Over the last few years, companies have progressed beyond the experimentation phase when it comes to implementing Artificial Intelligence (AI) technology. Bigger organizations, in particular, are achieving positive results and seeing tangible bottom-line impact —in other words, maximizing the ROI of AI.

According to a 2019 McKinsey survey, 63% of larger enterprises have increased revenues and 44% have reduced costs across business units that adopted AI. At the same time, AI and Machine Learning (ML) initiatives continue to fall short for a significant number of companies. A recent IDC survey found that 28% of AI/ML initiatives failed, as reported by 2,000 enterprise IT leaders and decision-makers. 

One area where we believe leaders still need support is around how to determine the true costs and benefits of implementing AI/ML at scale. Cost-benefit analyses for AI/ML projects are much more nuanced and layered than people realize.

In this post, we’ll explain why AI/ML projects are hard to evaluate from a financial standpoint, as well as discuss why it often makes sense to bring in an outside AI/ML expert who can help you define clear outcomes and put proper guardrails in place that ultimately lead to success.

Challenges Calculating the ROI of AI

It’s easier to understand why cost-benefit analyses are more complicated for AI/ML projects if we start at the beginning – with the classic return on investment (ROI) formula:

ROI: (gains – investment) / investment

Calculating ROI is one way that companies measure the value of their invested capital. The formula is powerful because it can be used across a wide range of industries and projects, including those involving digital operations. 

With AI/ML, calculating ROI requires more thought.

You have to define “gains” carefully, as there are many ways you can interpret this metric. For instance, “gains” could represent:

• Increased productivity
• Incremental revenue 
• Lower overhead
• Lower operating costs
• More customers
• More business per customer

The list goes on and on. 

So, you have to make sure you define your “gains” with as much specificity as possible in terms of your unique business model. Otherwise, you won’t be able to quantify or qualify success.

The “investment” variable can be even harder to nail down. Analysts tend to think solely in terms of the number of dollars spent, but there are other factors to consider, such as:

• Infrastructure costs: do you have to upgrade your IT infrastructure to support AI/ML?
• Taxes: how will your tax burden change if you hire new talent and/or purchase assets?
• Inflation: if the timeline for ROI is further out, how will inflation affect future returns?
• Opportunity cost: what else could you do with your time, resources and capital?
• Investment horizon: how long are you willing to invest in your AI/ML project, in particular as it moves from MVP to productization

To put things simply, back-of-the-envelope math won’t cut it for AI/ML projects. Your cost-benefit analysis should break down both the “gains” and “investment” variables into smaller, measurable units that are relevant to your business. 

A Broader Look at AI/ML ROI

Outside of determining the specific numbers to include in your AI/ML ROI calculation, there are other ways to think about the value of your investment. The AI market research firm Emerj¹ identifies three “types” of ROI that businesses can capture through AI/ML:

1. Measurable ROI
2. Strategic ROI
3. Capability ROI

Measurable AI ROI

Measurable ROI consists of both financial and qualitative outcomes. This bucket includes the “gains” referenced above — lower costs, lower risk, higher long-term value, etc. At the end of the day, what matters most is being able to identify specific metrics that you can track (e.g., our AI recommendation engine will increase customer satisfaction by 15% within 6 months as measured by customer surveys). Keep in mind that you may have to recalibrate your measurable goals as your projects evolve. 

We recommend sticking with endeavors that you can easily evaluate, as well as those that move the needle for your business, especially if you are trying to prove that AI/ML should have a role to play in your organization’s future.

Strategic AI ROI

Strategic ROI includes outcomes that help differentiate a business and/or position it for long-term success. It’s important for your leadership team to articulate how AI/ML creates strategic value for your enterprise, as this is where the real bread and butter lives.

AI/ML has the potential to augment existing capabilities, accelerate digital transformation, enhance customer experiences, and much more for those who can harness it’s full potential. Consequently, you have to keep strategic targets on your radar. 

Capability AI ROI

Capability ROI refers to value derived from building a solid foundation of AI/ML understanding within your organization. In other words, this is how you measure your progress towards achieving AI maturity.

When your business has the talent to implement AI/ML projects and your executive team has a conceptual grasp of the underlying technology, you have the potential to change the game. Although it’s difficult to justify AI/ML initiatives solely based on the capability ROI category, it’s a crucial element to consider.

Get Internal Financial Stakeholders On Board

We should also mention here how important it is to get your company’s financial stakeholders on board with your AI/ML projects as early as possible. Getting buy-in gives your team an additional opportunity to validate the outcomes you are pursuing. It also lays the groundwork for future projects, in which executive sponsorship will be critical to success. 

With that in mind, let’s move on to why many enterprises are choosing to hire outside AI/ML expertise to support their projects. 

When Does It Make Sense to Work With Outside AI/ML Experts?

To ensure your team sets reasonable expectations and stays on track, it often makes sense to work with AI/ML experts who have followed hundreds of projects from start to finish. Regardless of what resources and skills exist in-house, having end-to-end AI/ML support can be invaluable for achieving desired outcomes.

After all, leveraging AI/ML successfully over the long term requires highly skilled resources, multi-disciplinary teams, quality data, and ongoing collaboration. It can be difficult to train up existing employees or hire the talent you need across these areas. 

When you hire an outside firm with demonstrated success in the AI/ML world, you get the best of all worlds at a fraction of the cost. You can hire the talent you need, exactly when you need it. 

Third-party AI/ML experts also know how to set measurable goals and milestones according to your company’s unique objectives. Furthermore, they should operate under commercial contracts that hold them accountable to the outcomes they generate for your organization.

Earlier, we mentioned that AI/ML is experimental by nature – there is no standard playbook or tried-and-true method for success. Consequently, AI/ML-specific experience is paramount. 

AI/ML consultancies that have executed countless projects in different industries can draw from previous engagements and significantly de-risk your initiatives. They can better anticipate pitfalls, define dependencies, and bring in the ideal resources to maximize success.

Lastly, partnering with an AI/ML company can make a lot of sense for smaller companies with tighter budgets. Large enterprises can afford to recruit world-class ML talent and experiment continuously. 

Wrap Up

When evaluating potential AI partners it’s helpful to have a tool to assess their capabilities. You can use this AI/ML Assessment Matrix to score your internal organization’s skills as well as,  rank potential outside AI partners. 

¹ Emerj.com Three Kinds of AI ROI – Emerj’s Trinity Model