As an engineer, you probably don’t like the documentation side of things. You should at least see its value, but, in my experience, engineers are usually not fans.
We know that it protects the organisation, supports future work, and helps others understand what we have done. And yet, under time pressure, documentation is often the first thing to be condensed, deferred, or quietly abandoned.
This is not new.
But there is going to be a new pressure to create even more documentation.
Not because management suddenly cares more. Not because standards have become dramatically stricter.
But because the value of documentation has fundamentally changed. And it’s because of AI.
How AI Will Make Documentation More Valuable
Historically, documentation has been hard to use well.
It was often written because it needed to be and then just left. People might refer to meeting minutes to double-check their deliverables. But usually, it would be left until there was an audit or something official like that.
And that meant it was written in a way that was only useful for such things. Which in turn meant that it was hard to use for other things. Things like:

• Understanding why an engineering system was set up the way it is. Ideal for new members of an engineering team.
• Knowing if options had been considered or tried – and if they would work or not.
• Explaining how the system works – to engineers and non-engineers.

This was, and is, a great loss. People could easily make the same mistakes or not understand the system well enough to think of improvements.
AI changes that completely.
With comprehensive documentation of an engineering project (design decisions, options considered, trade studies, constraints, assumptions, experiments tried, and outcomes), it is possible to interrogate all of it quickly.
Not by reading everything line by line, but by asking the question you want answered.
Interrogating Engineering History
Imagine you are looking at an existing engineering system and considering a change.
If you have good documentation, you can now ask an AI system questions like:

• Was this kind of change considered previously?
• What alternatives were explored at the time?
• Was something similar attempted and found to fail?
• What constraints drove the original decision?
• What assumptions were critical—and are they still valid?

The AI does not invent the answers. It mines your own engineering record.
This documentation becomes your operational knowledge.
You won’t repeat the same mistakes and you can better assess new ideas with the knowledge you now have. It’s like you have become the old guy in the company who has seen it all!
Documentation Will Be Demanded More – Because It’s Easier
There is definitely a certain irony here.
The same technology that makes documentation more valuable also makes it easier to produce. Engineers can now:

• Dictate notes instead of typing them
• Generate concise summaries from long discussions and disjointed meetings
• Convert rough thoughts into structured explanations
• Maintain logs with minimal friction

As I mentioned in the previous article on one-pagers, AI can help you turn raw notes into something readable and useful. That capability will remove many of the traditional excuses for under-documentation.
And once that happens, the expectation will shift.
If documentation is easy and highly valuable, it becomes harder to justify not doing it.
For the Global Engineer (and Company)
This matters even more in a global engineering context.
When documentation exists, AI allows it to be repurposed for different audiences.
Language barriers are reduced. Differences in writing style, cultural expectations, and technical depth can be adapted on demand.
That means documentation no longer has to be perfect for everyone. It just has to exist. Making it easier again for an engineer to work anywhere in the world.
It also means an engineering company can be more robust.
Engineering teams change – sometimes when you least expect it.
People get reassigned. Projects ramp down. An entire engineering team gets taken out by food poisoning at a corporate barbecue.
When a new team comes in, good documentation plus AI dramatically reduces the recovery time. New engineers can interrogate the history of the project instead of starting with fragments and assumptions.
Think about all the lunar exploration knowledge that needs to be relearned with the recent efforts to return to the Moon.
Knowledge that can be reused, transferred, explained, and interrogated is far more valuable than knowledge locked inside a few people’s heads. When all past experience within a company is documented and easy to use, it increases the value of the company’s intellectual property by orders of magnitude.
And an engineering firm would be foolish not to demand all knowledge now be documented.
The Shift That’s Coming
Engineers are probably going to be asked to document more than they ever have before.
You might not like it. You might resist. But, as the above shows, the payoff will be real.
The best you can do now is start using AI to make this documentation easier to generate and then be mindful to use AI to access that documentation (and others) in the future.

I know I am not the first person to take on the topic of AI of late. But the conversation on how it could affect engineering in the broadest sense has not, as far as I have witnessed, been explored as much as it could be. In this post, I am going to go over some recent experiences I have had with AI in engineering, and then, from that, talk about what we could expect.

Can AI do engineering?
Recently I have been developing an AI agent to help people like you think of ways they can be better engineers. I have trained it on the knowledge I have documented on best engineering practice and given examples of how to reply in certain cases. 

The agent is called Ingeny. Try it out here.

This is the first version, and it is still evolving. So if you take the time to explore it to see how it can help you with things like your engineering skill development, career progression and working with other engineers, then you will help make it better. I would appreciate you taking the time to help evolve it. By the way, the plan is to keep it free so engineers can always use it as they wish and need.

