We can’t afford to ignore AI when it comes to planning processes
Waikiki beach in Honolulu, Hawaii. In December 2025, Honolulu launched its CivCheck AI tool to guide applicants through the city's specific application process.
Late last year, Jeff Bezos offered the city of Miami a piece of advice: use AI to review building permit applications and deliver yes-or-no decisions in seconds.
It was a quick aside in a wide-ranging conversation about AI but for people active in the planning world it struck an alarming chord. Bezos was signalling that AI may soon have the ability to decide on planning applications in Miami and, by implication, everywhere else as well. Coming from Bezos, we'd do well to take note.
The world's planning systems are all quite different from each other and they defy direct comparison.
A building permit in the United States, a regulatory submission in Singapore, and an Irish planning application do not operate within the same legal framework, but they do share a common feature: they require drawings, documents and proposals to be tested against rules, policies and procedures.
Some of them, like the rules-based system that operates in Singapore, lend themselves easily to review by machine. The Irish system, which is closely related to the one that operates in Britain, presents more of a challenge.
But despite the differences, planning authorities around the world are introducing AI-driven tools to speed up their approval processes.
In December 2025, Honolulu launched its CivCheck tool to guide applicants through the city's specific application process, helping them assemble complete and comprehensive applications before submitting them for review.
The automated prescreening is reported to have sped up the review process significantly.
In Los Angeles, where last year's wildfires destroyed whole neighbourhoods, a similar tool called Archistar AI PreCheck is being used to speed up the massive rebuilding programme.
In addition to checking applications for completeness, PreCheck also looks at how proposals adhere to zoning regulations before applications enter the system.
Archistar is, as it happens, an Australian company. And Australia turns out to be unusually advanced in this area: New South Wales has been running a state-funded AI Solutions Panel across 16 of its local councils since 2024, with significantly reduced approval times reported for new homes.
Similar initiatives are being trialled in other parts of the country.
Tech-forward Singapore no longer simply checks that an application is complete: it has begun to automate parts of the technical review itself, verifying, for instance, that floor area requirements have been met before a proposal moves up the ladder.
Singapore's approach is more rules-based than purely AI-driven, but the direction of travel is the same — more of the work, more of the time, being done by machine.
Closer to home, the UK, which has a planning system similar to our own, has been moving quickly too. In June 2025, the British government launched a tool called Extract which turns the great mess of historic planning information — handwritten notes, scanned maps, blurry PDFs — into clean digital data.
Its scope is narrow: it doesn't decide anything, it simply unlocks the data that decisions are based on. It sounds small but it's a significant cog in the wheel and directly relevant to our own context, where decades of similarly trapped information are still sitting in council filing cabinets.
In November 2025, the UK further announced it was in the market for what it calls Augmented Decision Making software — AI tools intended to help planners make decisions on domestic scale applications. The intent is to cut decision times from upwards of eight weeks to around four.
Most authorities are keen to insist that their AI tools are not intended to replace human decision-makers, only to speed up the dull bits. But the line between speeding up the dull bits and shaping the actual decision is being quietly crossed. Seen in that light, Bezos's quip takes on another significance.
Which brings us home to Ireland, where AI has so far made very little impression on the planning process. For all the recent digitisation, the system remains quite analogue.
It's true that our system faces a particular challenge when applying new technologies. Our development plans rarely express preferences in clear, mathematical, computable terms.
Decisions remain matters of judgement expressed in language that has little of the logic or consistency an AI model would need.
Proposed developments are routinely described as "contrary to the proper planning and sustainable development of the area," which may well be true, but is not a phrase that lends itself to computation. (A non-Irish colleague once remarked that our system was "well-meaning, but like wading through treacle.")
So while our particular system presents a challenge to technological enhancement, there may — just may — be a way forward.
The Planning and Development Act 2024 introduced the concept of Urban Development Zones (UDZ) — designated areas intended for large-scale, plan-led development within a structured framework.
The form these UDZs will take remains to be seen, but if the Department of Housing moves in the direction of jurisdictions like New York or Amsterdam, they could herald a shift where rules are stated in terms a computer can use: maximum heights set in metres rather than "appropriate to context"; density ranges stated numerically rather than "in keeping with the character of the area".
It would not remove the element of human judgement from the system, but it would expose at least one part of it to the kind of structure required to let AI speed up the process.
Even if AI tools were brought in alongside well-considered UDZs, they wouldn't remove all our development bottlenecks. There are other issues besides plan approval slowing the system down.
There’s also a new regulatory framework heading our way from Brussels which will shape whatever AI-assisted planning system we devise. The EU AI Act — the first comprehensive piece of AI legislation anywhere in the world — begins to take effect in August 2026.
It’s a complex piece of legislation and we've yet to see how all of this will play out, but in all likelihood the Irish planning regime will not be as free to adopt algorithmic solutions to knotty planning problems as our Australian and North American colleagues currently appear to be.
And quite apart from the possible effects of EU regulation, there are deeper questions about bias in systems trained on past decisions, and about the difficulty of meaningfully challenging an opaque algorithm — questions which will hopefully be the topics of future articles.
But none of this should stop us making a start on our AI in planning journey. Bezos's Miami quip should be taken as a wake-up call. AI will eventually change our planning system whether we like it or not.
If we engage quickly, we'll have some control over the shape it takes. If we dither, changes will be imposed on us from elsewhere.
AI won't slow down to match the pace of Irish public discourse. We need to up the tempo, even if, in the end, we decide we don't want AI involved in our planning decisions at all. One way or the other, we have to engage.
- Gary Miley is an architect based in Killaloe, Co Clare
CONNECT WITH US TODAY
Be the first to know the latest news and updates
