How to build AI into your projects

Praised for its innovation, the Intelligent Project Prediction (IPP) initiative – a collaboration between MIGSO-PCUBED, greyfly.ai and DHL – won two APM awards in 2024. What made the cutting-edge project such a success?

James Martin-Young, Head of Digital and Data at MIGSO-PCUBED, began his journey into artificial intelligence (AI) in 2023, when DHL Supply Chain (UK) decided it was time to rethink its approach to project management, as it was struggling to deliver its projects on time and to budget. The logistics firm asked MIGSO-PCUBED to form and lead a team of data and project specialists tasked with finding ways to reduce the gap between DHL’s plans and final delivery.

In 2024, the project went on to scoop both the APM Technology Project of the Year Award and the APM Innovation in Project Management Award – and it was able to demonstrate the potential value of AI to the project profession along the way. The tool of choice: the IPP platform, which uses advanced analytics and machine learning to identify underlying drivers of a project’s success, predict its outcomes and improve strategic risk management and decision-making.

Insights to make better calls

“If you can identify projects at risk and highlight areas for improvement, you can reduce the risk of schedule and budget over-runs,” says Martin-Young. “Meanwhile, driving more mature project data gives project teams the insights to make better calls.

“We definitely see this becoming part of the digital PMO [project management office], providing a prediction service back to project executives, so they can use those insights to accelerate project performance across the portfolio. It can also help them harness wider data points that deliver project success – for example, bringing in greater data around benefits realisation and sustainability.”

The DHL project involved building a single integrated team that shared common goals and processes. MIGSO-PCUBED’s delivery manager and experienced data, technical and change experts worked alongside representatives from DHL’s PMO and project wings. MIGSO-PCUBED also partnered with AI and project portfolio management specialist greyfly.ai. They shared their knowledge and experience of AI’s project portfolio management capabilities to build a common framework.

Fantastic results

The project lasted 18 months and was completed in January 2024. While no minimum requirements or specific outcomes had been specified for the project, it doesn’t take computer hyper-intelligence to process the benefits. The level of data assessed as “poor quality” dropped from 78% to just 9%. Forecasting of a project’s overspend reached 95% accuracy.

This information helped the team reduce the proportion of projects exceeding budgetary targets from 33% to a mere 6%. Meanwhile, the average budgetary overspend dropped from 74% to 33%. There were other knock-on improvements. Reports got better, as did the clarity of communication. Decision-making became more confident and focused. What’s not to like?

Five key lessons of AI adoption from IPP

1. Data quality is everything

If you want reliable AI predictions, you need to put data quality at the heart of the solution. At the start of the DHL project, data quality had to be improved. Driving those improvements took relentless focus and collaboration with project executives and the PMO.

2. Change management is critical for AI adoptions

AI introduces a new data-led way of working, and with that comes scepticism. The project team learned that AI adoption isn’t just about the tool; it’s about helping people trust the outputs. Directed communication, transparency and tailored training reinforced by solution design were key to embedding insights into decision-making processes.

3. Executive sponsorship makes the difference

Having strong, consistent sponsorship was critical to success. Sponsors play a pivotal role in building momentum, navigating ambiguity, combating scepticism and bringing users, senior stakeholders and anyone else affected together on the journey. Without them, adoption stalls.

4. Project management processes may need to evolve

Data and insights quality is driven by the project method, processes and systems. Some data-quality issues, and ways the team wished to present and garner insights, were adversely affected by existing processes. Plans to iterate processes were identified at the start and throughout implementation. In order to develop the solution, early planning of proposed changes, supported by understanding any restrictions and implications, was critical.

5. Insights are useless unless they drive action

Generating insights isn’t necessarily the hardest part – translating them into meaningful action is. Obtaining stakeholder support of the implications and determining actions required the team to adapt governance processes. This allowed teams to act on the AI outputs and solve persistent issues, which led to measurable ROI and improved project performance.