It's no surprise AI projects are top-of-mind for the C-Suite and pressure is mounting to do more: 84 percent of C-Suite executives believe they need to leverage AI to achieve their growth objectives. And organizations that successfully scale AI projects, pilot and scale more initiatives than their counterparts at a rate of 2:1; these successful firms also report shorter timelines, lower costs and higher returns as a result of doing so.
With more and more AI projects on the agenda, businesses know they need to be strategic but may find themselves with some challenges: running many AI pilots can be unwieldly, and some may be deployed by leaders for the sake of using AI rather than to meet a business need. It can also become difficult to know which AI projects to prioritize and how to calculate the return on investment of each initiative.
But there is good news. AI pilots, when part of a collective portfolio, are worth more than any pilot alone. Therefore, creating an AI portfolio can help leaders alleviate some of those challenges.
Like in an investment portfolio—where the investor will have a mix of assets like stocks, bonds and commodities to reflect their risk tolerance and financial goals—businesses should diversify their AI portfolios based on factors like probability of success and potential ROI. And when it comes to calculating that return, the beauty of a portfolio is that a company can still achieve their projected ROI collectively even if one or two projects don’t yield the results they expected.
So, how do leading organizations choose which projects to include in their portfolio to maximize the potential for success?
1. Evaluate strategic fit
For an AI portfolio to work, the business strategy should act as the portfolio’s North Star, guiding the selection process by qualifying which projects align to the overall goals of the organization.
Start by defining what ‘value’ means in the context of your company and the readiness of different parts of the business to work with AI to unlock that value. Ask yourself: What customer goals are we trying to meet? What competitive advantages are we looking to gain?
A financial services company, for instance, might want to use AI to improve customer satisfaction or create new revenue lines, while a gas and oil company might want to reduce its dependency on hydrocarbons. Once you are clear on the outcome you want, it becomes easier to judge whether the individual AI pilot is aligned with that goal.
Companies that fail to effectively evaluate strategic fit often end up with a long list of use cases that aren’t focused on driving a specific result, reducing the chances of overall success.
2. Determine value
While some pilots may look like a poor investment initially, the long-term gains could far outweigh this. Conversely, other pilots might yield good short-term value but not be something you want to invest in for the long run. Both long- and short-term benefits should be taken into account when building your portfolio.
So how do you determine the value of your AI use cases? The larger the organization, the harder it becomes to attribute value to any single pilot as separate from the business outcome that it serves. To overcome this, you must demonstrate the impact your AI is having using metrics – like net revenue, market share and new product lines – that demonstrate value to the company’s stakeholders.
Secondly, take into consideration that AI is continuously learning, and this will affect the value it provides. For instance, as data and real-world circumstances change, algorithms might have to adapt and the end result will evolve with it.
Third, the right modelling techniques to do these calculations will allow you to compare like for like across the portfolio.
And finally, you should always evaluate the AI or ML in the context in which it’s used. You need to look at the contextual influences which come from human involvement in applying that AI-driven decision to the situation at hand.
Say AI is evaluating credit data at a lending firm, which an employee will then use to make a decision. You cannot attribute the value of that process and decision to the AI alone; it also needs to include factors such as how the employee used the AI and whether it added value to their work.
3. Assess feasibility
The final step to fine-tuning your portfolio is to assess feasibility—the conditions required for success—and prioritize which pilots to execute.
Priority pilots should be applied use cases that you know you can get into production. If you think of the pilot as a car and the potential roadblocks as traffic lights, essentially, you are looking for the pilots with the highest number of “green lights”.
When assessing if a pilot has a high number of green lights, you must consider the factors that will increase the possibility of success:
- Is the right data available in house—and is it accessible and clean?
- Do you have a clear “route to live”? In other words, the knowledge, people, technology and processes necessary to take a pilot to production?
- Would you benefit from buying, building or partnering to get the expertise or technologies required?
By doing this assessment on the AI pilots you’ve already deemed strategically fit and valuable, you can easily see which have the most green lights – and focus on those first. Doing so buys you some time to gather all the parts—such as missing data, technology, talent or partners—needed for orange- or red-light projects to succeed.
Importantly, this time also gives you an opportunity to do some research and understand what’s really at the heart of the problem from a human and business perspective. Using some human-centric design thinking, you may realize that the focus of the project itself may need to change or that you have to remove the project from the portfolio altogether because you don’t have the right factors to move forward.
By creating a portfolio of pilots that considers the AI’s strategic fit, value and feasibility, you can manage your AI projects holistically and thoughtfully. It’s an approach we’ve seen our clients adopt to ensure projects move quickly to production – and ultimately, to scale.