Is technology ever just a tool?

One question I still can’t get away from is trying to figure out what my academic background brings to the church. If my PhD had been in theology, politics or social studies, it would probably be relatively easy to see how it fits, but a PhD in the history and philosophy of science and technology, specialising in the history of electricity, isn’t quite so obvious.

Recent discussions around the use of AI by churches have caused me to re-evaluate that. As I’ve been drawn into conversations about large language models, image generation and the kinds of work we might use AI for, I’ve started to recognise questions that are very familiar from the history of technology. Who decides what a technology is for? How much control do we really have over the way it develops? What happens when something which began as an optional tool becomes embedded in everyday life?

There are two fairly basic ideas from the history and philosophy of technology which can help us think about this: technological instrumentalism and technological momentum.

For this post, when I refer to AI I’m mainly talking about generative AI, particularly large language models and image-generation systems, rather than AI in the broadest sense. That includes tools such as ChatGPT, Copilot and image generators, rather than predictive systems or hypothetical forms of general or superintelligent AI.

Is technology simply a tool?

Technological instrumentalism is the idea that technology is essentially a neutral tool, with the ethical question resting mainly on how people choose to use it. Electricity, for example, can power a hospital, light a home, run a factory or operate a weapon. From an instrumentalist perspective, electricity itself is neutral; what matters is the purpose to which people put it.

Technological momentum complicates this by looking at what happens once a technology becomes part of a much larger system. Once electricity becomes bound up with power stations, transmission networks, appliances, industries and the habits of everyday life, it begins to shape the choices available to us. Nobody is literally forced to use electricity, but living in a society built around it makes opting out increasingly difficult, while activities which depend upon electricity come to seem entirely normal.

So where instrumentalism asks, “What are we choosing to do with electricity?”, technological momentum asks, “How has building a society around electricity changed the choices we think we have?”

The point is that technological momentum doesn’t require us to believe that technology somehow takes over and determines what happens. People continue to make choices, but those choices are being made within systems created by previous decisions. Once infrastructure, investment, working practices and expectations have grown around a technology, some choices become considerably easier than others and AI seems to be moving through that process remarkably quickly.

When AI stops being something we choose to open

An instrumentalist approach to a large language model is fairly straightforward. The same model might help somebody organise research, improve accessibility, proofread a document, draft an email, generate misinformation or produce work which would otherwise have been done by another person. The ethical questions concern why we are using it, whether it is appropriate for the task and what we do with the result.

But what happens when AI becomes embedded in the other technologies around us. Its speed and relatively low cost can begin to alter how long we think a task ought to take, which activities organisations expect to be automated and which skills continue to seem worth spending time on.

We’re already some way into that process. A few years ago, using a large language model usually meant deliberately opening ChatGPT or something similar and asking it a question. Google now integrates generative AI directly into Search through AI Overviews and AI Mode, and in May 2026 described a redesigned Search experience which brings AI much further into the ordinary search box. Google also reported that AI Mode had already passed a billion monthly users.

Microsoft is following a similar route. Copilot is now integrated across Word, Excel, PowerPoint, Outlook and Teams rather than existing simply as a separate service that somebody has to decide to visit.

The choice is gradually moving from “shall I use AI?” towards “shall I make the effort not to use AI?” There are still alternatives, and we can still decide how much attention we give to AI-generated results, but the technology is becoming part of the environment within which those decisions are made.

Electricity followed a much slower version of that process. There was a period when connecting a house to an electricity supply was a deliberate technological choice, followed eventually by a point at which living without access to electricity became increasingly difficult. We don’t usually think of ourselves as making a decision about electricity every time we switch on a light or charge a phone because the technological system has become part of the assumptions around which everyday life is organised.

I’m not suggesting AI has quite reached that point, but I’m no longer sure that the question “should Christians use AI?” is the one we should be asking.

My own use of AI

I do make use of AI models myself, and some of the reasons are closely related to ADHD. My thought processes can be fairly chaotic, and in research I often make intuitive leaps between ideas where I can see that there is a connection before I’ve worked through how to explain all the steps between them. A language model can help me organise that material, test the connections I’m making and identify the bits I’ve skipped over when I try to turn what is in my head into something another person can follow.