In the process of training Ingeny, I needed to train it to respond differently depending upon whether questions are about improving engineering expertise or about actually doing engineering. In the former it does not need to provide a warning that it is not a qualified engineer. In the latter, it should let you know that while it has offered as much insight as it can, into say a design for EMC, it is still up to you to do the engineering work. 

This proved to be demanding. AI is not good with that kind of nuance: is it an engineering question or a question about engineering?

If AI can’t easily deal with the subtle difference between doing engineering and talking about engineering, then it’s probably not going to do well with the subtle differences within engineering problems that can have huge effects upon the nature of the optimum outcome. It is also going to have difficulty assessing all the systemic issues present, the best way to apply first principles, or the best way to frame. This is because the AI that seems to reason is, at this time, based on Large Language Models - words - and engineering, as I have noted before, is very visual (it’s often in the “mind’s eye”).

Does this mean engineers are safe?

For now, I can’t see AI understanding the challenges of something like reviewing the landscape to determine (or create) the best approach to building a bridge to cross an expanse of deep water. However, I understand that AI is only getting better. It is therefore only a matter of time until AI can do such things. 

I remain unsure how long that will be.

Until then, it will be the same for engineers as what has been said for writers, medical practitioners and others. The question of whether it will be engineers or AI is the wrong question. The premise is that it will be engineers and AI, so the question then becomes: how will engineers use AI?

And it’s probably going to be pretty good.
Be the engineer you wanted to be
Anecdote time. 
When I was in academia, I would sometimes ask my students who of them wanted to be the type of engineer who understands the fundamentals of theory and first principles, but wants to be more an ideas person who then gets other engineers to make the ideas happen. Everyone would put their hand up. I did this to show my students that the chances of them getting such a job, given the popularity of such a job, is minimal. That means they were all going to have to make the ideas happen as well as coming up with the ideas.
However, AI has probably proved me wrong.
While I think it will be some time until AI can truly do engineering work, I can see it, fairly soon, doing a lot of the grunt work so that we can be the ideas engineer.
How would this work?
Software engineers are showing us how this could play out. They have felt the brunt of AI more than others, because, out of all the engineering disciplines, software engineering is the most language based. And a Large Language Model is ideal for that kind of work.
But while software engineers have been hit the hardest, they have also shown that they are valuable when it comes to the initial idea and providing the right prompt.
That’s what could very well be the case for the rest of us engineers. Consider the following:

• Providing the description for the part you want, and AI then generating the CAD features to produce that part. You can then take a look at the first iteration and describe the changes you want made - then, they would be executed with the speed of AI. If you have been a CAD monkey like me, then you know how much time this could save and the ideas you could generate when you only need to consider the final outcome as opposed to how to generate it.
• Asking AI to generate a circuit board design based on the inputs, outputs, and the available space. Once again, you can review the initial iteration and make suggested changes.
• Using AI to take your initial free body diagram - maybe hand drawn - generate all the force and moment balance equations, solve them for the key parameters you have identified and then optimise for the one parameter you are interested in. You could draw up a number of ideas and have them all analysed near instantly.
• Importing the bill of materials of a product you need to make to an AI system and then asking it to generate the layout of a production line that could make the product. You could then use it to develop details for each work station, estimate the time to complete work, then step back out to see how that affects the overall work flow, iterate, and then do other work stations. With AI to execute your instructions, you could design an optimised production line at an incredible speed.
• ​Have all the parameters set up instantly and to best practice for a CFD simulation. So you need to only focus on analysisng the results, assessing them for validity and then contemplating what changes to implement next.

You can see in the above, that you are automatically like the chief engineer getting other engineers to execute the detailed work for you to review. Imagine also that you can likely talk to the AI system and use peripheral devices to augment your instructions - like when you sketch and use hand gestures to work with colleagues. This is the type of work that many engineers I know would love to do.
The major challenge I see is that we will need to establish how we will train engineers to get to the level where they can be the ones providing the initial instructions. Maybe that is a post for another time.

What will come after this?
There is a chance that one day AI can do our work. I am not one of those people who naively says “AI can never do my job.” But I am not saying I know it can either. So the best thing is to be ready for what could come.

And if AI can one day do all the engineering, including the ingenious stuff, then it has probably also become smarter than us. And If it is that smart, then, like other smart entities (us), it will likely have pets. So work on being adorable to AI so it wants to keep you as a pet and make you happy - not a bad life really.
What are your concerns?
Do you have any thoughts about AI in engineering or any concerns? Share them with me in the comments. And also let me know how you go with Ingeny.
Note
And so you know, while I do get AI to help check my articles, I do write them. Usually on a Sunday night. But I will use AI to often generate my images - if I can’t find a suitable one online - and I sometimes dictate the article content to be cleaned up and written (this one was typed though).