The Integration Layer Arrives

Trend Spotlight

The companies making news this week were not building new AI models — they were building the plumbing that connects AI to construction's fragmented data, and investors are pouring capital into the pipes.

Revizto launched the first Model Context Protocol (MCP) server built for AEC: an open-standard connectivity layer that lets ChatGPT, Claude and Microsoft Copilot query live BIM data through a secure, standardised interface, with no custom integrations or proprietary APIs to maintain. The comparison doing the rounds is "USB-C for AI in construction" — apt, because the value sits in the universal port rather than the cable. At the same time, DroneDeploy released Progress AI, which uses vision-language models to identify installed work from drone and 360-degree imagery without needing a BIM model to compare against. Previous progress-tracking tools required that comparison; Progress AI identifies more than 80 trade types directly from visual data, opening the technology to the majority of projects that never carry a full BIM workflow.

Elsewhere, Sledge emerged from beta with an AI-native construction operating system that reads emails, PDFs and bank feeds and generates structured actions across estimating, contracts, RFIs, submittals and financials, and NavigateAI — Eric Wu's venture after Opendoor — launched with $25 million in seed funding to put AI coaching directly on the jobsite via phones and Meta smart glasses. The pattern across all four launches is the same: the bottleneck has shifted from "can AI do this?" to "can AI connect to the data and workflows where the work actually happens?" The model layer is commoditising; the integration layer is where value is accruing.

Venture capital is following the same logic. ConTech funding reached $6.1 billion across 410 rounds in 2025, and 80% of Q3 capital went to post-Series A companies building this kind of infrastructure. Robotics alone attracted $1.36 billion, and speciality trades — historically ignored by software investors — pulled in $750 million in a single quarter. For incumbents, the implication is uncomfortable: data locked in proprietary silos behind custom APIs is now the problem a new generation of companies is solving, and any startup that can plug into that data through an open protocol shrinks the incumbent's moat.

This Week's Headlines

🟢 Revizto ships the first open AI-to-BIM standard

Revizto announced three integrated components on 28 July: an enterprise MCP server, a Model and Object Properties API, and a Developer Portal, unveiled at the Made Right 2026 conference in San Diego [1][2]. Together they let AEC teams connect AI assistants from OpenAI, Anthropic, Google and Microsoft directly to live Revizto project data through the Model Context Protocol — an open standard maintained by the MCP organisation, not a proprietary Revizto format, and already supported by Claude, ChatGPT, Microsoft Copilot, VS Code and Cursor [3]. Data stays inside the Revizto environment; the AI client queries through the server with OAuth 2.0 authentication, addressing concerns raised in Revizto's own Bridging the Gap Report 2026, which found that 96% of organisations report growing worries about data ownership and control, and 39% plan to simplify their technology stacks [2][4].

Why it matters: Revizto is the first AEC vendor to implement MCP. If other major platforms follow, firms gain a standardised way to expose project data to AI tools without vendor lock-in at the integration layer; if Revizto remains the only implementer, MCP stays a single-vendor feature rather than an industry standard. Either way, building custom API integrations to individual AI platforms looks like a weakening bet now that an open standard exists.

🟢 ConTech venture funding hits $6.1bn for 2025, most of it going to scale-ups

Construction technology startups attracted $6.1 billion across 410 announced funding rounds in 2025, according to ConTech Roundup's full-year tally, which uses a stricter definition that excludes proptech and grid tech [7]. The broader "built environment" picture is larger still: Nymbl Ventures reports $4.4 billion in Q3 alone, a 66% year-on-year increase [5][6]. Four independent sources — Construction Dive, Nymbl Ventures, AEC Business and Construction Defect Journal — confirm that 80% of Q3 2025 ConTech funding went to post-Series A companies [5][6][8]. Robotics absorbed $1.36 billion year-to-date through Q3, up 125% on the prior year, and speciality trade technology took $750 million in Q3 alone — more than double any previous quarter [5][6][8].

