Hello Project AI enthusiasts,
This week's edition opens with the UK government's new AI planning tools, a concrete, AEC-relevant step that could shave weeks off householder applications and free up officer time for the bigger schemes that matter. Elsewhere, Microsoft has flipped Copilot Cowork to usage-based billing, forcing a harder conversation about what agentic AI actually costs to run. SpaceX has used its fresh IPO currency to buy Cursor for $60B, a coding-market land grab that goes well beyond last week's IPO headlines. KPMG had to pull its own AI report after the citations turned out to be hallucinated, a cautionary tale for anyone treating AI output as research. And RIB has launched an AI-native platform built specifically for civil and infrastructure delivery. Read on for what each of these means for your projects.
📌 Featured This Week
AI Planning Tools to Accelerate Government's 1.5m Homes Target
The UK government unveiled two AI tools on 17 June to speed up England's planning system, part of its pledge to build 1.5 million homes this Parliament. The headline prototype, Augmented Planning Decisions, is being trialled with Barnet, Camden and Dorset councils and aims to halve decision times on routine householder applications, from eight weeks to four, by triaging applications and giving officers a provisional assessment to review before signing off. A second tool, Extract, now live across every English council, converts decades of paper planning records into usable data, saving an estimated 250,000 officer hours a year. Both were built with Google DeepMind, Google Cloud and Faculty, backed by an £8.2 million contract, with a nationwide rollout targeted for 2027 if trials succeed.
This is the most concrete AEC story of the week. Householder applications make up nearly 70% of the roughly 350,000 filed annually, so freeing up officer time here should mean faster movement on the bigger housing and infrastructure decisions. The open question is whether councils trust a Google-built triage layer enough to lean on it at scale.
🔗 Editor’s Picks
This week's must-read stories on AI, project delivery, and infrastructure:
Microsoft Copilot Cowork Hits General Availability, and Switches to Usage-Based Pricing
Microsoft made Copilot Cowork generally available worldwide on 16 June, ending its run through the Frontier early access programme. The agentic system runs long, multi-tool tasks end-to-end across Outlook, Teams, Word and Excel and now bills on a usage basis through Copilot Credits at $0.01 each, layered on top of the existing per-seat licence. The shift to consumption pricing, Microsoft's first major pricing change in nearly two decades, is the real story here: as agentic tools chain together dozens of model calls, flat-rate billing stops making sense, and AEC firms running similar tools should expect the same conversation about cost predictability soon.
SpaceX to Acquire Cursor (Anysphere) for $60B, Days After Record IPO
SpaceX has agreed to buy AI coding startup Anysphere, maker of Cursor, for $60B in an all-stock deal, folding it into the xAI ecosystem under an option agreed back in April. The real story is not about the IPO timing; it's about AI-coding consolidation: Musk now has a direct entry into enterprise developer tooling, going head to head with Anthropic and OpenAI, at a moment when Cursor's own market share has been sliding against Claude Code. For AEC technology buyers, it's a reminder that the coding-assistant layer underneath much of today's construction software tooling is consolidating fast, and vendor lock-in risk is rising with it.
KPMG Pulls Report on AI Usage Due to Apparent Hallucinations
KPMG withdrew its October 2025 report on agentic AI adoption after research group GPTZero found that only five of its 45 citations pointed to real sources, with named organisations including UBS and the NHS disputing claims made about their own AI use. In other words, a professional services firm used AI to write a report about AI, and the AI made things up. For an industry already nervous about contractual and professional-indemnity exposure from AI-assisted deliverables, this is exactly the kind of governance failure that should sharpen review discipline before a similar document goes out under a project's name.
RIB Unify Platform Launches for Construction
RIB has launched RIB Unify, a cloud-native, AI-native platform aimed initially at the UK civil engineering and infrastructure market, built on Microsoft Azure AI Foundry and AKS. It consolidates document management, process management and estimating into a single browser-based environment, with AI built into the architecture from the outset instead of added later. The estimating module links quantities directly to cost items so that changes propagate automatically, aiming to improve estimate defensibility, a detail that matters more to commercial teams than the AI branding around it.
