5 AI-Adjacent Features Already Shipping in Autodesk, Bluebeam, and Esri

Feature 1: Construction IQ in Autodesk Construction Cloud (Forma)

Construction IQ is a generally available risk-prediction system inside Autodesk Construction Cloud, which is now part of Autodesk Forma⁵. It uses machine learning to surface and prioritize cost, schedule, quality, and safety risks across project issues, RFIs, and checklists. Autodesk reports it has been "trusted over 5 million times in the last year"³.

This is the most mature shipping AI in the Autodesk construction stack. It's been in production long enough that Autodesk's customer-trust metric is in the millions, not the hundreds. Construction IQ is a feature within Autodesk Construction Cloud / Forma, not a standalone product— so if you already have an ACC Premium-tier project, you likely have access to it today.

There's a real limit worth knowing. Construction IQ's models are currently optimized for Commercial, Healthcare, Institutional, and Residential project types⁴. Tenant fit-out, light industrial, specialty contractors— pilot with realistic expectations.

• Good for: Mid-project triage of issues that would otherwise be buried in dashboards; weekly risk reviews; superintendent prep meetings
• Falls short on: Project types outside Commercial/Healthcare/Institutional/Residential; transparency on training-data bias (Autodesk's customer base is the corpus); requires Premium-tier ACC on most setups
• Pilot move: Turn it on for one active project that fits the supported types. Assign a single owner to review the weekly risk feed. Compare flagged issues to your actual issue log over 60 days.

The model is real, the metric is real, and the limits are real. If you're running commercial or healthcare work, Construction IQ is probably the lowest-friction AI pilot available to you this quarter.

Feature 2: Autodesk Assistant Across Forma Build, Takeoff, and Docs

Autodesk Assistant is a generally available conversational AI inside Autodesk Construction Cloud and Forma. It answers natural-language questions about project specs, RFIs, submittals, and drawing data— citing the specific source documents— across Forma Build, Takeoff, and Docs⁶.

This is where the "is this just ChatGPT?" question comes up. And the answer is no. Autodesk Assistant is not a frontier model; it's a domain-specific retrieval layer trained on your project data. That's exactly what makes it useful for a busy superintendent looking up spec section 09 91 23 at 6:47 a.m.

Think of it as a source of truth for project documents. Where ChatGPT will confidently invent a specification reference, Autodesk Assistant points you to the specific page in your project's spec book. The citation discipline is the feature.

• Good for: RFI prep, spec lookup, submittal review, jobsite questions where pulling up a 400-page PDF is the bottleneck
• Falls short on: General-purpose analysis (it's sized for AEC project-data lookup, not strategy); quality of citations depends on how clean your document model is in ACC
• Pilot move: Have one project team use Autodesk Assistant as their default lookup tool for two weeks. Track time-to-answer against their old search workflow. Look for the moments when the citation discipline saves a callback.

Construction IQ and Autodesk Assistant live in the construction phase. The next feature lives upstream, in early design.

Feature 3: Autodesk Forma's ML-Powered Environmental Analyses

Autodesk Forma's wind, solar, and microclimate site analyses are powered by machine learning and deliver near-instantaneous environmental predictions during early design⁷. The same analyses traditionally required hours of CFD simulation by a specialist.

Forma's rapid wind analysis turns a workflow that used to require a CFD specialist into something an architect can run before lunch. Microclimate analysis goes further, combining sun, wind, sky exposure, and localized weather data into a perceived-temperature map of a site— useful for outdoor-space programming and climate-resilient design.

This is the surprise feature for many mid-market firms. They associate Forma with a marketing rebrand and miss that there's already-shipping ML inside the site-analysis layer. The workflow shift matters more than the technology label.

• Good for: Early-design feasibility, owner conversations, code/zoning trade-offs, pitching alternative massing options before you've committed engineering hours
• Falls short on: Approximate by design— not a replacement for code-required engineering analyses; requires Forma access (separate from ACC in many subscriptions)
• Pilot move: Run wind and microclimate on the next site your design team is evaluating. Compare to whatever environmental-analysis vendor or workflow you currently use.

Just because it's easy doesn't mean it's good. But when it's both easy and useful, the question becomes whether your design team's workflow can absorb a 10-minute analysis where a 4-hour one used to live.

Autodesk's stack is one part of the picture. Bluebeam— the PDF tool everyone in the field actually opens— has its own AI story.

Feature 4: Bluebeam VisualSearch (and the Imminent Bluebeam Max)

Bluebeam VisualSearch is the AI feature already shipping in Revu 21 today— it finds and counts repeating symbols in PDF drawings to speed quantity takeoffs⁸. Bluebeam Max, a premium subscription tier rolling out globally in early 2026, extends Revu with Anthropic Claude-powered natural-language interaction via Model Context Protocol (MCP). Bluebeam is part of the Nemetschek Group; Bluebeam Max is the premium tier of Bluebeam Revu.

VisualSearch (shipping today)

VisualSearch is best understood as "AI-adjacent." Bluebeam markets it as AI-powered, the underlying mechanic is closer to high-quality pattern matching on PDF vector geometry, and the workflow benefit is real either way. When you're counting door symbols across a 200-page drawing set, the mechanic matters less than the time saved.

