a16z on Revit: AI’s AEC opportunity is review, MEP and rework

The a16z essay title does exactly what a good title should do. Every Building You’ve Ever Been In Was Designed By Software Built in 1997 is provocative, specific and just uncomfortable enough to be true in spirit.

The article, written by Joe Schmidt, David Haber, Caroline Goggins and Zabie Elmgren, argues that the architecture, engineering and construction industry remains a $13 trillion market that technology has largely bypassed. Its core claim is that legacy design tools shape the economics, capacity and error patterns of the built environment.

The essay lands because it describes a reality many AEC professionals recognise. Buildings are coordinated through a chain of architects, engineers, MEP consultants, contractors and specialist trades, each using their own systems, models, documents, emails and uploads. The model is never quite the whole truth. The PDF is never quite current. The calculation may sit outside the authoring tool. The design decision may live in a meeting note. The consequence arrives later as rework, dispute or delay.

The a16z authors put the problem bluntly:

That is the part that matters. The built environment is not short of software. It is short of software that understands the work, connects the work and changes the economics of delivering the work.

a16z traces Revit back to 1997, then to Autodesk’s 2002 acquisition for $133 million. The essay says Revit now produces roughly $3 billion in annual recurring revenue and holds more than 95 per cent market share as the BIM authoring standard. It is taught in architecture and engineering schools. It carries firm libraries, templates and project histories. It is expected by clients and known by staff.

This is why replacing Revit is harder than most disruption stories suggest. The product is not loved in every workflow, but it is deeply embedded. The authors describe it well:

That comparison is helpful. Excel survives because it is flexible, familiar and everywhere. Revit has a similar kind of gravity. Any serious challenger must overcome not only features but also trust, training, libraries, file formats, client expectations and professional risk.

The a16z essay leans on familiar but still brutal numbers. Autodesk and FMI research found that construction professionals spend 35 per cent of their time, more than 14 hours per week, on nonproductive activities such as looking for project information, resolving conflicts and dealing with mistakes and rework.

The essay also cites 85 per cent of construction projects exceeding budgets, three quarters finishing late, North American disputes averaging $60.1 million and nearly 12.5 months, and $177 billion of annual US cost from rework, miscommunication and project data hunting.

Some of these figures are repeated so often that they risk becoming wallpaper. They should not. They describe an industry where the cost of poor coordination is structural.

The most useful part of the a16z essay is its three-lane view of how AI might attack the market. This avoids the lazy assumption that the only route is to build a new Revit and wait for everyone to switch.

The first route is to attack Revit directly by building a cloud-native, AI-enabled BIM authoring platform. That is the hardest path because it requires trust, retraining, feature completeness and migration from entrenched libraries. The second route is to build around Revit by owning workflows it handles poorly, such as design document review. The third route is to automate services work, particularly rule-based MEP and documentation tasks, where capacity constraints are acute but professional oversight, liability clarity and code compliance confidence are essential.

The second and third lanes look more immediately credible. a16z says document review before a bid can take three to six weeks, cost $50,000 to $100,000 per project and still catch only about 30 per cent of issues that become field change orders. It points to LightTable as an example of a company attacking that gap. It also highlights MEP design as a services market where much of the work is rule-based, code-constrained and still handled by large teams working on top of or outside Revit.

The strategic point is sharp: if AI can expand delivery capacity, the value is much larger than a better seat licence.

The a16z authors are saying, "When AI expands capacity rather than just improving tools, the pricing model changes fundamentally."

That sentence should make incumbent software firms uncomfortable. Legacy AEC software is still heavily priced around seats, modules and subscriptions. AI-native workflows may be priced around outputs, savings, speed, risk reduction or change orders avoided. If a tool prevents a £500,000 coordination issue, the value conversation is different from whether another engineer needs another licence.

The a16z essay argues that LLMs and vision models can now parse unstructured BIM metadata, classify it semantically and hand structured inputs to engineering algorithms. That is important because buildings combine geometry with code requirements, occupancy classifications, equipment specifications, spatial relationships, load assumptions, manufacturer data and coordination constraints.

MEP is the obvious pressure point. Data centres, electrification, life sciences buildings, hospitals and complex infrastructure all increase demand for mechanical, electrical and plumbing expertise. The talent pipeline is not elastic enough to absorb that demand through hiring alone. If AI can automate parts of the design and checking workflow under professional supervision, it may allow firms to accept work they would otherwise turn away.

This is where Maureen Ehrenberg’s framing in the RICS AI in Construction report is useful:

That is the balance the sector needs. The opportunity is real. The hype is also real. Trusted and safe practice will decide which AI-native tools become part of delivery and which remain impressive demos.

It is easy to blame Autodesk, and sometimes fair. But buyers have also trained the market to accept fragmented workflows. The sector still tolerates email-driven coordination, manual checking, disconnected calculations and duplicated information because changing the system feels risky. AI native challengers will need more than clever models. They will need commercial models, assurance processes and adoption paths that let cautious organisations move without betting on the project.

That probably means the first winners will not ask clients to throw away Revit. They will sit around it, audit it, enrich it, check it and automate painful services work until the centre of gravity starts to move.

• Do not read the a16z essay as a simple Revit replacement story. The more immediate opportunity may be workflow layers around Revit and AI-assisted services.

• Follow pricing model changes. If tools' prices are against outcomes, avoided change orders or recovered margin, the software budget conversation changes.

• MEP is a critical battleground. It combines high demand, rules-based work, capacity constraints and expensive coordination risk.

• Buyers need assurance alongside innovation. AI native tools must prove accuracy, auditability, professional oversight and contractual clarity.

• Incumbents are vulnerable where workflows are manual. The biggest openings are not always in authoring. They are in checking, coordination, documentation and repeatable design work.

For more analysis on AI-native software, AEC workflows and the future of project delivery, subscribe to the Project Flux newsletter. We follow the tools, business models and delivery signals that deserve attention.

a16z essay

Construction Disconnected report PDF

RICS AI in Construction report

Motif

LightTable

All content reflects our personal views and is not intended as professional advice or to represent any organisation.