🔨 AI in AEC and Projects
Here are some intriguing reads, specifically for AEC and project delivery professionals:
Neural CAD AI Foundational Models: AEC Magazine on the rise of "neural CAD," foundational models trained to generate and edit 3D and CAD geometry directly, pointing towards a shift from parametric modelling to AI-native design generation for architecture and engineering teams.
AI-Powered Project Management Helps iQIYI Cut Production Time on a New Feature Film: Streaming giant iQIYI used AI-powered project management to compress production schedules on a complex, multidisciplinary delivery, a programme-compression case study that maps directly onto construction and engineering PM challenges.
McCarthy Building Cos. Signs Multiyear AI Deal With Palantir: McCarthy has signed a multiyear, multimillion-dollar agreement with Palantir to build a connected AI operating system spanning design through field execution, the latest sign that enterprise-level AI agreements are gaining ground among large contractors.
mbue Launches AI-Powered Submittals: A new platform uses computer vision to read drawings and specifications and automatically generate compliant product submittals for electrical contractors, aiming to cut a traditionally slow, error-prone paperwork process down to minutes.
🧭 Curated Links
Five stories worth reading before you get back to delivery:
Neura Robotics Raises up to $1.4B at $7B Valuation for Humanoid "Physical AI"
German cognitive-robotics firm Neura Robotics has closed a Series C of up to $1.4B, billed as the largest full-stack robotics round on record, led by Tether with Nvidia, Amazon, Qualcomm, Bosch and Schaeffler among the backers. The capital is earmarked for humanoid "physical AI" and data-collection facilities, reinforcing robots-on-site as one of the few AI themes with proven reader interest.
AT&T Caps Employee AI Spend as "Tokenminimizing" Hits the Enterprise
AT&T has begun throttling some employees' AI access, as agentic tools that call models repeatedly have tripled enterprise AI bills despite falling per-token prices, with Meta and Uber reportedly doing the same. The industry's shift from "tokenmaxxing" to "tokenminimizing" is a live budgeting problem for any firm scaling agentic workflows internally.
A Derbyshire Police Officer Investigated Over Alleged Use of AI to Fabricate Evidence
A Derbyshire police officer is under investigation for allegedly using AI to create evidence across multiple cases, a stark real-world example of AI misuse in a professional and legal setting that sits alongside this week's KPMG story as a governance and integrity warning.
Z.ai Launches GLM-5.2, an Open-Weight Model Matching GPT-5.5 on Coding at a Fraction of the Cost
Chinese lab Z.ai has released GLM-5.2 with a usable 1M-token context and MIT-licensed open weights, matching or beating GPT-5.5 on several long-horizon coding benchmarks at roughly one-sixth the cost. It's another data point in the open-weight price compression squeezing what frontier labs can charge for comparable performance.
AI Not Yet Good Enough to Mark University Essays, Cambridge Research Finds
Cambridge researchers find AI grading tools are not yet reliable enough to mark university essays, tending to reward style over substance. A useful, sober data point on the limits of AI judgement in professional assessment generally.
🔧 Tool of the Week
Google's AI "Information Agents"
Google's new Information Agents, part of AI Mode, let you set up persistent monitoring on topics you care about and get alerted when something changes, working quietly in the background instead of waiting for a query. For project teams, it's a low-effort way to keep tabs on planning policy changes, standards updates or competitor moves without manually re-checking sources every week.
👀 Also This Week
Lucrezia Noli from Google shares insights on AI's rapid evolution, and more [Podcast]
📰 Event of The Week
RICS Global Construction & Infrastructure Conference, 23 June 2026
RICS' annual online conference runs across APAC, EMEA and Americas time zones, covering circular construction, green investment, geothermal energy, smart connectivity and an AI best-practice session. Worth flagging for this audience: Project Flux's own James Garner is speaking on putting AI into practice in construction, covering compliance, value and real-world application.
📊 Hiring: Power BI & Project Data Reporting Specialists
Pace Global is recruiting data reporting talent for major project delivery programmes across London, Manchester and Birmingham.
If you know your way around Power BI and you've worked on the data behind real projects — schedules, cost, performance — they want to talk to you.
£45k–£70k · permanent · project controls
See the roles → https://www.paceglobal.org/hot-jobs
Till next time,
Project Flux
All content reflects our personal views and is not intended as professional advice or to represent any organisation.