Beyond Copilots: Rethinking Decision-Making in Construction with AI

Christian Pallaria, Global Computational Design Manager, AECOM

A few months ago, we examined the emerging role of generative AI and autonomous agents in construction project management, framing AI as a response to increasing project complexity, data fragmentation, and decision-making pressure. At that time, much of the discussion centered on conceptual capabilities and early signals of change.

Since then, rapid advances in AI technologies and applied experimentation have shifted the conversation from possibility to practice. This article revisits the theme to reflect on insights from applied research and early real-world implementation, examining how AI agents perform in live project environments, where they add measurable value, and where current limitations persist. The earlier article was developed in collaboration with Sydney Mudau, whose perspective continues to inform ongoing discussions on the evolving role of AI in project management.

The Core Challenge in Project Management

In every construction project, the majority of project knowledge sits below the surface, embedded in unstructured data. While schedules, budgets, and cost reports can be analysed quantitatively, the bulk of project intelligence resides in emails, documents, meeting minutes, reports, drawings, site records, and informal communications. The volume and fragmentation of this information quickly overwhelm even experienced project managers.

Recent Industry Surveys Highlight the Scale of this Challenge:

If we consider only two of these findings (the 3–4 hours per day spent on email triage and the nearly 18% of time devoted to administrative work) it becomes clear that well over half of a project manager’s time is consumed by non-strategic activities. During this time, project managers are not exercising judgment, managing risk, or leading decision-making; they are acting as information processors.

Given the cost and seniority of project management roles, this raises a fundamental question: does it make sense for organisations to invest in highly skilled professionals, only for the majority of their time to be spent on administrative coordination rather than strategic decision-making?

This is where generative AI becomes potentially transformative. Used correctly, AI can take over systematic, repeatable information-handling tasks (aggregating updates, reading documents, tracking communications), freeing project managers to focus on leadership, foresight, and decisions that genuinely require human judgment.

What the Market Offers Today

Today, several construction technology platforms incorporate AI, typically by integrating large language models (LLMs) to enhance existing capabilities or introduce new AI-driven features. This improves interaction speed, automates repetitive tasks, or simplifies information access.

AI is often positioned as a copilot that enhances established processes rather than as a catalyst for process design. As a result, we are seeing familiar products and workflows augmented by AI-driven automation, delivering incremental efficiency gains while leaving the core operating model intact.

“The future of AI in construction is not about wrapping an LLM around tools that were designed without AI in mind. It’s about rethinking workflows from first principles.”

For example, Oracle Primavera P6 has introduced AI-supported task forecasting and resource allocation within its Gantt-based scheduling framework. While these features improve planning accuracy, they remain rooted in traditional project control and do not fundamentally alter how project intelligence is generated or consumed.

Similarly, Microsoft Project’s Copilot treats AI primarily as an enhancement to pre-existing workflows. Despite its power, it remains largely generic and insufficiently aligned with the fragmented, multi-system reality of construction data.

Oracle Aconex takes a more conservative approach, relying mainly on external analytics and integrations. Procore and Autodesk Construction Cloud go further by embedding AI to automate and streamline operations, but their focus remains primarily on optimizing execution within established workflows rather than rethinking information synthesis and decision-making.

Yet generative AI has the potential to do far more than optimise legacy processes. Its real impact lies in enabling a fundamental shift in how workflows are conceived and how information is managed. Achieving this requires rethinking processes from first principles, rather than retrofitting AI onto tools designed in a pre-AI era.

The Real Opportunity: Data and Its Location

At the core of this opportunity lie two fundamental problems. The first is the sheer volume of unstructured data generated by construction projects. Emails, documents, drawings, reports, site records, and informal communications are difficult to process using traditional technologies and extremely time-consuming for humans to analyse comprehensively.

The second problem is data fragmentation. Project information is scattered across multiple platforms (document management systems, scheduling tools, BIM environments, email clients, collaboration platforms) that do not naturally communicate with one another. Each system holds only a partial view of reality, making it difficult for project teams to maintain a coherent, up-to-date understanding of project status.

Addressing these challenges requires more than isolated AI features. It calls for an intelligence layer capable of connecting disparate data sources, interpreting unstructured information, and maintaining a continuous, contextual understanding of the project.

Some recent approaches are beginning to explore this direction by introducing AI-driven orchestration layers that sit above existing tools rather than replacing them. In these models, AI acts as a connective tissue across platforms, linking schedules, documents, communications, and models into a unified project context. This enables teams to continue using familiar tools while benefiting from a consolidated, intelligence-driven view of the project. Such architectures reduce cognitive load, improve situational awareness, and support more informed decision-making, shifting the role of AI from task automation to project intelligence.

Conclusion

The future of AI in construction is not about wrapping an LLM around tools designed without AI in mind. Real transformation requires rethinking workflows from the ground up, starting with problems that AI is uniquely positioned to solve. By addressing unstructured data and fragmented information systematically, AI can move beyond incremental efficiency gains and reshape how projects are managed. The most promising developments treat AI not as an add-on, but as an enabling layer for better judgment, clearer visibility, and more confident decision-making in an increasingly complex project environment.

The author is also the founder of WisyPlan. The views expressed are personal and do not necessarily reflect those of AECOM.