / HOW glidepath [Web application]. Replit. https://governance-data-viewer.replit.app/glidepath

Horace Dediu / Technology Adoption

Dediu, H. (2013, November 18). Seeing what’s next. Asymco. https://asymco.com/2013/11/18/seeing-whats-next-2/

Dediu, H. (2014, June 23). Horace Dediu: “Transformation of business and society through technology” [Video]. Asymco. https://asymco.com/2014/06/23/%E2%96%B6-horace-dediu-transformation-of-business-and-society-through-technology/

Censhare. (2014). Futureday 2014. https://www.censhare.com/en/news/futureday-2014

Competing Values Framework / Cameron, Quinn, and DeGraff

Cameron, K. S., & Quinn, R. E. (2011). Diagnosing and changing organizational culture: Based on the competing values framework (3rd ed.). Jossey-Bass. https://www.wiley.com/en-us/shop/general-introductory-business-management/diagnosing-and-changing-organizational-culture-based-on-the-competing-values-framework-3rd-edition-p-9781118003329

Cameron, K. S., Quinn, R. E., DeGraff, J., & Thakor, A. V. (2022). Competing values leadership (3rd ed.). Edward Elgar Publishing. https://www.e-elgar.com/shop/gbp/competing-values-leadership-9781800888944.html

DeGraff, J. (n.d.). Competing values leadership. Jeff DeGraff. Retrieved June 5, 2026, from https://jeffdegraff.com/books/competing-values-leadership/

DeGraff, J., & DeGraff, S. (2020). The creative mindset: Mastering the six skills that empower innovation. Berrett‑Koehler. https://jeffdegraff.com/books/the-creative-mindset/

Quinn, R. E., & Rohrbaugh, J. (1983). A spatial model of effectiveness criteria: Towards a competing values approach to organizational analysis. Management Science, 29(3), 363–377. https://doi.org/10.1287/mnsc.29.3.363

Kris Wheaton / Intelligence Analysis and Foresight

U.S. Army War College. (n.d.). Kristan Wheaton. War Room. Retrieved June 5, 2026, from https://warroom.armywarcollege.edu/author/kris-wheaton/

Wheaton, K. J. (2011, May 21). Let’s kill the intelligence cycle (original research). Sources and Methods. https://sourcesandmethods.blogspot.com/2011/05/lets-kill-intelligence-cycle-original.html

Wheaton, K. J. (n.d.). Sources and methods. Blogger. Retrieved June 5, 2026, from https://sourcesandmethods.blogspot.com/

Effectual Entrepreneurship

Darden School of Business. (2026, May 11). How Saras Sarasvathy rewrote the rules of entrepreneurship. University of Virginia. https://news.darden.virginia.edu/2026/05/11/how-saras-sarasvathy-rewrote-the-rules-of-entrepreneurship/

Effectuation. (n.d.). Effectuation. Retrieved June 5, 2026, from https://effectuation.org/

Sarasvathy, S. D. (2001). Causation and effectuation: Toward a theoretical shift from economic inevitability to entrepreneurial contingency. Academy of Management Review, 26(2), 243–263. https://doi.org/10.5465/amr.2001.4378020

Sarasvathy, S. D. (2008). Effectuation: Elements of entrepreneurial expertise. Edward Elgar Publishing. https://www.e-elgar.com/shop/usd/effectuation-9781848440197.html