5 Bold Predictions for the Construction Industry with Mirco Bianchini

Most predictions about AI in construction promise the same thing: faster outputs. Mirco Bianchini, Senior Product Manager at AECOM, aims somewhere more interesting. An architect turned parametric design specialist, he came on the show with five specific shifts, and each one challenges a default the industry has quietly accepted for years. Not "AI will replace people," but something more useful: the tools, roles, and business models around the work are about to be rearranged.

Prediction 1: The Point-Solution App Store Dies

Construction runs on a separate login for every small problem. One tool for this, another for that, each owning a slice of your data.

Mirco's first prediction is that this model is ending. As agentic tools mature, value stops living in any single feature and moves to orchestrating data across many sources through one entry point.

The tell is where it began elsewhere. The first functions to absorb agentic workflows outside construction were operational ones like finance and HR, not the glamorous front end.

That is the pattern to watch. When the value sits in orchestration, owning more tools stops being a strength and becomes a liability.

Prediction 2: Collaboration Rises, Headcount Does Not Fall

Much of the anxiety about AI shows up as job-loss fear. Mirco thinks that misreads where the constraint actually sits.

Construction rarely suffers from too few people. It suffers from coordination across disciplines, stakeholders, and project phases.

Used well, AI widens that bottleneck instead of cutting staff, taking one discipline's intent and translating it into a version another can act on. The caveat is structural, not cultural. A wrong call here can mean a collapsed slab, so the expert stays in the loop while deterministic checks force the agent toward a verifiably correct answer.

The goal is not to remove the expert. It is to remove the coordination drag around the expert.

Prediction 3: Projects Are Graphs, Not Static Lists

We manage projects as flat lists, yet the real work is a web of dependencies. Mirco's third prediction is that we start treating projects as graphs of related information and finally map how the work actually happens.

This is also where the data conversation gets honest. The industry does not lack data. The unanswered question is how we capture it and turn it into something usable.

That reframes a tired debate. Stop asking for more data, and start structuring the data and the decisions you already produce every day.

Prediction 4: A New Role, the Internal Forward Deployed Engineer

If teams start running fleets of agents, someone has to manage them. Mirco's fourth prediction is a new internal role, effectively an HR manager for AI agents.

This person introduces the technology across departments and orchestrates agents much like a manager handles graduates. You set guardrails, feed the right context, and maintain the system as it drifts.

It is less a brand new profession than the next step for the people who already sit between technology and the business. The work is constant gardening, not one-time setup.

Agents behave like junior staff, so orchestrating and maintaining them becomes a real job, not a background task.

Prediction 5: The Architectural Flip

The last prediction is a software architecture flip. Construction firms have long outsourced workflows to SaaS platforms like Procore and Autodesk, for good reasons.

As internal teams gain the ability to build what they need, the per-seat model comes under pressure, because no one charges an AI agent for a seat.

SaaS and startups still matter, and most internal builds are prototypes rather than products. The signal to watch is a prototype that keeps getting reused across projects, because that repetition is where real workflow value hides.

Key Takeaways

• Audit your tool stack for places where one orchestration layer could replace several point solutions.
• Treat coordination, not headcount, as the real constraint, and use AI to translate intent across disciplines.
• Keep a qualified expert in the loop on high-risk outputs, with deterministic checks the agent must pass.
• Reframe your data effort around capture and structure, and treat projects as graphs rather than flat lists.
• Give someone clear ownership for orchestrating and maintaining the agents your team depends on.
• Track prototypes that keep getting reused, and productise only what repetition proves is valuable.