From Smartphonesto Personal Agents

The central issue here is the transfer of human influence and skill to systems that can act on our behalf, and what that does to humans in the process.

1. The smartphone and the normalization of disempowerment

For all intents and purposes, the smartphone normalized routine cognitive outsourcing. We stopped memorizing phone numbers, navigating without GPS, retaining facts, tolerating boredom, managing attention without constant notifications, and coordinating social life without a device. None of these shifts is inherently bad. In fact, many of them are enormously beneficial. But together they habituated humans to embrace a powerful assumption: if technology can make life more convenient, then delegating effort should be viewed as progress.

The bottom line is that the agentic AI age has not come into a culturally neutral environment. Humans have already spent roughly two decades learning that the convenient life is the best life. That friction is inefficient, and that digital systems can help us safely maneuver everyday life. Therefore, personal AI agents could extend the convenience humans have become accustomed to.

2. A phone gives access; an AI agent exercises agency

3. Democratized labor and the rise of agency security

If AI models become cheap enough for pervasive personal deployment, a person may effectively possess a small digital staff: researchers, schedulers, writers, analysts, coders, and administrative assistants. This could be considered democratized labor.

That changes the security landscape. Traditional cybersecurity asks whether an adversary can access a system or data. Agentic environments add another layer: what authority has been delegated to software, and can that authority be manipulated, expanded, impersonated, or exercised without meaningful human review? This suggests an emerging category of agency security: protecting the boundaries, provenance, verification, and accountability of delegated machine action.

4. The agency paradox

Personal agents can genuinely increase human agency. They can expand what an individual can research, create, coordinate, and accomplish. Yet this same delegation can gradually diminish the human capacities and influence that expand this reach. A person may become more productive while becoming less able to perform, understand, contest, or supervise the work being done.

This creates a sort of paradox: AI can increase operational agency while decreasing developmental agency. The problem is gradual disempowerment through convenience.

5. Frictionless AI Land

If AI Land continues to be frictionless, asking people simply to choose friction will sound silly on the face of it. Why take the time to draft, remember, schedule, research, or decide when an AI agent can do these things faster and better? The answer cannot be that friction is virtuous, so humans should embrace it.

The more useful distinction is between three kinds of friction:

• Wasteful friction: effort that consumes time without empowering humans or protecting their values.
• Protective friction: pauses in projects, permissions, verification, and human review that prevent unsafe or irresponsible delegation.
• Formative friction: effort through which people cultivate judgment, empathy, patience, wisdom, and responsibility.

The strategic objective should be to remove wasteful friction, design protective friction, and preserve formative friction.

6. What AI agents cannot do for us

This notion actually connects the AI question to an older one: what is the good life, and what kind of person should technology help us become? In the classical tradition, the good life involves the cultivation of virtues such as wisdom, courage, temperance, and justice. Quite frankly, there is no convenient way to cultivate the virtues. They are refined through practice, restraint, relationships, choices, and sometimes suffering.

An agent may help a person identify a virtuous course of action, but the AI itself cannot cultivate virtue. It may recommend patience, but it cannot become patient on our behalf. It may generate an apology, but it cannot take responsibility for us. An AI agent simply cannot bear the moral weight of a person having to make difficult choices. The human must do this.

7. A different way to resist the tyranny of convenience

The Benedictine way provides us with an incredible example of what truly matters in life. They intentionally structured life around practices that resisted being governed by immediacy, appetite, productivity, or convenience. Prayer interrupted work. Silence interrupted noise. Community overrode excessive acts of autonomy. Rhythms of worship constrained the tyranny of the urgent.

The lesson for us in the age of AI is that character development strong enough to resist a culture of convenience usually requires certain practices. A Christian response to AI Land may therefore require something analogous to a rule of life for interacting with AI agents: explicit practices governing what we delegate, what we retain, when to pause, where humans remain accountable, and which capacities we intentionally continue cultivating.

8. Use case: the AI chief of staff

Imagine the executive director of a nonprofit organization using a personal AI chief of staff. The agent triages email, manages scheduling, researches donors, prepares briefing books, drafts correspondence, summarizes meetings, and identifies operational risks. This is legitimate removal of wasteful friction and can free the leader for higher-value human work.