Why it matters: The sector is maturing and capital is concentrating in companies with proven product-market fit that are scaling. Early-stage founders describe a widening "Series A cliff" between seed funding and the next round; for late-stage companies, capital remains abundant. That bifurcation — scarcity at seed, abundance at scale — is the defining shape of the current ConTech market.

🟡 DroneDeploy's Progress AI tracks 80-plus trades without a BIM model

DroneDeploy's Progress AI uses vision-language models to identify installed work from drone flights and 360-degree walkthroughs, claiming 95% accuracy across more than 80 trade categories, trained on a dataset of 20 trillion square feet of visual site data and 34 million user-generated annotations [9][11]. The technical advance is the removal of the BIM dependency: earlier progress-tracking tools needed a BIM model to compare physical progress against, which limited adoption to large-scale commercial or infrastructure work with mature BIM workflows. Progress AI identifies work directly from imagery instead, extending automated tracking to projects that never carry a full BIM model [10]. The 95% accuracy figure and the dataset-size claim both come from DroneDeploy itself; AEC Magazine corroborates the product's capability but not the specific metrics [10]. A separate figure — 63% drone adoption on site — appears in coverage sourced from Birdi, a competing reality-capture vendor, and should be read with that in mind [16]. Four AI agents are now active on the DroneDeploy platform — Progress AI, Safety AI, Inspection AI and Embodied AI, which drives autonomous capture with ground robots and docked drones [11].

Why it matters: This is a platform play rather than a point solution, and the open question is whether VLM-based tracking becomes a commodity feature across reality-capture tools generally or stays a DroneDeploy advantage built on proprietary data scale. Vendor-sourced accuracy claims of this kind need independent validation before they can be treated as established fact.

🔶 Sledge launches an AI-native construction OS with $20m in beta volume

Sledge announced the launch of its AI-native construction operating system on 30 July, reporting nearly $20 million in project volume under management while still in private beta [15]. The platform aims to run a contractor's entire operation from bid to build to paid — estimating, contracts, RFIs, submittals and financials — with AI reading emails, PDFs and bank feeds to generate structured actions. The $20 million figure is self-reported via press release, with no independent verification available [15]. Sledge is one of several AI-native platforms this year attempting to unify entire construction workflows rather than augmenting a single existing tool.

Why it matters: If the reported beta volume is real, it suggests contractors are willing to run live project financials through an unproven platform — a strong adoption signal. If the figure includes pipeline rather than committed volume, it is closer to marketing than evidence. Self-reported metrics from companies still in beta should be read with that caveat until independently confirmed.

🟡 Eric Wu's NavigateAI puts AI on the jobsite via Meta smart glasses

Eric Wu, who built and took public Opendoor (NASDAQ: OPEN), launched NavigateAI with $25 million in seed funding from Elad Gil, Khosla Ventures and Fifth Wall [12][13]. The product is a field copilot running on phones and Meta smart glasses, offering real-time upskilling, quality control and scoping through vision AI [14]. Named partners include Lennar, deployed enterprise-wide across residential operations, and Tishman Speyer on the commercial side [13]. The company targets the residential construction labour gap, which the Home Builders Institute estimates at 723,000 workers in residential construction specifically [14]. Forbes cites a $225 million valuation that is not corroborated by any other source; the $25 million raise itself is confirmed by GlobeNewswire [13].

Why it matters: Smart glasses put AI assistance directly on a worker's face rather than in their hand — a genuine interface shift from phone-based field tools. Whether construction workers actually wear the hardware on site day to day remains the open question that funding alone cannot answer.

🟢 Speciality trades become ConTech's funding frontier

Speciality trade technology attracted $750 million in Q3 2025, more than double any prior quarter on record [6]. Nymbl Ventures and AEC Business independently identify trades as the emerging funding frontier: the 5.24 million US trade workers represent 63% of the construction workforce across a $6.5 trillion addressable market [6][8]. ConTech investors have historically focused on general-contractor tools — project management, collaboration platforms, field supervision. The pivot toward electricians, plumbers, HVAC and other trades marks a shift in how the market thinks about construction software, since these workers operate in fragmented, owner-operated businesses with thin margins and minimal IT infrastructure.

