The New Paradigm in Project Controls

How AI is being adopted in Project Controls and why it will revolutionise the function forever.

Artificial Intelligence has the potential to transform the Project Controls function and the way in which decisions are made on complex capital projects. Using advanced machine learning and autonomous agents, this technology will go well beyond monthly progress reporting, earned value management and Monte Carlo analysis to dynamically optimise construction schedules, automate complex cost control tasks, and unlock the predictive power of ‘dark data’ to deliver projects more effectively.

By Brent Boden,
December 2025

Executive Summary

The approach to managing large-scale capital projects is currently undergoing a fundamental shift. For decades, the project controls function has depended on collecting historical data and extrapolating trends to predict a project’s future performance. In today’s environment of escalating project complexity, this lag-indicator and trend model can no longer be considered best or even good practice.

This article provides a practical review of AI applications already demonstrating tangible value in project controls, examines AI’s impact on scheduling, cost, and risk management, and highlights the emerging capabilities of key tools.

While these AI tools won’t replace project controllers, they are the catalyst for a fundamental role transformation. Adopting these technologies requires a significant shift in skills, elevating the function from essential progress tracking and status reporting using metrics and KPI’s to a data-driven strategic advisor.

To fully leverage these tools, organisations will need to commit to a Common Data Environment (CDE) and integration layer, update project management and project controls processes, upskill team members, and even rethink organisational structures, roles and governance models.

From Reactive to Predictive

Traditionally, the project controls function has relied on deterministic methodologies to monitor and forecast performance. While established frameworks like Earned Value Management (EVM) and Quantitative Risk Analysis (QRA) indeed offer forecasting capabilities, they operate on fixed logic.

EVM linearly extrapolates future outcomes based on past performance indices (such as CPI and SPI), whereas QRA relies on predefined, static probability distributions. These tools are effective at telling us what should happen if history repeats itself mathematically, but the methods used lack the ability to “learn” from new data, identify non-linear correlations, or adapt to emerging complexity in the way that AI can.

While the project controls function is essential for governance and reporting, it is inherently reactive. By the time a schedule delay or cost overrun is formally reported, the window for low-cost mitigation has often closed.

The Illusion of Control

Despite the universal adoption of sophisticated scheduling software, detailed risk registers, and EVM standards, capital project outcomes have not materially improved. Research by Oxford University’s Bent Flyvbjerg and others demonstrates that projects are overwhelmingly over budget, over time, over and over again.

The presence of control artifacts – such as a P6 schedule or a Monte Carlo simulation – often creates a dangerous “illusion of control.” These tools satisfy governance requirements but fail to prevent the systemic optimism bias and linear thinking that drive failure.

The Uncomfortable Truth

The Project Controls function, in its current form, fails to deliver on its promise of improving project certainty.

Capitalising on the Underutilised Data

The volume of digital data in modern projects is rapidly increasing, driven by the use of Commons Data Environments (such as Oracle Aconex), Building Information Modelling (BIM), remote sensors, photo and video feeds from drones and onsite cameras, and (digital) daily activity reports.

Projects aren’t failing due to a lack of data, but because of human bias and our inability to objectively analyse such vast amounts of information. No team, regardless of size or skill, can manually process the terabytes of structured and unstructured data generated by a modern project to identify forward-looking, predictive insights in real-time.

This analytical gap is a strategic opportunity. Agile organisations that can implement a sound AI strategy will be able to manage risks and project outcomes more effectively, gaining a competitive advantage over peers still reliant on lag indicators.

Demystifying AI in Project Controls

In this context, AI is a practical set of technologies designed to find patterns and make predictions at a scale far beyond human capability. For project controls, three categories of AI are most relevant:

Machine Learning (ML)

This is the core engine of prediction. ML algorithms are trained on vast datasets of historical projects, their past schedules, cost reports, and their outcomes to identify complex patterns that correlate with future success or failure.

