Delivering AI projects:
structure, control and governance are imperative for ROI

There are several ways of looking at the intersecting worlds of AI and project management:

How will AI affect project management? How will AI affect project managers? How will project management affect AI?

Having an AI strategy is critical for organizations, as the potential risks associated with it become increasingly clear. According to international risk and insurance business, Aon: “The growing use of AI is also transforming the threat landscape, enabling more sophisticated attacks and increasing the speed and scale of cyber incidents. At the same time, evolving cyber and AI regulations are pressuring organizations to adapt to new disclosure rules and strengthen compliance to avoid fines and reputational damage.”

But beyond managing risk, there is also the question of delivering value and return on investment (ROI) from introducing AI in organizations.

The impact of AI on project managers/project management

Despite the current hype around AI – and the uncertainty this has created across many so-called “white collar” professions, whose activities comprise mainly “knowledge work” – the human professional remains an essential, component part.

Writing in Consultancy.co.uk, Ruth Corcoran-Henry – Client Delivery Director at Skarbek and Senior Leader in its Portfolio and Programme Management practice – is adamant about what she terms the “humanology” necessary in projects:

“It is people who drive projects and the soft skills…that create the cornerstone for effective project delivery,” she said.

Critical human elements that contribute to this, Corcoran-Henry explains, include:

• Influencing without formal authority
• Resolving conflicting priorities and objectives within cross-functional teams
• Mastery in active listening
• Diplomatic dialogue
• Empathy and issue resolution
• Shifting opposing views into collaborative problem-solving

And while she does not dismiss AI within project management – describing new technology as “an ever-advancing enabler that strengthens how project managers deliver” – she insists that its value should come from being “an assistant that amplifies human capability, not as a replacement for creative problem solving, collaborative decision making or leadership judgement”.

So, for project managers and project management, AI offers new options to improve governance, control, and reduce administrative tasks, while combining the technology with human intervention.

For example, using AI can cut down document drafting time from days to hours, allowing project managers to reallocate valuable time to human-focused activities. AI tools may also offer early alerts to project elements going wrong, enabling managers to put things back on track before getting out of control.

Using AI as a way of capturing risk management data from projects is a valuable use, emphasized by Larry English in Forbes. For example, project documentation – including lessons learned (a key element in PRINCE2 Project Management) – is, English adds, a “goldmine of institutional knowledge” to “guide new projects” and “assess risks”. However, most of it, he claims, “goes unused”.

Instead, he suggests that “AI in project management can transform static documentation into real-time, actionable insights that help project managers expertly navigate risks and avoid repeating past mistakes”.

In fact, PRINCE2 Project Management lends itself well to working with AI, and the specific synergies between AI tools and this project management method – and how project managers can benefit from becoming familiar with them – are explored in the free to download PeopleCert paper: How AI can benefit the PRINCE2 Project Manager.

Project management for AI – making a robust case before deployment

As can happen with fast-moving technology innovation, what it delivers doesn’t always live up to its promise.

According to MIT Sloan School of Management, the AI “hype cycle is slowing, as organizations confront the challenges of enterprise AI deployment and the need to drive tangible business value”.

For example, despite 75% of UK companies using AI tools, fewer than a third are seeing positive return on investment.

And the problem – says Ritam Gandhi of digital product studio, Studio Graphene – is that in the “rush to adopt AI”, companies are missing some important elements, such as defining:

• Where it sits within the workflow
• What decisions it will inform
• Which processes it will support
• The criteria for measuring success

“Without defining these things, building a long-term business case for AI and realizing its value will be difficult,” he added.

Key actions this year for organizations striving to gain benefits from AI – according to Thomas Davenport ,professor and fellow at the MIT Initiative on the Digital Economy and Randy Bean, adviser to Fortune 1000 companies – include:

1. Create use cases for AI agents that can go enterprise-wide.
2. Think how AI can go wider than individual productivity to areas such as new product development, customer experience, etc.
3. Develop capabilities to build and test agents.
4. Leverage existing AI while considering how further investments will affect future business strategies.
5. Allocate responsibility for reporting on AI to senior leadership.
6. Create “AI factories” that combine tools and business processes to expand AI use cases and obtain more value.

But while progressing these actions may be important for organizations, they should also be managing them as a project output, according to digital learning services provider and PeopleCert partner, ILX.

A project management approach to implementing AI

Offering advice on how to effectively manage an AI project, ILX explains how “implementing AI” is “a multidisciplinary initiative that often involves new technologies, uncertain outcomes, and significant cultural change”.

With such initiatives likely to involve multiple departments, the necessary “governance and control…to manage its risks” should come from a structured project management method such as PRINCE2 Project Management.

But what do ILX’s suggestions mean in practical terms?

