RFIs alone cost Australian projects close to $910,000 each on average. Here's what AI can realistically automate in construction admin in 2026, and what it can't.
.JPG)
Kshitij Dhamala
A mid-tier builder running three sites at once has, at any given moment, a stack of RFIs waiting on answers, a handful of subcontractor sign-offs sitting in someone's inbox, a variation that needs pricing, and a WHS form that has to be filed before Friday. None of that is the job anyone pictured when they got into construction. It's also, according to the closest thing the industry has to a proper study on the subject, one of the most expensive parts of running a construction business in Australia.
Here's the number that should get your attention: researchers who analysed 168 Australian and New Zealand construction projects found that RFIs alone cost the average Australian project close to AU$910,000 in processing costs over its life. Not delays, not disputes, just the administrative cost of asking questions and waiting for answers (Global Construction Review, reporting on research by Dr Ajibade Aibinu and Aconex Data Analytics). A separate, earlier Navigant Consulting study, cited by Procore, put the average cost of answering a single RFI at around AU$1,624, with almost 22% never getting a response at all.
This article isn't another "AI is transforming construction" piece. It's a specific look at where construction admin actually costs Australian builders time and money, in plain terms: what AI can genuinely automate today, what it shouldn't be trusted with, how to decide between off-the-shelf software and something custom, and what a sensible first step actually looks like.
Construction workflow automation, in plain terms, means using software and AI to move information (an RFI, an approval, a compliance form) between the people who need to act on it, automatically, instead of relying on email, memory and manual follow-up. That's the whole idea. The value isn't the technology itself, it's the hours it gives back.
And there are real hours at stake. Payapps' "Building the Future" survey of more than 1,000 construction professionals found Australian firms spend over 10 hours a week on repetitive administrative tasks, roughly three times more than firms that have already digitised those processes. That's a quarter of a working week, for someone whose actual job is running a site or a project, spent on data entry and paperwork instead of building.
Zoom out and the scale of the industry makes that number matter even more. Construction contributed $163 billion in value to the Australian economy in 2022-23, the third largest industry sector in the country, and employs around 1.3 million people (Australian Bureau of Statistics). Labour costs across the sector reached $100 billion that year and were still climbing. On an industry that size, a quarter of a working week lost to admin compounds into a genuinely large number, even before you get to the specific cost of RFIs above.
None of this means builders are being careless. It means the tools most companies are still running on, email threads, shared drives, paper folders, a shared spreadsheet everyone half-trusts, were never built for how many moving parts a modern job actually has.
Ask a project manager where their week disappears and you'll hear some version of the same five things.
An RFI is a formal request for clarification when the drawings, specifications or contract documents don't give a contractor enough to proceed safely or accurately. In Australia these are sometimes called technical queries. Handled well, they're a normal part of construction. Handled in an inbox, they're where the numbers above come from: Navigant's research found average response times of 6.4 to 10 days, and that roughly a tenth of RFIs on any project turn out to be "unjustifiable", meaning they shouldn't have needed asking in the first place, at an average cost of over $170,000 per project just for those.
An RFI that sits too long often turns into a variation: a formal change to scope, cost or program. Variations need their own paper trail and client sign-off, and if the RFI behind it took two weeks to answer, the variation starts two weeks behind before it's even begun.
This is where a lot of the day-to-day friction actually sits, and it's also the workflow Australian construction businesses are most actively searching for help with right now. Every trade on site generates its own stream of requests, sign-offs and questions. Without a system to route and track them, a subcontractor query gets lost in a text thread until someone chases it up, usually after it's already caused a delay.
Programs shift constantly as tasks run early, late, or get held up by weather or a missing delivery. When the master schedule lives in one spreadsheet that only one person updates, everyone downstream is working off information that's already out of date by the time they see it.
Safety paperwork isn't optional, and it shouldn't be treated as an afterthought. Construction remains one of the industries that together account for roughly 80% of Australia's work-related traumatic injury fatalities each year, out of 188 total workplace deaths recorded nationally in 2024 (Safe Work Australia). Standardised, well-kept WHS records aren't just an audit requirement, they're part of how that number comes down over time. Five different problems, one underlying cause: information that has to move between people, and currently doesn't move fast enough.
It's worth being specific here, because a lot of AI-in-construction content overstates what's realistic right now. These are genuine, currently available applications, not speculative ones:
It's easier to see the value in a specific example than a list of features. Here's what a connected RFI workflow looks like once the pieces above are wired together:
And here's the same idea applied to subcontractor approvals, the workflow Australian builders are actively looking for help with:
Neither of these workflows removes the human decision. Both remove the manual chasing, filing and cross-referencing around that decision, which is where the hours in the Payapps survey are actually going.
Trust matters more here than almost anywhere else in this article, so it's worth being blunt.