Natural Language Processing (NLP)

This technology enables computers to read and understand unstructured text. In construction, this means AI can read thousands of daily logs, technical queries, and emails to identify emerging issues, sentiment, and hidden risks.

Generative AI

This rapidly advancing technology excels at synthesis and summarisation. It can be fed complex, multi-source data, such as cost sheets, schedule fragments, and progress photos, and prompted to create new content, such as drafting the initial narrative for a weekly progress report.

While each of these technologies offers unique value, their convergence is where real transformation occurs. When Machine Learning (predicting delays) is combined with NLP (understanding root causes from daily logs) and Generative AI (creating a recovery plan), we go well beyond the benefits of advanced forecasting. This integration of technologies enables the system to not only identify potential risks but also to propose detailed, actionable mitigation strategies.

Building a New Paradigm

The primary value of AI in project controls is its ability to create a fundamental paradigm shift:

From Reactive Reporting (“What happened?”)
To Predictive Forecasting (“What will happen?”)
And ultimately, to Prescriptive Guidance (“What should we do?”)

This fundamental shift moves the project control function from a data collector reliant on metrics and KPIs to a strategic partner, augmented with advanced, data-driven insights to focus on project optimisation and proactive intervention.

The AI Apps Transforming Project Controls

Project Scheduling & Time Management

In scheduling, AI represents a step change from the reliance on subjective inputs to empirically derived forecasting, where machine learning algorithms analyse project data to predict optimal strategies based on actual performance.

nPlan: This platform uses a sophisticated ML engine trained on a database of hundreds of thousands of past project schedules. By analysing a new schedule against these historical outcomes, it provides a probabilistic forecast for every activity and milestone, identifying risks that human planners and traditional CPM analysis consistently miss. It moves the discussion from a deterministic “finish date” to a more realistic “probability of finishing on time.”
Alice Technologies: This platform leverages “generative scheduling.” Instead of merely checking a manually created schedule, Alice lets planners enter project constraints such as resources, site layout, and procurement lead times. The AI engine then generates and simulates thousands of viable, resource-optimised schedule options. This allows teams to select the most efficient and robust construction methodology before breaking ground.
InEight Schedule: Unlike “black box” tools, InEight uses a “Human-in-the-Loop” approach. It builds a “Knowledge Library” of historical project data. As a planner adds an activity, the AI uses NLP to “read” the intent and queries this library to suggest realistic durations and logic based on actual past performance. It acts as an intelligent guardrail, flagging optimism bias (e.g., “This activity historically takes 12 days, not 8”) while leaving the final decision to the planner.
Oracle Primavera Cloud: While Oracle Primavera Cloud provides the scheduling interface, the predictive capabilities are powered by the separate Construction Intelligence Cloud (CIC) service. CIC connects to the project database (P6 or Primavera Cloud) and mines historical performance data to identify patterns that would otherwise be invisible to the standard CPM algorithm.

Project Cost Control & Estimating

AI improves cost control by automating low-value manual tasks and delivering greater accuracy in high-value forecasting, particularly for the Estimate at Completion (EAC).

Procore Helix: functions as a unified “AI intelligence layer” embedded across the entire Procore platform, capable of analysing diverse project data—from financials and RFIs to drawings and site diaries—to automate workflows and identify real-time insights. It leverages tools like Procore Assist (for natural language querying and reporting) and customisable (no-code) AI Agents that can autonomously handle repetitive multi-step tasks, such as drafting RFIs or managing submittals.
Autodesk Take-off: utilises machine learning to automate the quantification process. This capability significantly reduces manual counting errors and accelerates the estimation workflow, allowing estimators to focus on high-value cost analysis rather than repetitive administrative tasks, all within the unified Autodesk Construction Cloud environment.