• Good for: First-pass takeoffs, MEP fixture counts, door and window schedules, sanity-checking estimator work
• Falls short on: Image quality matters (poor scans degrade results); not a replacement for an estimator's review; the "AI" label is generous
• Pilot move: Use VisualSearch on the next takeoff your estimating team runs. Measure the time saved against their current workflow. Decide if it justifies a standardized estimator process.

Bluebeam Max (rolling out)

Bluebeam Max's pitch isn't a new AI capability. It's a workflow integration: bringing Claude into the PDF tool where the work actually happens⁹. The platform is built on Model Context Protocol, designed to support ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity later¹⁰. In practical terms, the underlying LLM is swappable— so your firm isn't locked into one vendor's model roadmap.

As of May 2026, the Bluebeam Max vendor page still says "Once it's available" for the trial, which tells you regional availability varies. Plan around "rolling out," not "fully GA worldwide."

• Good for: Estimating teams already living inside Revu who want Claude as a sidecar; firms standardizing on the best AI tools for business workflows that don't lock them into one model
• Falls short on: Date-sensitive availability (check your region); pricing not publicly disclosed
• Pilot move: Track Max availability in your region. Pilot it on one estimator's seat first. Measure the workflow friction reduction before you scale licenses.

Bluebeam handles the PDFs your estimating team lives in. Esri handles the imagery your civil and site teams analyze— and the AI story there is older and more mature than most operations leaders realize.

Feature 5: Esri ArcGIS Pretrained Deep Learning Models

Esri ArcGIS ships more than 75 pretrained deep learning models¹¹ that extract features— building footprints, vegetation, vehicles, solar panels, swimming pools— automatically from satellite and aerial imagery. These are generally available with any ArcGIS Pro, ArcGIS Online, or ArcGIS Enterprise license.

If you do civil, site, or infrastructure work, the most production-ready AI in your stack today is the ArcGIS pretrained deep learning model library. It's not the ArcGIS Pro AI Assistant. That product and the four ArcGIS Online assistants— Notebooks, Solutions, Item Details, and Arcade— all remained in public beta as of the February 2026 release¹².

That distinction matters. Most readers expect the AI Assistant to be the answer. It's not, yet. The deep learning library is the mature offering, and the Building Footprint Extraction model alone replaces hours of manual digitizing on 10–40 cm-resolution imagery.

• Good for: Site analysis, asset inventory, change detection, due diligence, vegetation surveys— anywhere your firm currently pays a junior staffer to manually digitize from imagery
• Falls short on: Imagery resolution matters (10–40 cm for many models); regional training data quality varies; some models are US-focused
• Pilot move: Pick one site or asset-inventory task that consumes junior staff time. Run Building Footprint Extraction (or the most relevant model for your work). Benchmark accuracy and time saved against the manual baseline.

Why this is #5, not the AI Assistant: The ArcGIS AI assistants are promising, but as of February 2026, the production-ready AI in ArcGIS is the model library. Don't pilot what's labeled "beta" when you have a 75-model mature library sitting next to it.

Five features, three vendors, every one of them generally available. The harder question is what to do with them.

Why Most AEC AI Pilots Fail (And What to Change)

Most AEC AI pilots fail because firms turn the feature on without changing the workflow around it. Shipping AI features are a learning floor— they reveal what's possible— but ROI comes from redesigning how the work flows, not from the feature itself.

People are the answer, not AI. The feature doesn't do the work. An operator with a workflow that uses the feature does the work. That's why owned-tool pilots beat new-platform pilots almost every time on a 90-day horizon: friction is lower, change management is lighter, and the cost of being wrong is small.

Here's the pilot anatomy that actually produces a learning signal:

1. Single owner. One person on your team owns the pilot end-to-end. Not a committee. Not "everyone."
2. One project. Pick a single active project that fits the feature's supported scope. Resist the urge to test on three.
3. One workflow change. Decide in advance what stops happening because the AI feature now happens. Construction IQ is useless if no one reviews the weekly risk feed.
4. 90-day measurement. Set a fixed window with a measurable AI success framework. Track time, errors, or dollars— pick one.
5. Written go/no-go criteria. Before the pilot starts, write down what "yes, scale this" and "no, kill it" look like. Avoid the hidden costs of AI projects that don't have exit criteria.

A pilot without an owner, a workflow change, and a 90-day measurement plan is not a pilot— it's a license cost.

Once shipped vendor AI is in workflow, custom AI— firm-trained models, internal agents, MCP integrations— becomes the next horizon. That's the strategic ceiling. The floor is the inventory above.

Where to Start (and What Comes After)

Start with one feature, one project, one owner. That's the cheapest, fastest way to learn what AI inside your existing AEC stack can actually do. Once you have a working pilot, the next move is firm-level AI strategy— the part that requires more than vendor features.