Now consider the following scenario. The organization must eliminate several positions. The agent can analyze budgets, rank options, draft the restructuring plan, prepare severance packages, schedule the meetings, predict employee reactions, and generate talking points. Eventually it could conduct much of the communication itself.

The efficiency case for delegation may be obvious. But the difficult human conversation is also a character development moment. The leader must exercise courage, compassion, judgment, accountability, and presence. If the agent disrupts this encounter, it has disrupted a place where leadership itself is practiced.

9. Implications for Attention Inc.

Attention Inc. exists to help individuals and institutions decide where working with AI agents increases human agency and where it diminishes it. We ask a different set of questions to measure this concept:

• We shift the governance question from “What can AI do?” to “What must humans still be able to know, decide, do, contest, and become?”
• We add security questions to govern not only information access, but delegated authority.
• We equip humans to define which tasks may be automated, which require human review, and which should remain human-led alone.
• We use the three-friction framework to distinguish efficiency gains from protective and developmental practices.
• For Christian organizations, we discuss the virtues and spiritual cultivation as a rule of life in AI Land.

10. Four important questions about AI agents

Imagine a world in which nearly everyone has a personal AI agent. The right question goes beyond whether this world is simply better or worse. It is whether pervasive delegated intelligence increases human agency or makes humans more vulnerable. The smartphone era provides an important insight: greater access does not produce wisdom or truth.

These four questions are also published on their own as a discussion guide for boards, staff teams, and small groups.

Is this world safer or more dangerous?

Both. Personal agents could improve safety by detecting fraud, translating warnings, monitoring threats, helping vulnerable people, and giving small organizations capabilities once available only to large institutions. At the same time, pervasive AI agents multiply attack surfaces such as impersonation, persuasion, and the speed at which bad information can become action.

The crucial distinction is that smartphones largely democratized access to information and services; AI agents democratize the capacity to act. Safety therefore increasingly depends on protecting human agency so that people can ask the important questions: who or what may act, within what boundaries, with what verification, and with what human accountability?

Is this world more open to the gospel?

Technically, yes; spiritually, not necessarily. AI agents can translate the Bible, answer questions in various languages, assist missionaries, connect seekers with communities, contextualize teaching, and extend access into places where human workers face barriers. Those possibilities are significant.

But access to the gospel is not the same as receptivity to the gospel. In an environment of infinite personalized content, religious claims can also be generated, optimized, countered, and drowned in competing narratives. AI may lower barriers to the gospel while raising barriers to attention, trust, and wisdom.

Is this world more open to truth or ministry?

It may be vastly more open to information without being more open to the truth. When every person has an AI agent capable of producing persuasive arguments, synthetic evidence, personalized interpretations, and endless counterclaims, the central epistemic problem shifts from finding information to determining what is good information. What is knowledge? And whom can we trust?

Ministry can likewise scale as it becomes easier to simulate. An AI agent can draft sermons, answer theological questions, generate discipleship materials, and conduct follow-up. But Christian ministry also involves embodied presence, accountability, sacrifice, responsibility, and unconditional love. Scalable communication should not be confused with discipleship.

What did the smartphone teach us?

The smartphone made safety, truth, ministry, education, Bible access, and global communication much more possible. It also made surveillance, distraction, misinformation, attention problems, pornography, social comparison, polarization, and cognitive outsourcing more possible. The smartphone exacerbated competing interests without really solving any of the conflicts.

The more defensible claim about personal AI agents is not that they will make the world more open to the gospel, truth, or retaining human agency. It is that they could make the world more contentious and more flooded with competing interests: the gospel versus propaganda, truth versus deception, ministry versus manipulation, flourishing versus disempowerment.

Conclusion

The smartphone put the digital world in nearly everyone’s pocket. Personal AI agents may put delegated power in nearly everyone’s hands. This is a qualitative change. The major risk is that successful delegation will make the surrender of human agency and autonomy feel sensible and normal. The aim is to make human agency, autonomy, and the cultivation of character a main theme and goal, and to expose the tyranny of convenience for what it is.