• Clarifying what an AI project will deliver

  The business case and project brief in PRINCE2 Project Management will “define clear objectives and align the project with real organizational needs and ensure it delivers meaningful value…”

• Establishing governance

  Accountability is supported by appointing roles, e.g., project executive, project manager, and senior users.

• Taking a stage-by-stage approach

  Managing an AI project in stages enables the ability to “reassess progress, review lessons learned and decide whether to proceed, pivot or halt the project”.

• Managing risks

  With the potential for risk across data quality, system integration, ethics, compliance and change resistance, PRINCE2 Project Management’s risk register ensures any issues are addressed.

• Having a long-term view of value

  Beyond a successful implementation, an AI project needs to deliver long-term value. Therefore, the benefits management approach – defining benefits to be managed and measured – helps this, along with post-project reviews to “assess outcomes, gather feedback and inform future phases or projects”.

AI projects: putting PRINCE2 Project Management and PRINCE2 Agile into practice - a practitioner’s view

Sagar Zilpe is an AI/PMO adviser and PeopleCert Ambassador at ZITRIX® Technologies and WorKnoW®. Here, he draws on his experience across enterprise-scale product development and transformation programmes – including AI-enabled platforms, intelligent automation initiatives, and implementations within regulated environments such as healthcare – to share his perspective on deploying PRINCE2 Project Management and PRINCE2 Agile for AI projects.

PeopleCert (PC): Have you used PRINCE2 Project Management to help organizations implement an AI project?

Sagar Zilpe (SZ): Yes, PRINCE2 Project Management – supported by PRINCE2 Agile – has been consistently applied across AI-led initiatives.

In conventional projects, you can generally define outcomes at the outset while execution focuses on delivering against those specifications. In contrast, AI initiatives are inherently exploratory. The behaviour of the solution evolves as models are trained, data quality improves, and real-world use reveals new patterns; therefore, uncertainty is a continuous characteristic of the delivery lifecycle.

PRINCE2 Project Management provides a method that establishes clear decision points, accountability structures, and alignment with business value, ensuring that progress is intentional. When combined with agile practices, it allows teams to iterate and learn while maintaining clarity on objectives, boundaries, and acceptable risk.

In effect, the method enables organizations to treat AI as a governed learning process, where each iteration contributes to progress.

PC: What elements of the method were most valuable in either supporting delivery of the AI project, or stopping the project?

SZ: The most valuable elements are those that introduce decision discipline in environments where outcomes cannot be fully predicted.

Continued business justification plays a central role because AI initiatives often begin with strong potential but uncertainty about what will be realized. By revisiting the business case periodically, organizations can evaluate whether the initiative continues to justify investment. This shifts decision making from assumption-based to evidence-based, which is critical to prevent ongoing investment where value is unlikely to happen.

The principle of managing by exception is powerful in the context of AI. Traditional tolerances focus on time, cost, and scope, but AI introduces additional dimensions such as accuracy, confidence, consistency, and drift. By explicitly defining acceptable parameters, teams are empowered to experiment within safe boundaries. This allows innovation to progress rapidly while ensuring prompt escalation for any deviations beyond acceptable limits.

A focus on products helps address a fundamental challenge in AI delivery by defining what “success” actually means. AI solutions are evaluated both by how they function and how well they perform and through defining acceptance criteria in terms of outcomes, organizations ensure that delivery aligns with real-world expectations.

By using stage-based control, organizations create specific points where they can reassess feasibility, risk, and impact – ensuring that progression is justified.

And, in AI environments where small adjustments can create large effects, issue and change control mean that changes to, for example, data, are managed so their impact is understood and aligned with overall objectives.

PC: What was the impact of using PRINCE2 Project Management for this purpose?

SZ: The most meaningful impact has been the shift from activity-driven progress to evidence-driven progress.

This results in more purposeful iteration as teams can experiment while guided by clearer signals on what is working and what isn’t and rework is more targeted and efficient. Meanwhile, stakeholder confidence is supported as PRINCE2 Project Management makes decisions more visible. Also, the ability to stop an AI project that is not realizing value is important, so organizations can redirect resources to initiatives with greater potential.

One of the most significant lessons learned from using the method in AI projects is expressing baselines as acceptable performance ranges rather than fixed outputs, allowing them to adapt while still providing a reference for decision-making.

And with underlying issues in data quality, assumptions, or design, quality should be treated as an early indicator rather than a final validation step. Controlling quality through verification, validation, testing, etc, helps prevent issues from becoming larger problems later in the lifecycle.

Ultimately, PRINCE2 Project Management – especially combined with agile delivery methods – allows organizations to learn, adapt, and progress while maintaining alignment with value, quality and accountability; ensuring innovation remains focused and controlled.