Why it matters: Tools built for trade contractors need to be radically simpler than GC-oriented platforms. The funding data suggests investors believe the next wave of vertical SaaS winners will emerge from this segment — though whether the product landscape can absorb $750 million a quarter in productive deployment is a separate question from whether investors can deploy the capital.

🟡 Data sovereignty becomes a selling point

Revizto's Bridging the Gap Report 2026 found that 96% of organisations report growing concerns about data ownership and control, and 39% plan to simplify their technology stacks [2][4]. It is a single-sourced, vendor-commissioned statistic, so the numbers should be treated as directional rather than definitive — but the strategic point holds regardless of the exact percentage. As AI tools require access to more project data, firms that control that data gain leverage; architectures where data stays in-house while AI queries flow through a controlled interface turn data sovereignty into a product feature.

Why it matters: This is the logic behind Revizto positioning MCP as a sovereignty solution rather than a convenience feature. Whichever vendors solve "how do I let AI see my data without giving it away" are likely to have an edge in the next procurement cycle.

Data Point of the Week

80% of all ConTech venture capital in Q3 2025 flowed to post-Series A companies. Nymbl Ventures / Construction Dive [5][6]

Three years ago, ConTech funding was broadly distributed across stages, with seed and Series A deals attracting meaningful capital. Now, four independent sources confirm that proven scale-ups absorb the majority of it, and investors widely describe a "Series A cliff" between seed funding and the next round: seed-stage founders report harder fundraising, longer diligence cycles and higher revenue bars. Meanwhile, companies such as Higharc ($95m Series C), DroneDeploy (acquired by Procore for $845m) and Avoca ($1bn valuation) show that late-stage capital is abundant for businesses with proven traction. The practical implication: pre-Series-A companies need a path through revenue and retention metrics rather than vision decks, while post-Series-A companies are finding a market ready to fund their scale.

The Longer View

Will MCP become an industry standard or stay a single-vendor feature?

The Model Context Protocol is new to AEC, and Revizto is currently the only vendor to have implemented it. The protocol itself is mature, with support from major AI clients including Claude, ChatGPT and Microsoft Copilot, but protocol adoption in construction tends to move slowly — Revizto's competitors, including Autodesk, Trimble and Nemetschek, have not publicly committed to it. If they do, the industry gains a standardised AI connectivity layer comparable to what IFC did for file-level interoperability two decades ago. If they don't, MCP stays a Revizto differentiator rather than infrastructure. Worth tracking: which other AEC vendors are evaluating MCP, what a construction-specific MCP roadmap (clash detection, RFIs, submittals, schedule data) would look like, and whether client organisations start requiring MCP compliance in procurement the way some now require IFC compliance.

The vertical-OS thesis and its weaknesses

Sledge and NavigateAI represent a pattern: AI-native platforms that try to unify an entire construction workflow rather than solve one problem well. Sledge covers bid to build to paid; NavigateAI covers scoping to QC to training. Both assume a single AI system can handle enough of the workflow to justify replacing incumbent point solutions. The thesis has surface appeal — contractors are tired of managing a dozen disconnected tools — but the execution risk is real: construction workflows vary by trade, project type, contract structure and geography, and a platform that tries to cover everything risks doing none of it well. Worth watching: how Sledge's beta users actually use the platform, whether NavigateAI's Lennar deployment produces measurable productivity data, and how retention compares between vertical-OS platforms and point solutions.

Vision-language models and the end of BIM dependency

Removing the BIM requirement for progress tracking could open automated visual intelligence to a much broader market: if installed work can be tracked from drone imagery without a BIM baseline, the addressable market expands from projects with mature BIM workflows to essentially any project with drone or 360-degree camera access. OpenSpace, Buildots and other progress-tracking platforms built their workflows around BIM comparison, so if VLM-based tracking works at the claimed accuracy levels, their differentiation narrows — though DroneDeploy's 20-trillion-square-foot training dataset is a genuine moat on model quality, and open-source VLMs are improving quickly. Worth watching: independent accuracy benchmarks across VLM-based and BIM-comparison systems, whether OpenSpace or Buildots develop comparable BIM-free capability, and how quickly the trade-identification catalogue expands beyond its current 80-plus categories.