AI should not be making the actual engineering or design decision an RFI is asking for. It should not be signing off on a variation's cost or scope. It should not be replacing a site manager's judgement on a genuine safety call, and it should never be the last check on a WHS incident report before it's filed. And none of it works reliably if it's bolted onto a business that has no consistent process to automate in the first place: if three site managers currently handle RFIs three different ways, software won't fix that, it will just make the inconsistency happen faster.
Two other things worth being honest about: data quality and security. An AI system is only as useful as the documents and records it's working from; feed it incomplete or outdated drawings and it will draft confidently wrong answers just as fast as correct ones, which is exactly why a human review step before anything gets sent is not optional. And any system handling contract documents, financial approvals or subcontractor data needs a straight answer about where that data lives, who can see it, and what happens to it if you switch providers, before it goes anywhere near a live project.
These terms get used interchangeably in marketing, and that's part of why builders are reasonably sceptical of the whole category. They're not the same thing.
| Layer | What it actually does |
|---|---|
| Traditional software | Records information. A document register or a spreadsheet stores what happened, but someone still has to enter, move and chase every update by hand. |
| Workflow automation | Moves information between systems and people based on fixed rules, without a human manually copying it each time. |
| AI | Can read, classify, summarise, extract and draft within defined boundaries, turning unstructured information (an email, a photo, a voice note) into something structured and useful. |
| AI agents | Can carry out a multi-step sequence, like the RFI example above, under defined permissions and with a human checkpoint built in, rather than requiring a person to trigger every step. |
Most construction businesses already have the first layer. The real opportunity right now is in the second and third, with AI agents being worth considering only once a workflow is well defined enough to hand over in the first place.
This is the fork in the road most builders eventually hit, and the honest answer depends on the business, not on which option sounds more advanced.
An off-the-shelf platform is usually the right call if your admin problems are fairly standard: you need RFI tracking, a shared schedule and a document register, and you're comfortable adapting your process to fit the software rather than the other way round. Construction project management software is a large, established category in its own right, so there's no shortage of options built for exactly this.
A custom system starts to make more sense once any of the following is true: you're running jobs across more than one state with different compliance requirements; your business does more than one thing under one roof and needs a single connected record instead of two disconnected tools; you're already running legacy systems or spreadsheets that a generic platform won't talk to; or you want RFIs, variations, subcontractor approvals and WHS records connected end to end in one place, instead of living in four separate tools that don't share data.
That last point is where AI agents earn their keep over a simple workflow tool: a well-built system can receive a request, check it against the relevant contract clause, classify its urgency, and route it to the right person, without a project manager touching every step in between. That's the difference between software that records what happened and a system that does some of the work.
Not every admin problem deserves equal priority. Based on where the time and money actually go, this is a reasonable order to work through:
| Priority | Workflow | Why it comes first |
|---|---|---|
| 1 | Subcontractor management and approvals | High frequency, day-to-day friction, and a fast win once routing is automated |
| 2 | RFI logging and first-draft responses | Highest cost per instance (around $1,624 each) and slowest average turnaround (6.4 to 10 days) |
| 3 | Variation tracking and sign-off | Directly downstream of slow RFIs; compounds schedule and cost risk if left manual |
| 4 | WHS and compliance record-keeping | Non-negotiable regardless of scale; automate the capture and reporting, never the judgement calls |
Start with whichever row is causing the most friction on your actual jobs right now, not whichever sounds most impressive in a sales pitch.
However you go about this, off-the-shelf or custom, the sequence looks roughly the same:
The most common failure isn't choosing the wrong tool, it's automating a process that was never clearly defined in the first place. Fix the process first, even while it's still manual, then automate it. The second common mistake is trying to automate everything at once instead of proving one workflow works before moving to the next. The third is skipping the human checkpoint because the AI-drafted response looks right often enough to feel safe. It won't always be right, and the moment it isn't, you want a person positioned to catch it before it goes to a client or a regulator.
None of this requires a construction company to become a technology company. It requires being honest about where the RFIs, subcontractor approvals and compliance paperwork are actually going each week, and fixing one of those properly before moving to the next. Beyond Himalaya Tech works with Australian construction businesses on exactly that: mapping where the admin time is really going, then building the system, off-the-shelf or custom, that actually fits how the business works.
If your business is on the real estate side of this same problem rather than construction, here's how real estate agencies are automating their own admin.
Either way, if you want a clear-eyed look at where your own jobs are losing time, book a free AI audit and we'll walk through it with you.
Kshitij Dhamala
AI Strategist & Digital Marketing Specialist
Kshitij is a Computer Engineer and Lead AI Strategist at Beyond Himalaya Tech. He specializes in architecting advanced multi-agent AI systems and driving digital growth through modern search strategies, including Technical SEO, Answer Engine Optimization (AEO), and Generative Engine Optimization (GEO)
Topics
RPA breaks when a process varies; AI agents adapt and escalate exceptions. Here's how Australian businesses should actually choose between them.
UX audit checklist for NDIS, aged care and government software in Australia. What WCAG 2.2 AA actually requires, and what it costs.
