How to maximize success and ROI with AI projects

A recent survey of 2,000 CEOs revealed that just 25 per cent believe that their AI projects have achieved ROI. Further to this, the IBM poll revealed that just 16 per cent of AI projects have scaled enterprise wide. In some ways, these are stark findings, but they are not entirely surprising.

In financial services, many organizations are still grappling with legacy systems and code bases, providing a further layer of complication to the usage of AI-powered technologies, making it a bit more challenging for some to seamlessly deliver their high-value initiatives.

But that doesn’t mean that organizations should not seek to explore the opportunities AI is delivering for competitors, challengers and disruptors. Indeed, 64% of CEOs surveyed acknowledge that the risk of falling behind their competitors drives investment in some technologies before their potential value is clear.

Missing out is clearly not an option, even if only a quarter of projects succeed. But with careful planning and a dose of pragmatism, there are a number of things organizations can do to ensure that the time and effort spent on less successful projects is not in vain.

Setting parameters for success

The reason so many AI projects are failing to deliver ROI is that there is clearly still a lack of focus when it comes to identifying high-value strategic initiatives over broader applications of the technology. I also suspect that many projects lack clear objectives and have not had the right metrics put in place, against which success can be judged.

When the aims of any technology-led initiative are not well defined, the project will lack focus and unravel. This is the earliest sign that a project is not viable, however, it’s also early enough in the process to stop, regroup, and redefine objectives.

For any project to succeed, buy-in and validation of the use case must be secured from internal stakeholders and delivery teams, from product owners, data science teams and developers, to the C-suite, with all agreeing clearly defined KPIs. This will ensure the right level of investment is in place to secure the appropriate development and resource bandwidth.

AI projects are often put on ice as data scientists wait for the necessary permissions to use certain datasets. This is why legal and governance clearance for projects is so essential, so these teams should be included in the early planning stages of projects.  

In my experience, AI projects are rarely dropped entirely, as there are always elements of the work that can be reused elsewhere to cut down rework. Sometimes, it’s also just a case of waiting a little longer than expected for higher quality datasets to become available, but these projects can be picked up again when they are delivered.

Failure is never total

Experimentation and failure are both essential elements of innovation and it’s important for technical teams to test themselves and explore the art of the possible with advanced technologies.

Most organizations today have internal innovation labs, or at least some allowance for experimentation with new technologies. In financial services, this experimentation is of course conducted in secure test environments, particularly when it comes to generative AI use cases. But pushing the boundaries is crucial for unlocking truly transformative innovation.

The important thing when it comes to experimentation, is to fail fast—which usually means putting the project on hold until certain requirements are met, such as data access, legal clearance or the availability of a new tool.

For data scientists, projects may not always be successful, but as they explore a new technology, they may identify new learnings that will serve them and their teams well into the future. With research and development in AI increasing by the day, it’s always better to pause a project that may not be on track, rather than abandon it entirely. New capabilities and techniques are emerging all the time, which unlock new possibilities and use cases. With the level of research and development that is going on in AI, it’s likely that a solution that solves a problem with a paused project may become available in the near future.

In this respect, time is always on our side. This is borne out by the more positive outlook also revealed by the IBM survey in relation to the future of AI investments. Of the 2,000 CEOs polled, 85% expect their investments in scaled AI efficiency and cost savings to have returned a positive ROI by 2027, and 77% expect to see a positive return from their investments in scaled AI growth and expansion. There are a number of factors fueling this more optimistic view, not least the expected growth in AI-based skills that will naturally occur in the intervening years. Another is the upward trend of AI maturity and AI-ready infrastructure that will certainly continue.

So, while some organizations may not find ROI to be where they currently expect it, these insights suggest that the initial lag is somewhat priced into technology adoption roadmaps, at least at a macro level. My advice in the meantime is to first establish a balance between ambitious and well-defined objectives and then be bold in how they are achieved.