Project Risk Management

Traditional risk registers are static, subjective, and notoriously poor at modelling complex, interdependent risks. AI makes risk management dynamic, quantitative, and proactive.

nPlan (Deep Learning): nPlan’s engine quantifies the specific schedule and cost impact of risks by running thousands of simulations based on historical project data. It allows teams to move beyond subjective “High, Medium, Low” ratings to a data-driven view, prioritising the specific risks that truly threaten the completion date.
InEight Risk (Augmented Intelligence): InEight applies its “Knowledge Library” to risk management. When a user drafts a new risk or issue, the AI suggests potential causes and mitigation strategies based on how similar risks were handled in past projects. It ensures that “lessons learned” are not filed away in a forgotten document but are actively surfaced to the risk manager when needed.
Atlassian Jira (Generative AI): Atlassian (Rovo AI) now allows project controllers to use Natural Language Processing to mine this data. The AI instantly reads and synthesises this unstructured text, surfacing hidden trends and emerging risks that would otherwise remain buried in the database.

Project Reporting & Document Control

Here, the application of AI addresses the “dark data” problem: the thousands of documents and reports stored in a Common Data Environment (CDE) that no one has time to read. AI uses NLP and computer vision to unlock the value hidden in this data.

Oracle Aconex: Within the core CDE, AI features are being deployed to enable intelligent search. This will allow project teams to instantly find a specific technical query, contractual clause, or decision across a database of thousands of documents, saving valuable time. It can also automate routine tasks, such as intelligent mail routing and classifying documents, ensuring that information reaches the correct reviewer efficiently.
OpenSpace: This tool uses computer vision, a form of AI, to automatically track and quantify site progress. A team member simply walks the site with a 360° camera, and the OpenSpace platform stitches the photos together, maps them to the project plan/ BIM model, and uses AI to automatically calculate the percentage complete of work packages (e.g., “drywall installation is 42% complete in Area B”). This provides objective, verifiable progress data, removing subjectivity from progress claims.
Generative AI Apps: A project manager can feed a tool like Copilot, Gemini or ChatGPT the raw data—cost reports, schedule extracts, and automated progress data from OpenSpace—and issue a prompt: “Draft the executive summary for the weekly project progress report.” The AI generates the first draft, allowing the PM’s role to shift from authoring to editing and validating.

The Next Frontier: "Agentic AI" for Project Productivity

Beyond the specific tools discussed, the next wave of productivity will come from “Agentic AI.” This refers to using generative AI platforms, like Microsoft Copilot (Studio) and Google AI (Gemini Studio), to create autonomous agents that carry out multi-step tasks on the project controller’s behalf.

This now shifts AI use from analysis to action. An AI tool (such as those mentioned above) analyses a schedule you provide. An AI agent can be instructed: “Every Monday at 6:00 AM, access the Primavera P6 database, run a schedule integrity check, identify all new activities with high float, and email a summary of the three most critical items to the planning manager.”

These agents act as 24/7 digital assistants to automate high-volume processes.

Practical Examples:

Best Practices for Agentic AI Deployment

Maintain a Human-in-the-Loop: The goal is augmentation, not unchecked automation. Agents should flag, summarise, and recommend, with final financial or contractual decisions resting firmly with a qualified human.
Start with High-Volume, Low-Risk Tasks: The best initial candidates are rule-based, repetitive processes (like the invoice pre-check) where the cost of an error is low and easily corrected.
Ensure Audibility: All agent actions must be logged. When an agent flags an invoice, it must “show its work” (e.g., “Flagged: Line item 3 ($10,500) exceeds PO #789 ($10,000)”).

Organisational Readiness: People, Process, & Data

Purchasing an AI tool is not an AI strategy. The technology is an enabler, but its value can only be unlocked through a well-structured and secure data environment and a deep understanding of the impact on people and processes.

1. Data Strategy (The Foundation)

The “garbage in, garbage out” principle is even more pronounced with AI. A successful AI implementation must be built on a foundation of high-quality, trusted data.

Need for Structured Data: This is a non-negotiable prerequisite. The adoption of standardised Work Breakdown Structures (WBS), cost codes, and project templates (e.g., in Primavera) must be enforced throughout the project lifecycle. Without this standardisation, AI models cannot accurately compare performance across projects or learn from historical data.
Need for Clean Data: The CDE (Oracle Aconex) must be enforced as the “single source of truth” for all project documentation. Data silos and offline spreadsheets disrupt the data pipeline on which AI relies.

