Nemetschek Intelligence Platform: Three Products, One Bet on Openness

Nemetschek Intelligence Platform: Three Products, One Bet on Openness

One Nemetschek Intelligence Platform, Three Disciplines, and a Bet on Openness

Archicad Max, Allplan Max 2027 and Bluebeam Max share a common AI foundation and a public MCP server. The strategy behind Nemetschek Intelligence Platform is sound; the evidence is thinner than the company’s own May release.

Nemetschek released Max editions of three products on 29 October: Archicad Max for architecture, Allplan Max 2027 for engineering, and Bluebeam Max for construction. All three run on the Nemetschek Intelligence Platform, the group’s shared AI and data foundation, and all three follow the same principle the chief AI officer states in the release: AI proposes, automates and analyses, while the professional reviews, decides and remains accountable. That clause is doing more work than it appears to, and it is why this launch is more interesting than a feature list.

The commercial context matters too. Nemetschek reported revenue of 1,191.2 million euros in 2025, the first time it has crossed a billion, at an EBITDA margin of 31.2 per cent with recurring revenue at 92.2 per cent of the total. This is not a vendor experimenting with AI to find a business model. It is a profitable, subscription-based group putting AI into products people already pay for.

Construction is a difficult market for AI for reasons that have nothing to do with models. A single building project runs across an architect’s model, an engineer’s structural calculations, a contractor’s drawings and a client’s approvals, and each discipline carries its own file format, its own review cycle and its own liability. The information that would make an assistant useful is distributed across all of them. That is why vertical software vendors in this market have shipped agentic features more slowly than their counterparts in general enterprise software: the context problem arrives first, and no model solves it.

What actually launched

Archicad Max is a premium subscription from Graphisoft aimed at data- and AI-driven practices running complex projects. It carries an agentic AI Assistant in beta that completes multi-step tasks inside the Archicad model, AI Assistant Spaces that ground answers in a firm’s own standards and project knowledge, and an updated AI Visualizer. It also includes early access to Nemetschek Creator, a browser-based platform for early-stage design, with full access when the commercial version ships in 2027. The distinction between Archicad Max and the standard Archicad subscription matters for buyers: the agentic assistant, the grounded assistant spaces and the updated visualizer sit behind the premium tier, so a practice already paying for Archicad faces an upgrade decision rather than a feature update. The release does not disclose the price difference.

Allplan Max 2027 extends AI across the engineering portfolio, spanning Allplan, SCIA, FRILO, RISA and SDS2, and lists eight named capabilities: Feature Builder, AI BIM Modeler, AI Visualizer, AI Formula Writer, AI BIM Analyst, AI Structural Analyst, AI Drafter and AI Coach. Bluebeam Max, available globally since May 2026, adds Smart Review for catching design discrepancies and scope gaps, Smart Overlay for comparing drawing sets, Magic Markups for automating repetitive markup and takeoffs, and stitching to combine drawing sheets into one navigable view.

The platform argument

The reason to group three products under one announcement is that they share a foundation rather than merely a brand. The release traces the lineage: an AI Layer announced in 2024, the Nemetschek AI Assistant rolled out across brands since 2025, and now agentic, model-aware capabilities on a common platform. The commercial logic is that an investment in one discipline can scale to the others.

That is the argument every roll-up eventually has to make. Nemetschek’s portfolio grew largely by acquisition, and the risk in that model is that each brand develops its own AI separately, at its own cost, producing its own results. A shared platform is how a group avoids paying three times for the same capability, and it is also how it avoids shipping three incompatible assistants to customers who work across disciplines. That is what the Nemetschek Intelligence Platform exists to prevent. The acquisition history makes the work harder than it sounds: GoCanvas arrived in 2024, Firmus AI and Manufacton more recently, and each brought its own codebase, its own customers and its own assumptions about how a model should be structured. Unifying AI across that is a multi-year engineering programme rather than a launch.

Openness by MCP

The most concrete commitment in the release is interoperability, and it is specific enough to check. The Nemetschek AI Assistant uses the Model Context Protocol to interact with the software and to admit third-party AI applications. Archicad Max ships a public Archicad MCP Server in beta, connecting desktop AI applications directly to the Archicad model. Bluebeam Max supports MCP for Anthropic’s Claude, GitHub Copilot CLI and AnythingLLM, turning natural-language prompts into markups and project insights.

