How Artificial Intelligence Can Support Delivery: A White Paper 

By: Peter Major – Head of Service Delivery

Introduction 
From a project delivery perspective, artificial intelligence (AI) is increasingly in the spotlight for its capacity to support better business outcomes, and even more generally to provide value to a range of data-based aspects of modern-day life. It is safe to say that the power of AI is undeniable with it very likely to become an increasing part of our work lives and everyday lives. This is tempered in many quarters by concerns with security, data privacy, and the inherent bias of algorithms upon which AI is based.  
While the hype around AI is significant, this shouldn’t preclude thinking about the project delivery capabilities that AI can support and assist us with right now. Exploring these capabilities now will guide the ongoing evolution of AI within projects. With this, and over time, we would expect the use of AI to be fine-tuned and become more robust.  
This white paper explores the advantages and limitations of AI capabilities in supporting project delivery, and identifies a number of long term considerations as AI inevitably integrates further into the projects world.  
How Did We Get Here? 
Formal project management practices, such as Gantt Charts, started to appear early in the 20^th Century. More modern project management is considered to have started in the 1950s with the development of CPM and PERT methods. Since the 1970s, with increasing access to technology to not only provide business solutions, but to also help manage the delivery of projects, technology has become an increasingly critical success factor in the business world.  
The common element of foundational project management practices is data. Data supports the ability to build a schedule, to identify interdependencies, to budget, to report – in fact, most aspects of project management are data related. 
Of course, the internet and the increasingly mobile access to technology solutions has provided the next wave of evolution. The ability to store and process data has continued to grow exponentially. Increased volume and access to data repositories has meant an exponential growth in processing power (a foundational component of AI). With the significant repositories of data generated by the waves of evolution described above, and the ability to quickly access and summarise that data, it’s no surprise that we have reached the tipping point for AI concepts so quickly.   
The Current State of Delivery Success  
Before we get too excited about the amount of data now at our fingertips, the sobering reality of project delivery success trends indicates that we still have some way to go in understanding and applying insights from the data that is available. If we have greater ability to analyse and predict, shouldn’t we be getting better at delivery?  
According to Top Project Management Statistics for 2023: Trends and Insights – Project Management Report, “research by Project Management Institute shows 28% of projects are deemed failures based on them exceeding budget, timeline, or failing to meet goals. Standish Group’s yearly Chaos Report found only 29% of IT projects in 2021 were completed successfully, meaning on time and on budget. The remaining 71% were either failures or challenged, highlighting continued struggle in IT project delivery.” 

From the same site, “according to PricewaterhouseCooper’s annual survey, 20% of the 10,000 projects reviewed delivered less than half of the expected benefits. Gartner found in their research that 51% of projects miss their benefit targets. Another report from KPMG put the failure rate for projects at 50-60%, similar to other analyses.”  

The rates of success and failure are quite consistent. Further, the trend towards increasing delivery success in recent decades has only marginally changed, perhaps due to the increasing influence of agile thinking and delivering in bite-sized chunks. A number of commonly cited reasons for project failure still ultimately relate to data-driven elements (for example inadequate planning or unrealistic schedule). 

Research suggests other insights when it comes to project delivery: 

• The important of people, and relationships, in project delivery. People are not machines, and do not always behave consistently or rationally. If the people aspect of projects is undervalued (for example, if change management is deprioritised or if senior relationships are ignored) then project outcomes are often suboptimal. Other commonly cited reasons for project failure relate to people and relationships (for example poor communication and lack of sponsor support) 

• Projects operate in the real world. External business and environmental factors have impacts. Restructures happen, people change jobs, and governments change. It can be difficult to quantify impacts associated with the real world, let alone to respond to them. Its worth pondering how AI might ever be able to navigate these challenges – perhaps a value of AI is in tempering the optimism of schedules and budgets.  

Key Contributors to Delivery Success 

With a growing portfolio of analysis of project delivery success through PMI, Gartner, and industry specific research, its fair to say that a number of key reasons for delivery success (or failure) are recognised. These reasons apply at a level above the size of a project, whether it is delivered using waterfall or agile, or the level of technology required for the solution.  

Here are only a couple of related links to illustrate the point: 

• 7 Key Things for Large Project Success – Project Management Articles, Webinars, Templates and Jobs 

• What are the Critical Success Factors in Project Management? 

