80% failure rate in AI projects, can scope management help?

There’s a new wave of work in AI, which leading agencies and tech consulting firms are jumping on. It’s a move that makes sense – AI has never been more accessible, yet companies still need talent with the time and knowledge to implement it effectively.

In a recent study by the RAND Corporation, 97 percent of the business leaders surveyed reported that the urgency to deploy AI-powered technologies has increased. While only 14 percent of organisations were fully ready to adopt AI.

The study highlights the seismic shift toward AI that has occurred in the past several years. It’s no wonder that agencies are furiously paddling to catch the wave. However, the risk is as great as the reward, with the RAND study also noting that more than 80% of AI projects fail.

So, in the midst of a highly dynamic technology upheaval, what can you do to seize the opportunity without it sinking your business?

A key reason many AI projects fail is down to leaders struggling to articulate the exact problem they want to be solved and the metrics by which success should be judged. This issue is compounded if leaders frequently shift priorities, leaving engineering teams scrambling to keep up.

It’s a challenge that experienced professional service providers should be familiar with. Clients have the domain experience, but they’re less likely to have the technical experience required to solve their challenges. Nor do clients have the benefit of delivering many similar projects, so they don’t have the prior knowledge to help guide their decisions.

This is why it’s crucial to set the project up for success at the very beginning of an engagement.

Defining the problem AI aims to solve

AI projects often falter from the outset due to poorly defined or ambiguous problem statements. When leaders don’t clearly communicate the specific issue they want to address, teams may build AI solutions that don’t align with the business’s actual needs. This can be a costly mistake, especially considering the upfront effort required in AI projects.

The solution lies in adopting a problem-based approach to AI projects. AI initiatives should be based on the problems to be solved, not the solutions to be applied. All potential problems the organisation aims to address should be assessed upfront. If AI projects are strategically prioritised, they are less likely to be derailed by other challenges that hadn’t been considered. The project can then move on to defining how the chosen problem might be solved, which will inform the specific technological solutions.

It’s at this stage that agencies need to pin down the scope, while also putting a lot of thought into managing customer expectations.

Making project scopes work for you in AI projects

While recent tech advancements have made AI much more accessible, meaningful adaptation to unique business use cases is anything but simple.

A scope that provides both detail on the deliverables, while also clearly illustrating the variability of effort will help set customer expectations at the outset.

Like any project, breaking the work down into phases can help build out the scope and illustrate the scale of the engagement. Any parts of the project where the variables are known can be defined and quoted. Where there is complex work to be carried out, a range can be quoted instead – this documents how the cost might vary upfront and provides an opportunity to educate clients at the beginning of a project instead of explaining in retrospect. Note: even with a range, it is important to define inclusions and/or assumptions so as to avoid clients assuming the max range includes all possible scenarios.

For example, one of the biggest challenges in AI projects lies in data preparation, which requires intensive effort to ensure the data is clean, relevant, and has appropriate governance and security in place. This work is time-consuming and expensive, and often cannot be estimated upfront because it is largely dependent on the state of the client’s data.

Maintaining flexibility through ‘scope management’

The RAND study notes that 84% of AI project failures are rooted in leadership shortcomings, and that leaders who frequently change priorities can derail projects.

However, as innovation in AI is happening so quickly, defining the problem upfront is just the beginning. While AI itself isn’t a new concept, its mass adoption is, and it’s being applied to increasingly diverse use cases. Simultaneously, the technology’s capabilities are advancing rapidly, making it imperative for organisations to adapt their approaches as they go. No business wants to deliver a solution that is already out of date on the day of release, so companies delivering AI projects need robust systems in place to manage changes effectively.

When changes to the scope or priorities of an AI project are necessary, it’s critical to have a single source of truth that transparently records all changes along with the necessary approvals. This ensures that everyone involved in the project is on the same page, particularly when multiple stakeholders are involved in decision-making.

If a change goes awry, having a clear record of when and why the decision was made, along with who approved it, can be invaluable in diagnosing the issue and implementing a course correction. This level of transparency not only mitigates risks but also fosters accountability, ensuring that the project stays aligned with its original goals despite necessary adjustments.

This approach is known as scope management.

Can scope management increase the success rate of AI projects?

The success of AI projects hinges on the ability to balance clear problem definition with the flexibility to adapt to new developments. Leaders must ensure that their teams have a precise understanding of the problems they are tasked with solving. At the same time, they must be prepared to manage change effectively, ensuring that any changes to the scope are justified and clearly recorded.

By prioritising scope management organisations can create an environment where change is  anticipated, clients are empowered and informed, and new requests create opportunities instead of creating project-sinking icebergs.

For further insights into the challenges and strategies for successful AI project implementation, the full RAND Corporation report offers valuable lessons and recommendations for leaders across industries.

‍