Most of us interact with artificial intelligence (AI) through chat interfaces. These are the friendly, sometimes chippy, text boxes we use to query any of the major services. The more advanced among us use a voice interface. Lately, I have been using ChatGPT to practice my Serbo-Croatian. It is a 24/7language helper, ready to translate, converse, and correct my spoken language. Yet another way that we are using AI is image generation. These are the primary modes we currently use to benefit from AI.
What if I told you that your future AI experience is most likely to be none of these methods but is going to be agentic. Agentic AI is in its infancy, but the potential for ministry use is limited only by our imagination.
Agentic AI acts and reasons autonomously from us as human actors. Agentic AI typically relies on a “trigger” to get the process started. This might be the addition of a value to a spreadsheet, a text message, or an email. It might be a temperature increase or somebody walking through your front door. The AI takes this input and formats a request, sending it through a series of application programming interfaces. It might, for example, ask ChatGPT to answer a question. Then, the agentic AI process makes choices about what to do with that information, ultimately completing a task on our behalf.
When I publish a blog post, for example, I add the website address to a spreadsheet. A process on another platform is watching that spreadsheet, checking it every 15 minutes to see what might have changed. Upon seeing a new address, it calls up an AI tool and asks the tool to read the web page and write a 25-word summary. It then asks another AI platform to make an image based on the content of the article. That task reads the article and creates an image, saving the image temporarily for use in the process later. Then, it logs into three different social media platforms, uploads the short text and the image, and makes a report of its actions back in the original spreadsheet. All of this happens without me doing anything other than providing the original trigger.
Agentic AI combines the power of traditional computing processes with AI tools like ChatGPT or Microsoft Copilot. It relies on application programming interfaces (APIs) that define the methods and data that each unique system has available. By using APIs, complex data and queries can be sent between disparate systems, making for powerful integrations.
Agentic AI transforms many of the common ways that we interact with computers. One can use the enhanced capabilities of AI platforms at scale, meaning that replication and interaction are almost limitless. Of all the various ways that AI is currently being utilized, agentic AI has the most potential to replace or reduce the amount of human interaction required to solve complex problems.
Imagine a ministry website in Arabic that operates in the Middle East. A visitor chats with the AI interface and asks questions about the ministry. As the chat continues, the answers provided come from the content on the website itself. At some point, the visitor asks to speak with somebody. An AI agent gets this request and places a phone call to a team member. The AI agent summarizes the content of the chat session and suggests that the team member go online to interact with the visitor. This frees up the team member from being online all the time and makes a handoff to a real person possible.
Another application might focus on providing language instruction to new missionary candidates. I am aware of a mission agency that has already automated language testing for its cross-cultural staff. By triggering an AI agent with an email address, this organization starts the process whereby AI queries a missionary (in verbal form) on a series of standard questions. AI then produces a report about the accuracy and fluency of the spoken language, keeping the staff member accountable to a language learning goal.
Already I am aware of agentic AI being used by ministries for appointment scheduling, case management for immigrant ministry services, online course management, and various office processes.
Unfortunately, agentic AI suffers from the same limitations that AI in general experiences. In my example above, there is no human interaction that allows me to test or interrupt the process as it is being executed. This can make for some very bad outcomes. I once posted a short article about the effects of pornography on overseas missionaries. The images produced by AI were borderline pornographic, and inappropriate for my audience and message. Just as agentic AI has the potential for great good, it also can make harmful mistakes. Current tools for automating AI agents are hard to stop midstream and ask for approvals. This will no doubt change in the months and years ahead.
Another issue that agentic AI raises is how much this technology will further dehumanize us. Anybody who has tried to wade through a customer support system based on automated prompts knows how frustrating a dehumanized interaction with a machine can be. Will our future be filled with AI agents who are not sympathetic, understanding, or able to make exceptions for us?
Agentic AI is also challenging to program. APIs are intended for computer interaction and are prone to being poorly constructed if something does not work right. Even simple automations like the one outlined above take many hours to create and test. It is helpful to have a solid understanding of how computer logic works, and to know how to parse data structures. Fortunately, AI itself is built into many of the tools used to create these automations and can make this process somewhat easier.
Bill Gates recently said that agentic AI will be personal. He predicts that “we will all have an agent that will be a utility helper to get things done… your agent will determine which parts of that are important enough for you to spend time understanding.”1 If he is correct, this personal agent will have great influence over people. It is hard not think about Genesis 11:6b when we hear claims like this, “And nothing that they propose to do will now be impossible for them.”
For now, agents are far from delivering this reality. At this point in the development of AI, agents are tools which automate tasks on our behalf by calling up AI services and autonomously redirecting their output. Of all the current forms of AI that we have available, agentic AI has the greatest possibility of changing how we interact with machines. Even though the chatting and spoken forms of AI are important, ministry leaders should consider how agentic AI might propel the Great Commission forward.
1 Reid Hoffman, “Bill Gates on Possibility, AI, and Humanity,” YouTube, accessed 3/25/2025, www.youtube.com/ watch?v=KeGYI69sWvw.
Ted Esler is the President of Missio Nexus. He previously worked in IT, served cross-culturally in ministry, and has held various leadership roles in Pioneers.
Scripture References are from the ESV
Subscribe to Mission Frontiers
Please consider supporting Mission Frontiers by donating.
Subscribe to our Digital Newsletter and be notified when each new issue is published!