Bot Wars and Big Bangs: How AI is About to Transform Infrastructure Delivery
Key Takeaways
-
AI has the potential to have a significant impact on infrastructure delivery. These include:
- Contract analysis, claims and reviews.
- Project approvals.
- Design and engineering.
- Site safety
- Adapting to real-world conditions.
- Harmonising standards.
- We also need to consider the environmental and human costs associated with using AI.
What Next?
Are you studying how AI can help (or hinder) the delivery of infrastructure and applying it to your projects?
Message from Russell:
For guest blogs, I like to invite experts to paint visions of the future in transport. This one by Pete Church looks at AI and its application to infrastructure projects.
Pete Church is a transport infrastructure executive with over 30 years’ experience leading complex rail programs in the UK and Australia. His portfolio spans major infrastructure delivery, new rolling stock and train control systems, including leadership roles on the UK’s West Coast Route Modernisation and, in New South Wales, Australia, on More Trains More Services (MTMS), the Transport Access Program (TAP), and Digital Systems. At TfNSW, his signature contribution has been reshaping public sector project delivery through collaborative contracting, inclusive leadership, and a strong customer and community focus.
Pete is a long time amateur programmer, starting from self-taught BASIC on a Sinclair Spectrum over 40 years ago. Today, he is ‘dabbling’ with AI to understand its potential for changing how infrastructure projects are undertaken. It is fair to say that he thinks the impact can be dramatic, enjoy.
AI and Me: A Not-So-Distant Future for Infrastructure
While I’ve been away from the day-to-day workplace, one of the things I’ve done to keep myself occupied is a bit of training and dabbling with AI. Apparently, I’m an ‘early adopter.’ I’ve learned that there’s much more to AI than just ChatGPT. In creating my own 'bots', I’ve been genuinely amazed by what can be done intuitively and quickly and equally frustrated by how daft it can be at times.
Everyone seems to be talking about AI these days, and that’s how I got gently ‘mugged’ into writing this piece - the ravings of a madman - exploring the potential impacts on government, transport, and infrastructure construction.
Bot Wars!
Where to start? As any good movie - or criminal investigator - would say: follow the money! In construction, contracts (certainly in Australia) tend to be thick, verbose, and written in archaic English, a joy for lawyers, a minefield for clients and contractors. But that’s going to change fast.
Even with my rudimentary skills, I can create a GPT and load its knowledge base with the entirety of a contract, plus any relevant documents. I can train it on how to respond and where to pull in supplementary data, schedules, cost breakdowns, etc. and then automate the whole process.
Here’s where it gets interesting.
Imagine you’re a somewhat opportunistic contractor (not that I’ve ever met one!) who has built a bot trained to identify contract vulnerabilities. On site, you encounter something unexpected - how about the ground conditions? You send yourself a quick email, maybe with a photo. That information is instantly analysed against the contract and a claim is auto-generated. The claim letter is perfect, it correctly references the right contract clauses and includes schedule and cost impacts, and it’s all done almost instantly.
Amazing? Absolutely. And it’s not futuristic, it’s here now. The real question is: where does this lead us?
Will clients build their own GPTs to counter claims? Will we see a digital arms race as “bot wars” break out across contract negotiations? Or, more optimistically, will we simplify and standardise contracts, use shared data platforms (maybe blockchain-based), which will lead us to agree claims promptly - and speed up payment flows throughout the supply chain?
I’d like to hope for the latter: more collaboration, quicker payments, and less friction - meaning more time and energy focused on delivering the work.
Big Bang!
What if I told you we could save at least a month on every project start? Sounds mad, right?
Most projects begin with a dead zone, no site activity, while approvals are gained, typically via management plans. These outline processes for the execution of the project. Usually, someone pulls up a previous plan, changes the name, and then spends weeks updating it. The Client team then spends just as long reviewing and commenting before finally approving.
But imagine a GPT pre-loaded with every approved plan from the last decade, cross-referenced against current legislation and procedures. Within minutes, it could produce a site-specific draft for review and submission. And if you knew the reviewer’s preferences, the bot could even tailor the submission accordingly, surely guaranteeing approval in days!
Back to the Drawing Board?
What about AI’s role in design and engineering?
Engineering consultancies have long tried to cut design costs. Experienced engineers aren’t cheap, and many firms have turned to offshore resources with mixed results. But what if “offshore” was just AI running on your desktop?
It’s entirely possible to feed standards, specs, and prior designs into an AI and have it spit out drawings. So will this make engineers redundant?
Unlikely - at least in the short term. AI still produces the occasional bizarre outcome (the right-angled Indian road bridge might not have been AI designed, but things like this do still happen!). In fact, I think we’ll likely need more engineering oversight, not less - experienced professionals who can guide AI, spot errors, and ensure quality.
In the Dirt
Let’s talk about site work, where it really counts.
I’ve been involved in updating the old Dogging & Rigging guide from a 1970s paper relic into a modern, online resource. It was a great cross-industry effort to embed calculators, real-time guidance, and video ‘how-tos.’ But the next step is even better: link it to equipment.
Imagine cranes that take direct measurements, distance, load, wind speed, and use that with the guide to monitor and advise operators in real time. That frees people up to focus on what really matters, being safe around site.
Mismatch between design and real-world conditions is another constant. Site engineers and construction managers often have to make real-time calls based on their experience. But what if they could snap a photo and get instant feedback, with designs updated on the fly, inspection records auto-generated, and compliance and verification logged instantly?
It’s not science fiction. Railways already use train-mounted cameras and pattern-matching AI to detect faults. Drones have been used to track performance and measure excavations for over a decade. Real-time image-to-record workflows and documentation is now just a small step away.
e-Harmony
One last thought.
In Australia, efforts are underway to harmonise standards across jurisdictions, particularly in rail. Every state has its own operational rules, with their own nuances and history, which makes it nightmarish for those freight and passenger operators who span borders.
AI could ingest all existing standards, highlight the inconsistencies between states, and draft a harmonised, unified version in days. Of course, this wouldn’t resolve history, politics or opinion, but it might help the National Transport Commission (NTC) move the conversation along a bit quicker.
So What’s the Cost?
All of this is coming to a project near you, and faster than you’d think. There are oodles of benefits, but let’s not forget the costs either.
Sending a simple email emits around 0.3g of CO₂ (remember that next time you reply with “thanks!”). Add attachments, and that 0.3g grows. Now consider AI: running large models uses way more power. Data centres, which are needed to power these tools, are estimated to consume 1.7 trillion gallons of water globally by 2027. That’s another huge environmental burden.
And let’s not forget the human cost. Just this week, Atlassian announced 150 call centre redundancies; AI could now do the job more effectively. There will be impacts across all sectors, but hopefully, some will be able to plug some of the reported gaps in the construction sector. More positively, maybe AI will help to rebalance work and life, and we’ll get to welcome in the 4-day work week!
So change is coming, and “early adoption” is great, but let’s make sure we don’t ignore the impacts.
Footnote
This was written using a mix of:
- Voice-to-text via ChatGPT (though I started shouting as it kept interrupting me)
- Audio transcription tools (edited heavily to remove all the “so… right… anyway…” filler)
- Nebo handwriting-to-text app on iPad (best of the bunch - though tricky in a dynamic environment – on Sydney’s trains and metro)