From our review of these reports and a number of websites, some critical delivery contributors may be classified as people (interaction) based, some are process (controls) based, and some are data-based. Lets explore now how AI may help with different types of critical factors, by reviewing a specific example. 

According to (Top Project Management Statistics for 2023: Trends and Insights – Project Management Report), industry research and surveys of project management professionals reveal the leading contributors to project success / failure are:  

1. Level of executive support 

2. Communication and alignment with stakeholders  

3. Clarity of goals and scope 

4. Management of scope (in order to avoid scope creep) 

5. Adequacy of planning and forecasting 

6. Realism of delivery timelines 

7. Capacity and capability of resources assigned 

8. Focus on change management processes  

9. Focus on risk management 

AI and Delivery Success 

AI doesn’t seem to naturally apply to the first 3 dot points identified above (note: this statement is made with caution, as one might imagine that AI can assess Sponsor decision making or even the clarity of the scope of a vendor contract!). However as one progresses through the list, the opportunities for AI to contribute positively, through the collation and analysis of data, appear stronger. AI can significantly enhance project delivery success by improving efficiency, supporting realism of planning, providing data for decision-making, and supporting collaboration throughout the project lifecycle.  

Tangible Ways that AI Can Help With Better Delivery 

Here are several ways AI can generally help: 

1. Automating Routine Tasks 

• Task Management: AI tools can automate repetitive tasks such as scheduling, resource allocation, and status updates, reducing administrative burden and allowing teams to focus on high-value work. 

• Automated Reporting: AI can automatically generate progress reports, summaries, and performance dashboards, ensuring stakeholders stay informed without manual intervention. 

2. Predictive Analytics 

• Risk Identification: AI can analyse historical data and current project performance to predict potential risks, such as delays, cost overruns, or resource shortages. Early detection allows project managers to take proactive measures. 

• Forecasting: AI-powered tools can provide more accurate estimates for project timelines, resource needs, and budgets, based on past project data. 

3. Enhanced Decision-Making 

• Data-Driven Insights: AI can process vast amounts of project data and provide actionable insights, helping project managers make informed decisions. These insights can relate to task prioritisation, resource optimisation, or workflow efficiency. 

• Scenario Simulation: AI can model different scenarios and predict outcomes, allowing teams to explore the impact of decisions before they’re made. 

4. Improving Collaboration 

• AI-Powered Communication Tools: AI-driven communication platforms can streamline interactions between team members, ensuring faster and more efficient collaboration, even across distributed teams. 

• Natural Language Processing (NLP): AI can understand and categorise project-related communications, flagging urgent issues or bottlenecks, and ensuring critical conversations are addressed promptly. 

5. Real-Time Monitoring and Adjustments 

• Performance Tracking: AI can monitor key project metrics in real-time, identifying deviations from the plan and suggesting corrective actions. This helps maintain project alignment with deadlines and goals. 
• Resource Optimisation: AI can adjust resource allocations based on real-time data, ensuring the most efficient use of available resources and minimising downtime or overuse. 

6. Improving Stakeholder Engagement 

• Customisable Dashboards: AI-driven dashboards can provide stakeholders with personalised views of the project’s progress, tailored to their specific needs and concerns. 

• Sentiment Analysis: AI can analyse stakeholder communications and feedback to gauge sentiment and satisfaction, enabling project managers to address concerns early and maintain strong relationships. 

7. Enhancing Quality Control 

• Automated Testing and Quality Assurance: AI can help identify defects in product development or processes more efficiently through automated testing tools, ensuring higher quality deliverables. 

• Continuous Improvement: AI learns from past projects, helping teams refine processes and approaches, leading to continuous improvement in future project delivery. 

8. Resource and Talent Management 

• Skills Matching: AI can match team members’ skills with project needs, ensuring that the right people are assigned to the right tasks for optimal performance. 

• Workload Balancing: AI can analyse team members’ workloads and distribute tasks evenly to prevent burnout and improve overall productivity. 

By leveraging AI tools, project managers can streamline workflows, reduce risks, and ensure projects are delivered on time and within budget. Many would agree that these capabilities are already being established, and can be applied if the desire (and the data) is there.  

Limitations of AI 

Each of the areas listed above has some inherent limitations to be considered against their potential advantages. These broadly fit into the theme of the ‘human element’ of projects. When one considers the current state of project delivery, and related insights, it’s clear that while AI can assist in a number of meaningful, measurable improvements, there are some aspects that may not be a benefit in all cases, and also some elements that AI may not be able to assist at all. Limitations against each area are listed below: 

1. Automating Routine Tasks 

• Automation is reliant on defined standards and processes. Processes are often reliant on human data input or participation, which means that the automation is only as good as the data fed into it. 

