Construction AI: the scaffolder now talks to the form, and the agents reach the workforce planner

On 19 August, Smart Scaffolder launched a voice assistant that lets a scaffolder complete site records by talking to their phone. The Smart Forms AI Assistant listens, works out the context and populates the form, and the company says it cuts many routine records to around half the time. The first release covers the forms every scaffold contractor lives with: incident reports, harness and lanyard inspections, tie test records, toolbox talks, site audits, vehicle checks, permits.

Here's the detail that made me sit up. If a mandatory field is missing, or the AI isn't confident it's understood something, it highlights that field for review rather than filling it in and hoping. The competent person keeps full control before anything is submitted. That's the right way round. A tie test record isn't a blog post, it's a safety document that someone signs their name against, and the value of an assistant here is speed on the boring bits, not a machine deciding what "adequately tied" means. The half-time figure is Smart Scaffolder's own, so treat it as vendor-reported until a contractor tells you what it does on a wet Tuesday. But the design principle behind it is sound.

What that means on site is simple. The bloke who's been putting off his vehicle checks and his toolbox talk paperwork until the end of the shift can now knock them out by talking while he's still stood by the scaffold, gloves on. The forms are aligned to NASC guidance and the usual good practice, so nothing about the compliance changes, only the friction of getting it recorded. Smart Scaffolder says this is stage one, with an AI assistant for its Smart Estimator coming next year that would let an estimator describe a scaffold in plain English and get an initial model back to price. They'll be showing it at ScaffEx on 10 and 11 September at Manchester Central. That's what it's about: giving the person doing the records their time back, not replacing their judgement.

On 29 July, Bridgit, whose workforce-planning software is used by a big chunk of the top US contractors, released two things: a set of AI agents, and a Bridgit MCP. The MCP is the bit worth understanding. Model Context Protocol is an open standard, and what a Bridgit MCP does is let live workforce data flow into whichever AI assistant a contractor already uses, ChatGPT, Claude, Copilot or Gemini, without anyone exporting a spreadsheet or logging into another system. Ask your usual assistant who's available for a job starting Monday, and it answers from the real roster. The agents go a step further and build project teams, suggest staffing and draft workforce reports. The date's a few weeks back, but I'm flagging it now because the MCP piece is the significant part, and it's only just started showing up in the UK trade press.

The guardrail is the same one Smart Scaffolder built in. A planner reviews and approves before any change is made. Think of it like a very fast junior who drafts the resourcing plan and leaves it on your desk to sign, rather than one who reshuffles the crews and tells you afterwards. The comparison only goes so far, but the point is the human stays the decision-maker. I'm not sure every "agentic" launch this year has been that disciplined, so credit where it's due.

This lands on a real UK problem. On 14 August, Nadeem Mirza, resource and workforce planning director at Sir Robert McAlpine, argued in Construction Management that the skills shortage is a structural crisis, and that AI-gathered data, on people's strengths, competencies, even personality types, is how you place the right person on the right project and keep them from drifting off to a competitor. His line was that no one should leave the planning to the machines, they should use them to generate richer data that a human then interprets. Same instinct, different end of the job. For the site manager or the resourcing lead, the promise is less time keeping a spreadsheet alive and more time on the people it describes.

Put the two launches next to each other and the shape is obvious. One's a scaffolder finishing a tie-test record by talking. The other's a planner asking a chat assistant who's free next week. Neither is a robot, neither is a moonshot, and both take a slice of admin that quietly eats site hours and hand it back. The Building Safety Regulator's own April-to-June 2026 figures show approval rates climbing (89% for new higher-risk buildings, 85% for external remediation) but medians still running 22 to 35 weeks, so the record-keeping burden isn't going anywhere. If AI can shave the time it takes to keep those records straight, without touching who's accountable for them, that's a real gain rather than a demo.

The discipline that makes it safe is the same in both products, and it's worth writing on the wall: the machine drafts, a competent, named person signs. Smart Scaffolder flags the field it isn't sure about. Bridgit waits for approval. Buy on that basis and you get the time back without giving up the control.

Today's action: Next time a vendor demos an AI that fills in a form or a plan, ask one question before anything else: what does it do when it isn't sure? If the answer is "it flags it for a human," keep listening. If it's "it just fills it in," walk.