Vendor-shipped AI inside the tools you already own is the floor of what's possible, not the ceiling. The firms that build AI advantage in the next 24 months won't be the ones that adopted Construction IQ fastest. They'll be the ones who built a firm-level AI strategy on top of these shipped features and the custom workflows built around them.

Mapping which feature to pilot first is a 30-minute call. Building the firm-level AI strategy on top of these features is the 6-month commitment. Dan Cumberland Labs helps mid-market AEC firms work through both— starting with what to expect from an AI consultant versus building it in-house. Both are true. All of it matters.

FAQ: AI Features in Autodesk Construction, Bluebeam, and Esri

Common questions about AI features in Autodesk Construction Cloud, Bluebeam, and Esri ArcGIS— answered with the current GA status as of May 2026.

What AI features are generally available in Autodesk Construction Cloud right now?

Construction IQ, Autodesk Assistant, AutoSpecs, Automated Symbol Detection, Autotags on Photos, Automated Drawing Extraction, Specifications Tool, Sheets Tool, Bid Forwarding, Financial Data Extraction (TradeTapp), and the Recommendation Engine (BuildingConnected) are all generally available⁶. Construction IQ and Autodesk Assistant are the two most production-mature for mid-market firms.

Is Bluebeam Max shipping in 2026?

Yes— Bluebeam Max is rolling out globally in early 2026 as a premium subscription tier with AI features powered by Anthropic Claude via Model Context Protocol⁹. As of May 2026, vendor language still indicates "Once it's available" on the trial sign-up, which means regional rollout timing varies. Plan around availability in your region, not a single worldwide launch date.

Are Esri's AI Assistants in ArcGIS production-ready?

No. As of the February 2026 release, the Notebooks, Solutions, Item Details, and Arcade assistants in ArcGIS all remained in public beta¹². The mature, generally available AI in ArcGIS is the pretrained deep learning model library— 75+ models for feature extraction¹¹. If you need production-ready ArcGIS AI today, that library is the answer.

Which project types does Autodesk Construction IQ support?

Construction IQ's machine-learning models are currently optimized for Commercial, Healthcare, Institutional, and Residential project types⁴. Specialty contractors, tenant fit-out, and light-industrial work should pilot with that limitation in mind— the model output will be less reliable on project profiles outside the trained types.

Do Autodesk Construction Cloud and Forma require enterprise contracts for AI features?

No. Standard Autodesk subscriptions cover most of the listed AI features. Construction IQ requires a Premium tier on most ACC project setups, and Bluebeam Max requires the new premium subscription tier, but $20M–$100M AEC firms can adopt these features without enterprise-scale agreements. The friction is operational, not contractual.

⚠️ EVERYTHING BELOW IS PIPELINE METADATA — NOT PUBLISHED

References

1. American Society of Civil Engineers, "Architecture, engineering, construction sector slow to adopt AI, survey shows" (December 18, 2025) — https://www.asce.org/publications-and-news/civil-engineering-source/article/2025/12/18/architecture-engineering-construction-sector-slow-to-adapt-ai-survey-shows
2. Bluebeam, "New Bluebeam Report Shows Early AI Adopters in AEC Seeing Significant ROI Despite Uneven Adoption" (October 2025) — https://press.bluebeam.com/2025/10/new-bluebeam-report-shows-early-ai-adopters-in-aec-seeing-significant-roi-despite-uneven-adoption/
3. Autodesk, "Construction IQ — Autodesk Forma" (2026) — https://construction.autodesk.com/tools/construction-iq/
4. Autodesk, "Construction IQ FAQs — Autodesk Help" (2026) — https://help.autodesk.com/cloudhelp/ENU/Docs-Insight/files/construction-IQ/Construction-IQ-FAQs.html
5. Engineering News-Record, "Autodesk Folds Construction Cloud Into Forma, Will Release Geometry-Based AI Assistants" (2025) — https://www.enr.com/articles/61362-autodesk-folds-construction-cloud-into-forma-will-release-geometry-based-ai-assistants
6. Autodesk, "AI for Construction — Autodesk Forma" (2026) — https://construction.autodesk.com/workflows/artificial-intelligence-construction/
7. Autodesk, "Autodesk Forma — Environmental Impact Analysis for Architects" (2026) — https://www.autodesk.com/products/forma/environmental-impact-analysis
8. Bluebeam, "AI and Innovation" (2026) — https://www.bluebeam.com/product/ai-and-innovation/
9. Bluebeam, "Bluebeam Unveils Bluebeam Max, Next-Generation AI-Powered Innovations at Unbound 2025" (October 2025) — https://press.bluebeam.com/2025/10/bluebeam-unveils-bluebeam-max-next-generation-ai-powered-innovations-at-unbound-2025/
10. Bluebeam, "Bluebeam Max — Product Page" (2026) — https://www.bluebeam.com/bluebeam-max/
11. Esri, "Pretrained Deep Learning Models" (2026) — https://www.esri.com/en-us/arcgis/deep-learning-models
12. Esri, "What's New in AI Assistants (February 2026) — ArcGIS Blog" (February 2026) — https://www.esri.com/arcgis-blog/products/arcgis-online/geoai/whats-new-in-ai-assistants-february-2026