Every software vendor you talk to this year has an AI story. Every conference keynote has a demo. And if you're the one responsible for a campus's buildings or a manufacturing shop's drawings, you've probably had the same reaction more than once: that looks impressive, but would it actually work here?

That's a fair question, and it deserves a real answer — not a sales pitch. So, let's separate what's delivering measurable results today from what's still mostly a demo, across both facilities management and engineering/CAD.

Facilities Management: Where AI Is Actually Paying Off

The clearest wins in facilities AI right now aren't glamorous. They're boring, high-frequency tasks that eat staff time without requiring deep expertise — which turns out to be exactly the kind of problem AI is good at solving.

The pattern that works: focus on problems that take up significant staff time, occur frequently enough for automation to save real hours, and have clear success metrics. That is where AI in facilities is proving its value — not in broad “transform your operations” pilots.

What's Real

• Predictive maintenance. Sensor data flagging early signs of equipment wear — a pump vibrating abnormally, for instance — before it fails, so a work order gets triggered proactively instead of after a breakdown.

• Energy optimization. Coordinating HVAC, lighting, and other high-load systems against real occupancy and outdoor conditions, rather than running on fixed schedules — this is consistently cited as one of the fastest-ROI AI applications in facilities because it runs on data most teams already have.

• Troubleshooting assistance. A technician in the field asking a system to instantly pull relevant historical maintenance data and manufacturer specs, instead of digging through binders or calling someone back at the office.

What's Still Mostly Hype

• Fully autonomous “AI runs the building” narratives. The realistic version today is AI recommending an action — an adjusted maintenance schedule, a flagged compliance gap — with a human still approving it.

• Generic AI chatbots bolted onto a facilities workflow with no connection to your actual asset data. If the tool can't see your buildings, your equipment, and your work order history, it's answering questions in a vacuum.

One theme comes up in nearly every credible report on this topic: the biggest obstacle to AI in facilities isn't the AI itself. It's data quality and system integration — if your asset data is inconsistent or your systems don't talk to each other, no AI layer on top will fix that. That's worth sitting with before investing in any AI tool: the sequencing matters. Clean, connected data comes first.

Engineering and CAD: Where AI Is Actually Paying Off

The pattern in engineering software is strikingly similar to facilities. The flashiest AI demos aren't necessarily where the time savings live.

What's Real

• Documentation and support chatbots. Nearly every major CAD platform — including Autodesk's own in-product assistant — now has some version of this: ask a question about how to complete a task, get a direct answer without digging through help menus. It's a modest feature, but it's genuinely useful for reducing time lost to searching documentation.

• Generative design within a single ecosystem. For teams that live inside one platform, defining constraints and letting the software generate optimized geometry options can meaningfully speed up early-stage design work — with the caveat that it only works within that platform's ecosystem.

• Drawing and drafting automation. Newer AI-assisted features can scan an existing drawing and propose block conversions or flag repeated geometry automatically, cutting down repetitive drafting cleanup.

What's Still Mostly Hype

• “AI designs the part for you” as a general-purpose promise. Generative design has real value in specific, well-constrained scenarios — it hasn't replaced the core bottleneck most engineering teams actually have, which is finding and reusing what already exists rather than generating something from scratch.

• Any AI feature marketed as a drop-in replacement for engineering judgment. Every credible use case in production today still keeps a person in the loop reviewing the output.

A useful question to ask about any AI feature: does this save time on a task my team already does constantly and finds tedious — or is this solving a problem we don't actually have? The former is where the ROI shows up. The latter is usually the demo.

The Adoption-Risk Piece Nobody Talks About

Here's what most AI marketing skips entirely: adopting a new AI feature isn't free, even when the feature itself is free with your existing license. Every new tool your team turns on comes with a training curve, a change-management conversation, and — particularly for anything touching compliance data, drawings, or building systems — a question about who's accountable when the AI gets something wrong.

That doesn't mean don't adopt. It means adopt deliberately:

• Start with the single highest-friction, most repetitive task your team already complains about — not the most impressive demo.

• Prove it on that one use case before expanding.

• Make sure whoever approves the AI's output understands it's a recommendation, not a decision.

Where This Leaves You

AI in facilities and engineering software isn't one thing — it's a spectrum from genuinely useful, quietly-shipped features already inside tools you're paying for, to flashy demos that don't yet solve a problem you have. The teams getting real value aren't the ones chasing every new feature. They're the ones asking, for each one: what specific, recurring task does this remove from someone's plate?

Trying to figure out which AI features in your Autodesk or Accruent tools are actually worth turning on? Hagerman & Company can help you cut through the noise — we'll help you identify where AI genuinely fits your team's workflow, and where it's better to wait. Talk to a Hagerman advisor.