It also helps with proofreading, which is something I find exceptionally difficult to do reliably on my own. I can read something I’ve written repeatedly and continue to see what I intended to write rather than what is actually on the page. Using AI to flag mistakes or awkward sections can remove a genuine barrier while leaving the decisions about what I want to say with me.

I’ve also used AI to generate a first draft of minutes from a recording of a meeting, where recording has been agreed and appropriate. Minute taking is another task I find particularly difficult, and being able to work from a transcript and an automatically generated draft can save a considerable amount of time. That doesn’t mean I can simply send the generated minutes out afterwards, because names can be confused, decisions can be misunderstood and an apparently neat summary can lose an important qualification. I still have to work through the output carefully before I’m prepared to share it with anybody else.

That distinction between assistance and responsibility is important. An AI model can reduce the amount of work involved in something but can and should never be made responsible for whether the finished result is accurate or appropriate. The Methodist Conference’s 2026 report on AI takes a similar position, recognising that AI can assist organisational work while retaining human responsibility and judgement.

My experience with AI-generated images has been rather less settled. I recently used one on my blog because I couldn’t find an existing image which did what I wanted, while creating the image I had in mind myself would probably have taken several days. At the time it seemed like a reasonable use of the technology, but several people challenged that decision and I’m now talking with an artist about other possibilities for illustrating future posts.

I’m still not entirely convinced that my original choice was wrong, and I certainly don’t think the alternative to AI should be expecting artists to give their work away for free. The challenge did make me reconsider how quickly I had narrowed the available options to finding an existing image, creating one myself or generating one. Once generation became the easiest solution, I hadn’t looked particularly hard for other possibilities involving people.

I can give a reasonable instrumentalist account of why I used AI: I had a task which would otherwise have taken several days, and a tool allowed me to achieve it quickly. At the same time, the presence of the tool had already changed the way I viewed the alternatives.

What are we happy to hand over?

There are plenty of tasks where I’m quite happy for technology to reduce the amount of work involved. I don’t feel any particular need to preserve the hours I can spend struggling to turn notes into coherent meeting minutes simply because doing it manually develops character. If AI can produce a draft which I then check carefully, I can use that time somewhere else.

Other choices are less straightforward. Photography, design, writing and editing are skills which develop through practice, and an organisation which routinely generates material because it is quicker may gradually provide fewer opportunities for people to develop those skills. A person who might once have learned to design a poster by actually designing posters can find themselves checking things generated by a machine instead.

That doesn’t give us a rule against automation, because AI can also enable people to do things they previously couldn’t. It can help with language barriers, disability and lack of access to particular kinds of expertise, while a small organisation may be able to produce something it simply couldn’t have afforded before. My own use of it because of ADHD sits within that complexity.

For churches, there is another layer because we already talk a great deal about gifts, participation and formation. If we believe communities should provide opportunities for people to discover and develop what they can contribute, then questions about efficiency have to sit alongside questions about the opportunities we are preserving or removing.

The Methodist Church’s current work on AI reflects some of this uncertainty. The 2026 Conference report recognises AI as potentially valuable for routine and organisational tasks while being much more cautious where its use might replace pastoral care, theological reflection or personal judgement.

I suspect we will have to at least become comfortable with making these decisions rather than looking for a simple Christian position on AI. Asking whether AI is a good or bad tool only takes us so far if AI is simultaneously becoming part of the infrastructure through which we search, write, communicate and work.

That leaves another part of the problem which deserves rather more space than I can give it here. Technologies don’t only change the choices which seem available to us; large technological systems also have material costs which can become remarkably difficult for their end users to see. My own research on electricity has left me with some rather uncomfortable parallels between the development of the National Grid and the growth of the data centres now sitting behind AI. Which I will explore in the next post on this.

Thomas Hughes, Networks of Power
 Methodist Conference 2026 Artificial Intelligence Report
 Methodist Church: Conference debates AI