Chapter 2: AI Is Not *Merely* a Technological Challenge

Here is a scene you might recognize.

Two colleagues you trust pull you aside after a board meeting. One says, "People are asking whether we should block ChatGPT on organization-owned devices. A program lead is experimenting with curriculum drafts. A partner heard something shared in public and thinks it was 'AI.' We do not need a futurist manifesto—we need a decision."

If you treat that moment as a technical problem, the meeting ends with IT: permissions, filters, vendor contracts. Those matter. But the anxiety in the room is not really about tensors. It is about truth, trust, craft, and who we are becoming when sentences arrive faster than conscience can keep pace.

That mismatch is the point of this chapter.

Genre, briefly: This book is a formation-oriented field guide—for credibility, vocation, and how language-shaping tools press on trust. It is not a systematic theology of technology, not a substitute for your denominational polity, and not a prompt cookbook that stays fresh every quarter. I write as a fellow traveler; you supply Scripture, tradition, and local accountability.

Most leaders do not "interface" with AI the way engineers interface with a stack. You meet it through language—drafts, edits, prompts, summaries, emails, slides—sometimes spoken, usually written. The surface is words. Underneath are servers, weights, and policies you may never see. The lived work, though, is anthropological in the plain sense: it concerns persons, habits, power, and what counts as faithful speech in your community.

Chapter one traced how fluency and trust are drifting into a more complicated relationship. I am not going to replay that evidence here. I want to name what kind of leadership work the drift requires—so that when someone asks for a decision, you are not tricked into offering only a patch.

I am going to use Ronald Heifetz's language of technical versus adaptive challenges because it clears the fog—with one non-negotiable qualification: anything you ship in the real world is also sociotechnical. Data, security, procurement, and law are not optional side dishes. They are part of the meal. So the accurate headline is not "ignore technology." It is this:

Generative AI is primarily adaptive in how it rearranges identity, trust, and practice—and unavoidably technical in how it actually runs.

If you are "not technical," that sentence is not an insult. It is an invitation to lead the adaptive work while commissioning the technical work you cannot personally execute.

The technical side is not vapor. In serious institutions, "sociotechnical" increasingly means recognizable obligations even to a non-specialist board: data stewardship, vendor and incident discipline, and documented oversight for higher-risk uses—scaffolding you will see reflected in public frameworks such as the NIST AI Risk Management Framework and ISO/IEC 42001 (an AI management-system standard). You do not need to love acronyms. You do need to refuse the story that courage replaces documentation where accountability is real.

Technical vs. adaptive (plainly)

Technical problems have known solutions and can often be delegated: patch the server, fix the spreadsheet, replace the router.

Adaptive challenges require people to learn new habits, tolerate loss, revise values, and live in ambiguity long enough for new judgment to form. You cannot "install" adaptive change the way you install software.

Public AI fits the second bucket for mission-shaped organizations because it touches:

• Truthfulness and attribution — who said this, on what grounds, with what accountability?
• Direct human care — what happens when care is mediated or simulated?
• Speed vs. depth — which rhythms get silently selected when the tool rewards velocity?
• Justice and vulnerability — who bears the risk when something goes wrong: the leader, the admin, the least powerful person in the chain?

None of those questions are answered by a brighter engineer alone. They are answered by communities practicing discernment—slowly, imperfectly, together.

The same tools that strain credibility can, with boundaries, help a truthful voice carry. This book stays inside that tension—AI as both stressor and possible aid—without pretending one side cancels the other.

Why non-technical leaders are not disqualified

You might be thinking: I am not technical enough for this.

You may not need to train a model. You do need enough literacy to ask the right questions: Where did this output come from? Who verified it? What happens if it is wrong in public? Who is displaced if we automate this task?

That is not "learn Python or go home." It is minimum viable governance—the same instinct you already use when money, people, or public credibility is on the line.

Here is what I want you to hear: the interface most leaders meet is language—prompts, documents, edits, reviews. You are already trained, often deeply, in the hard parts: naming reality, reading tone, spotting when a paragraph is morally smooth but thin when you press for evidence or accountability. Those skills transfer. Optimizing how your community uses AI is not the same as optimizing weights in a lab.

A warning from recent research (please read twice)

It is tempting to assume the danger is only naivete—that "people who do not understand AI" will misuse it. Recent experimental work complicates that comfort.

Fernandes et al. (Computers in Human Behavior, 2026, DOI https://doi.org/10.1016/j.chb.2025.108779) report that with ChatGPT-class assistance on demanding reasoning tasks, participants gained performance but lost calibration—they overestimated how well they did. In the paper's framing, higher self-reported AI literacy was associated with greater overconfidence. Treat that association as reported finding, not folk wisdom dressed in a citation.

Translation for leaders: fluency and optimism can outrun judgment—including for sophisticated users. The folk story of Dunning–Kruger is not a perfect map here; what matters is the practical warning: the people who feel safest may be the ones who skip the second read.

That is why adaptive leadership on AI is not cheerleading adoption. It is also not sanctified avoidance. It is structured experimentation: hypothesis, bounded scope, human review, rollback plans—especially where harm is asymmetric (frontline counseling work, HR or personnel processes, public accusations, anything touching a vulnerable person).

Chapter one already named the organizational posture that matches this: experimentation without publication until review norms are real. I return to it here because adaptive leadership is where that posture gets kept under pressure—including the second bright line teams quietly violate once the first feels "handled": private counseling notes, personnel files, or similarly sensitive material does not belong in consumer-grade tools treated as if they were a private vault.

Institutions are already acting adaptively

You are not imagining the institutional swirl. Denominations, ecumenical bodies, and large nonprofits are issuing statements, study guides, and cautions—examples among many include Southern Baptist conversations, World Council of Churches materials, and the Vatican's "Rome Call for AI Ethics." They disagree on emphasis; they agree that formation and governance, not only tooling, are in play.

That is adaptive politics in public—messy, incomplete, and still more honest than pretending this is only an app update.

What adaptive leadership will cost you

Heifetz, in Leadership Without Easy Answers (1994) and The Practice of Adaptive Leadership (2009, with Grashow and Linsky), insists adaptive work includes loss: familiar roles, cherished rhythms, stories we told about who we are. If you lead here, you will need to:

• Name losses plainly rather than ridiculing resistance.
• Keep distress in a productive range—enough urgency to learn, not so much panic that people freeze.
• Protect dissenting voices—especially the ones warning about shortcuts.
• Model not-knowing without abandoning responsibility.

If that sounds like ordinary leadership under sustained moral pressure, good. You were never supposed to outsource discernment to a vendor roadmap.

Where we go next

We have named a credibility crisis in chapter one and a kind of crisis here: adaptive, moral, communal—and technical in execution.

The preface named the other side of the same pressure: many organizations have to reconstitute how knowledge, formation, and accountability actually run—and, for the first time in a long while, can, because the cost of building coherent internal systems has shifted. Adaptive leadership here is not cheerleading; it is the work of rebuilding judgment and practice so that tooling does not outpace integrity.

The next question is temporal: why this window, for you, now? Not to manufacture hype—to locate kairos with evidence: adoption curves, discoverability shifts, and the difference between healthy skepticism and dangerous complacency. Chapter three will widen the lens to where we actually are—postures, pace, and the discoverability layer—without asking you to become a futurist.

---

Reflection questions

1. Where is your community trying to solve an adaptive challenge with only a technical fix—or the reverse?

2. Who on your team names shortcuts early—and what would it take for those warnings to shape policy instead of getting buried under urgency?

3. What is one "minimum viable governance" question you could add to your next workflow where AI touches public words?