2. Predictive Analytics 

• The larger the portfolio of historical data and current project performance, the better forecasts are likely to be, and the better that some potential risks can be identified. However its important to remember that not all risks can be identified or analysed based on historical data. Risk management is an active, forward looking activity. 

3. Enhanced Decision-Making 

• While AI can provide actionable insights, helping project managers make informed decisions, the action of making decisions is still a human activity. If key stakeholders are not available, if there is an entrenched contrary view, or if there is insufficient capacity to support decision making, then a required decision may lie dormant. 

4. Improving Collaboration 

• Again, flagging urgent issues or bottlenecks, and ensuring critical conversations are addressed promptly, is essentially a prompt. If team members are not using tools the same way, or are not sufficiently trained, then collaboration is likely to be impacted. 

5. Real-Time Monitoring and Adjustments 

• Resource Optimisation: AI can monitor key project metrics in real-time, and can adjust resource allocations based on real-time data, but the real world implication of these recommendations may not be practical. If it is recommended that solution build is delayed for 3 months, and the build team is already contracted waiting to start, will they still be paid? What are the impacts to morale? These are considerations that are more challenging for AI to assist with. 

6. Improving Stakeholder Engagement 

• Sentiment Analysis: AI can analyse stakeholder communications and feedback to gauge sentiment and satisfaction. However stakeholder may be wary of technology reviewing communications and may change their tone to suit. Not all stakeholders would be expected to agree to having their communications assessed by AI for tone, effectiveness, or timeliness. On the other hand, there are already tools available that support real time subjective assessment of meeting effectiveness or presentation quality. The privacy implications of sentiment analysis are yet to be worked through. 

7. Enhancing Quality Control 

• While AI can help identify defects in product development or processes, if testing is reduced or compressed then defects may be consciously accepted or ignored. In other words, using AI for testing is not guarantee that testing is thorough. One may argue that AI can help prioritise defects for resolution, by analysing the criticality and source of a defect. AI may also be able to more easily pinpoint the location of a defect. This level of analysis will improve as underlying data sets and test results volumes increase. 

• AI learns from past projects, helping teams refine processes and approaches, leading to continuous improvement in future project delivery. There may be an inclination for new projects to ignore or downplay AI learnings by arguing that ‘this was before my time’ or ‘the learning is no longer relevant’. On the other hand, organisations that systematise the capture and application of learnings (whether AI led or not) are often more mature in delivering projects, as mistakes are less likely to recur. The point here is that human factors and changes in the real world may compromise the acceptability of AI learnings. 

8. Resource and Talent Management 

• AI can analyse team members’ workloads and distribute tasks evenly to prevent burnout and improve overall productivity. However AI will make initial assumptions about productivity that are not always based on reality. Perhaps it is better to argue that AI may assist in responding to changes in productivity, or that over time, AI can learn about effective workloads compared to initially planned workloads. 

Longer Term Considerations 

As the area identified above are explored and become more embedded, they will become more accepted and perhaps even something we don’t always consider consciously. The AI will simply happen behind the scenes. Especially in data-driven environments (and projects) the opportunity for efficiency is clear. Still, it is difficult to predict which areas generate the best value, and indeed that are likely to be other areas of value that are less obvious at this point in time.  

As implied earlier, projects are not simply about data. The human and ‘life’ element of projects cannot be ignored. For example: 

• Will AI help a Sponsor automatically be a more effective Sponsor? No (but AI can provide insights as input to decision making).  

• Will AI identify and engage all key stakeholders?  No (but AI can automate and personalise interaction points).  

• Will AI make sure decisions are timely and accurate? No (but AI can make sure input steps are completed or alerted, and collate a view based on available date) 

In the end, its important to bear in mind at AI is a tool (and an increasingly useful tool) but it doesn’t manage a project. AI doesn’t make decisions, people do. 

Conclusion 

AI has the potential to assist with uplift of a range of delivery success factors (perhaps more likely with process based, and data-based factors). In turn, this should positively impact overall delivery success.  The application of AI for more human-based factors is perhaps less obvious, but part of the excitement of AI is that the possibilities will grow and change over time. AI will likely assist with areas that are unexpected or novel, and it will be interesting to see how this space evolves in the project delivery world to come.