2. People & Process

The introduction of AI fundamentally changes the role of the project controller.

The Changing Role: The job focus must shift from data collection and report generation to data analysis and strategic interpretation. The controller of the future is a data analyst who can query AI models, interpret probabilistic forecasts, and advise the project manager on the highest-value interventions.
Need for Data Literacy: Teams must be trained on how to use, interpret, and trust AI-driven insights. They must develop the critical thinking skills to “question the black box” and understand the limitations and assumptions behind an AI-generated forecast.

3. Implementation Strategy

A pragmatic, phased approach is necessary to build momentum, demonstrate value, and encourage adoption.

Start Small: Do not attempt to “boil the ocean.” Select a single, high-impact problem to solve first.
Focus on a Pilot Project: Run a pilot on a small, non-critical project or a live tender. For example, “Let’s use an AI tool to improve our schedule risk analysis on one tender” or “Let’s automate invoice processing on one project.”
Measure and Scale: Define clear success metrics for the pilot (e.g., “reduce time spent on invoice processing by 30%” or “identify 15% more critical risks than the manual process”). Use the results of this pilot to build a business case for a broader, phased rollout.

Conclusions & Strategic Recommendations

While the traditional project controls toolkit of detailed project schedules, risk registers, and monthly progress reports has a place in ensuring good governance on projects, despite decades of refinement, it has had no material impact on improving project outcomes.

Add the growing complexity of modern capital projects and the exponential increase in digital data, and it becomes clear that the sheer volume of information is beyond the capacity of human teams to analyse effectively using these “traditional” methods.

The implementation of integrated AI solutions is the mechanism to unlock this data – its advanced analytical and deep learning capabilities have the potential to provide executives and project leaders with new, more targeted insights to inform decision-making and directly improve project outcomes.

The Unavoidable Evolution of the Project Controls Function

This technological shift demands a fundamental restructuring of the Project Controls function. The traditional model—heavily staffed by “data collectors” whose primary output is the compilation of retrospective reports—is obsolete.

As AI automates data aggregation and routine scheduling logic, the need for large teams of schedulers and cost controllers will diminish further. In their place, a leaner, more specialised workforce must emerge, characterised by two distinct profiles: the Strategic Analyst, capable of validating and interpreting the real-world project-specific application of probabilistic data and generative AI recommendations to guide executive decision-making; and the Technical Enabler, skilled in data integration, building custom AI agents and training project-specific models.

Leaders must recognise that this is a significant talent shift and any purchase of AI tools alone must be accompanied by the reskilling and/or reshaping of the project controls function.

How We Can Assist

To help you move beyond the hype and capture the tangible value of AI in project controls, bpma provides the independent expertise required to build a future-ready function. We assist clients in establishing the necessary foundations and scaling their capability through three key phases:

Establish & Implement a Data Stategy

Immediate: Start Now

Mandate the enterprise-wide adoption of standardised project templates and enforce the use of CDE as the single source of truth for all project data and documentation. All other data-related efforts depend on this foundation.

Launch a Focused Pilot Project

Medium-term: 3 – 6 mths

Select one domain (e.g., schedule risk analysis or cost estimating) and pilot one high-potential tool on a live tender or small project. Define clear, measurable success metrics to validate the business case and build internal advocacy.

Develop a "Digital Engineering & AI" Roadmap

Change Management and Capability Building Icon

Longer-term: 1- 3 yrs

Create a formal strategy that integrates your existing Digital Engineering standards and these new AI capabilities. This roadmap must align People, Process, and Data to build a sustainable, data-driven culture that scales these benefits across the entire organisation.

Whether you are looking to establish your initial data baseline or pilot advanced predictive agents, our team is ready to support your evolution. Reach out to learn more about how we can accelerate your transition to predictive project controls function.