Use of AI Guardrails in Missions

image credit: pixabay

I scroll through the headlines of multiple global news outlets each morning over my Cheerios and blueberries. I do not recall a day since January when I did not see a headline about artificial intelligence. U. S. Christians have been early adopters with 44% stating they use AI “very often” or “often.” In one sense AI is like the wild West, uncharted, unknown, and with an unexpected future. Unlimited power (not really, but go with the alliteration) needs boundaries.

True for AI use in general; true for AI use in missions.

Among organizations, the International Mission Board has been at the forefront of studying and applying AI for purposes related to the missionary task. One such tool they developed is Faithbot.io. This resource has been available since 2024. I have used it to train my students in evangelism and apologetics.

This October, I have the honor of co-presenting, with Don Barger and Chris Neal, a research paper at the annual Evangelical Missiological Society on AI titled: “Humanity and Artificial Intelligence: The International Mission Board’s use of Technology in Mission.” We also presented this paper last spring in Wake Forest, North Carolina. Our work examines Faithbot.io as a case study.

Don serves as the Director of Innovation and Artificial Intelligence at the IMB and Chris serves as the Digital Communities Team Leader of the IMB. These guys represent some of the best thinkers and practitioners when it comes to the marriage of theology, missiology, and praxis in relation to AI.

While I will not go into the details behind our research and content of our paper (come to EMS in Dallas or watch online), I do want to direct your attention toward the concept of “guardrails” I raised in my previous post.

Our work includes four particular dimensions with corresponding theological guardrails. Chris did an excellent job developing and explaining them for AI’s use in missions:

• Truthfulness and Fidelity — Is the AI output true? Does it faithfully represent reality?

• Guardrail: AI-assisted work must be tested for truthfulness and fidelity. AI output is a starting point, not final truth. It should be evaluated as to whether it produces clarity or misrepresentation.

• Presence and Communion — Does AI serve or sever connections between people?

• Guardrail: Embodiment and human-to-human relationships are both essential aspects of image-bearing. Technology may extend presence and communication, but it must not displace relational connections and love of neighbor.

• Agency and Accountability — Does AI empower or replace human image-bearing?

• Guardrail: AI may assist, but it cannot replace or obscure the human agent who is responsible before God for action, judgment, and communication. A human image-bearer should remain the identifiable author, decision maker, and accountable party whenever AI is used.

• Formation and Values (Ends, not just Efficiency) — What kind of people and priorities will this use of AI form over time?

  Guardrail: Technology must serve biblically informed values and priorities, not just speed or efficiency. Because AI tools shape what people notice, measure, and reward, they must be tested for formative pull toward control, spectacle, or convenience as ultimate goods.

Faithbot.io is a large langage model-based trained on Scripture, theology, and apologetics. It does not present itself as human but as an AI assistant, and serves four functions in digital engagement: 1) sharing evangelistic and apologetic messages; 2) providing discipleship support; 3) assisting with Bible study, sermon, and theological research; and 4) triaging inquiries which come through digital engagement opportunities. In the ten months prior to our paper, at least 450,000 users across the world, representing fifty languages, engaged with Faithbot.io.

While this AI tool has limitations, it maintains an excellent evaluation in view of the aforementioned dimensions and guardrails. In our Conclusion, Don writes:

What the first ten months have demonstrated is that AI-assisted missions engagement, when built on sound theological principles and deployed with disciplined human accountability, can meaningfully extend gospel access in ways that were not previously possible. It can reach people in contexts too dangerous for direct human contact. It can create space for honest inquiry that social pressure would otherwise foreclose. It can extend brief human encounters into sustained conversations. It can free human workers to focus their limited time on the conversations that most need their presence.

Now, the next time you use Faithbot.io, know that behind the interaction and results is a governing set of theological, ethical, and missiological parameters that help establish healthy boundaries for Kingdom work.