Sources

[1] AEC Magazine — "Revizto opens project data to external AI platforms via new API and MCP server", https://aecmag.com/news/revizto-opens-project-data-to-external-ai-platforms-via-new-api-and-mcp-server/ — 28 July 2026. 🟡

[2] Morningstar / PR Newswire — "Revizto expands connected project intelligence platform with enterprise AI integrations", https://www.morningstar.com/news/pr-newswire/20260728sf12510/revizto-expands-connected-project-intelligence-platform-with-enterprise-ai-integrations — 28 July 2026. 🔶

[3] Model Context Protocol — "What is the Model Context Protocol?", https://modelcontextprotocol.io/ — n.d. 🟢

[4] AOL / PR Newswire — "Revizto expands connected project intelligence platform with enterprise AI integrations", https://www.aol.com/articles/revizto-expands-connected-project-intelligence-080000000.html — 28 July 2026. 🔶

[5] Construction Dive — "AI, robotics push built environment tech funding to $4.4B in Q3 2025", https://www.constructiondive.com/news/ai-robotics-built-environment-funding-nymbl/805304/ — November 2025. 🟡

[6] Nymbl Ventures — "Q3 2025 ConTech Market Report", https://www.nymblventures.com/post/q3-2025-contech-market-report — October 2025. 🔵

[7] ConTech Roundup / Bhragan Paramanantham — "ConTech Funding Report: 2025", https://contechroundup.substack.com/p/contech-funding-report-2025 — January 2026. 🟡

[8] AEC Business / Aarni Heiskanen — "What Startup Funding Reveals About the Future of Construction Technology", https://aec-business.com/what-startup-funding-reveals-about-the-future-of-construction-technology/ — 2026. 🟡

[9] DroneDeploy — "DroneDeploy Redefines Progress Tracking in Construction with Launch of Progress AI", https://www.dronedeploy.com/blog/dronedeploy-redefines-progress-tracking-in-construction-with-launch-of-progress-ai — July 2026. 🔶

[10] AEC Magazine — "AI automates construction progress tracking", https://aecmag.com/news/ai-automates-construction-progress-tracking/ — July 2026. 🟡

[11] DroneDeploy — "DroneDeploy Surpasses 20 Trillion Square Feet of Visual Site Data, the Largest Visual Dataset in Construction", https://www.dronedeploy.com/blog/dronedeploy-surpasses-20-trillion-square-feet-of-visual-site-data-the-largest-visual-dataset-in-construction — April 2026. 🔶

[12] Forbes / Anna Tong — "Opendoor Co-Founder Eric Wu Launches AI For Construction Venture", https://www.forbes.com/sites/annatong/2026/05/26/opendoor-co-founder-eric-wu-launches-ai-for-construction-venture/ — 26 May 2026. 🟡

[13] GlobeNewswire — "NavigateAI Launches to Build the AI Copilot for the Physical World", https://www.globenewswire.com/news-release/2026/05/26/3301425/0/en/navigateai-launches-to-build-the-ai-copilot-for-the-physical-world.html — 26 May 2026. 🔶

[14] HousingWire — "A start-up with a deep tech pedigree launches AI to train the trades", https://www.housingwire.com/articles/a-start-up-with-a-deep-tech-pedigree-launches-ai-to-train-the-trades — May 2026. 🟡

[15] PRWeb — "Sledge launches Construction AI with nearly $20 million in project volume under management while still in beta", https://www.prweb.com/releases/sledge-launches-construction-ai-with-nearly-20-million-in-project-volume-under-management-while-still-in-beta-302839071.html — 30 July 2026. 🔶

[16] Birdi — "What's the best software for drone construction progress reporting in 2026?", https://www.birdi.io/blog-post/whats-the-best-software-for-drone-construction-progress-reporting-in-2026 — 2026. 🔶