Construction AI: the building safety levy starts on 1 October, and the week's contech money stayed small

Here's the date to put in front of every residential client you've got: 1 October. That's when the building safety levy comes into force in England, and it's a straight charge on building new homes. The client named on the building control application is the one who pays, the local authority collects it, and it bites on new residential developments of ten or more dwellings, or purpose-built student accommodation of thirty or more bedspaces. This was set out again across the trade and legal press on 27 September, ahead of Wednesday's start.

The bit that catches people out is that there's no single national figure. The rate is set by each local authority and varies a lot, from around £12 a square metre of chargeable floorspace at the bottom to just over £100 at the top, with Kensington and Chelsea sitting at £100.35 for schemes on land that hasn't been developed before. Build on previously developed land and a 50 per cent discount applies, provided most of the site meets the brownfield definition. And there's a decent list of exemptions: social and supported housing, homes for registered providers, care homes, NHS hospitals, and any scheme under the ten-dwelling threshold. So a good chunk of the affordable pipeline sits outside it, which is rather the point of it.

Now the part I'd underline, because it's the one that'll sting. An application you submitted before 1 October stays outside the levy. But if that application gets rejected at Gateway 2 and you resubmit after commencement, it can become liable. Given how many schemes are stuck in the Gateway 2 queue, that's not a hypothetical. It's a bit like beating the deadline to tax your car and then having the payment bounce; you thought you were covered, and you're not. What that means for the person carrying a residential scheme is dull and it matters: know your borough's rate, know whether you're in or out, and check that a resubmission won't quietly drag you back over the line. The client pays the levy, but you're the one who'll get asked why nobody saw it coming.

If you follow the funding, this was a quiet week. The construction-tech tally to 28 September came in at about $25m across five startups on the Bricks and Bytes count, which by recent standards is a slow one. The biggest single cheque was €9m for syte, a German firm whose AI reads land feasibility and zoning across more than sixty million German parcels, so a lender or developer can size up a site in minutes rather than weeks (that raise was announced mid-September and landed in this week's roundup). After that it drops off fast: about €5.5m for Arcos, a site and infrastructure security platform, and $4m for SmartBarrel, which does biometric time tracking and geofenced job costing for crews. No British firms in the list at all.

Two things stand out, and neither is the total. The first is where the money pointed, which is upstream, at feasibility and the deal, and at the workforce, rather than at the work on site. That's a pattern, not a one-off. The easy money has always preferred the clean bit before the mud. The second is what wasn't there. None of this week's raises went to what we've started calling the challengers, the broad, AI-first platforms taking on the big incumbent systems the way the app banks took on the high street. These were point tools, each doing one job, and a good point tool is a fine thing to buy. But it's worth being clear-eyed that a takeoff tool or a time-tracking app, however sharp, isn't the same animal as a platform meant to run a chunk of your project. Different purchase, different question.

I'll flag the obvious caveat: these figures are vendor and aggregator-reported, and a quiet week is just a quiet week, not a trend. But it's a useful reminder of how to buy. When the funding noise dies down, you're left judging software the plain way, on whether it does the job in front of you and whether you can walk away with your data if it doesn't.

Put the two together and they point the same way, oddly enough. The levy is another cost that rewards good records, because the schemes that sail through Gateway 2 are the ones whose paperwork is in order, and the ones that get bounced, and risk falling into the levy on resubmission, are usually the ones scrambling to reconstruct what happened. And the funding week says the market is still long on tools that do one narrow thing. So the job for the person actually running projects is to be picky. Buy the thing that produces the record you'll be asked for, at Gateway 2, at Gateway 3, by a client wanting proof, and make sure it does the work rather than just showing you a dashboard.

For your board pack: before 1 October, confirm which of your live residential schemes are in or out of the levy, note the rate for each borough, and check that nothing stuck in the Gateway 2 queue will drag back into it on resubmission.