There is a second commitment that is less fashionable and more useful. Connected Studio Sessions link Bluebeam markups to the corresponding location in Revit drawing and model views, bridging PDF and BIM regardless of which tool authored the drawing. That is the unglamorous interoperability that decides whether a construction team can use two products together, and it names a competitor’s format rather than a partner’s. There is a specific irony in a CAD vendor publishing an MCP server. These are companies whose commercial position has rested on the proprietary nature of their file formats, because a format a competitor cannot read is a customer who cannot leave. Publishing an open protocol that lets any AI client read and act on the model gives away a degree of that control, and the calculation appears to be that the control was eroding anyway.

The evidence, and a step backwards

Two quantified claims carry the launch. Allplan Max reports early beta users completing selected tasks up to 36 per cent faster, and Bluebeam Max reports more than 2,000 early adopters finding significant productivity gains in preconstruction, design coordination and construction management. Both are company-reported and both hedge in the way honest performance claims usually do: selected tasks, significant gains, no denominator.

Worth noting is that the company’s own May 2026 announcement of Bluebeam Max was more specific than this one. It named two customers using the product, Martin-Harris Construction in Las Vegas and KPFF’s Los Angeles civil office, and described what each was doing with it. The October release keeps the 2,000 figure and drops the names. That is a small regression in evidence, and the names were the more persuasive part.

The numbers behind it

The group’s financial position is what makes the launch credible rather than aspirational. Revenue rose 19.7 per cent to 1,191.2 million euros in 2025, currency-adjusted growth was 22.6 per cent, and EBITDA reached 371.1 million at a 31.2 per cent margin. Subscription and SaaS revenue grew 51.2 per cent to 858.7 million, taking recurring revenue to 92.2 per cent of the total. Annual recurring revenue stood at 1,199.2 million euros.

Worth noting is where that growth came from. The Build segment, which houses Bluebeam, grew 41.3 per cent to 481.3 million euros with EBITDA up 59 per cent, while Design grew 10.4 per cent to 539.8 million. Both figures reflect the GoCanvas acquisition and a temporary effect from Bluebeam’s completion of its subscription transition, which the company itself flags as non-recurring. The Max tier lands in the segment already growing fastest, which is one further reason the Nemetschek Intelligence Platform matters commercially rather than only architecturally.

One footnote on the boilerplate. The release describes more than 4,500 employees and more than 7.5 million users. The audited annual report records 4,061 employees at the end of 2025, and the company’s own May 2026 release said over 4,000 employees and more than 7 million users. The direction is plausible for a growing group and the figures may simply be newer than the last filing, but the release presents them without a date.

What to watch

Three markers. The first is whether the 36 per cent figure survives production use, since beta cohorts are self-selected and the vendor chooses which tasks count. The second is whether the public Archicad MCP Server leaves beta, because a server that stays in beta is a demonstration rather than an interface. The third is whether Nemetschek Creator, promised for 2027, arrives as a browser-based platform competing with the desktop products beside it, or as a companion to them. A fourth is whether the group publishes any cross-discipline result at all, since every figure in this release measures one product in isolation.

Nemetschek Intelligence Platform: Three Products, One Bet on Openness

What it signals

The launch marks a stage where vertical software vendors stop treating AI as a feature to announce and start treating it as a layer to standardise. Nemetschek is doing what a group of this shape should: building one foundation, shipping it across three disciplines, and using open protocols so customers need not choose between its AI and the tools they already run. The bet is that openness wins more accounts than lock-in would. In a market where the buyer is an architecture practice with strong opinions about file formats, that is a reasonable bet to make.

The harder question is whether a shared platform produces consistent answers: an architect asking about a wall assembly and an engineer asking about the same wall pose different questions of the same model, and an assistant grounded in one firm’s standards may not know the other discipline’s. Every claim about the Nemetschek Intelligence Platform here is really a claim about that consistency, and consistency is the part a launch cannot demonstrate.