I am often asked what is the ROI in AI. For company leaders and executives, this is a crucial question. It is a huge topic and there is no easy answer, TL;DR It’s big, as long as you know what to measure.
AI is changing business and how all the industry leading companies are investing millions in tools to improve their business. Don’t take my word for it, just take a look at any company leading index? and all the leading companies in the list are investing heavily in AI/ML. Why? For me, at this point it is because of differentiation. AI allows companies to do more things, more efficiently, or create new revenue streams which fundamentally transform their companies and sets them apart from their competitors. Not long from now, however, it won’t be about differentiation, but rather about survival. The companies who won’t leverage AI in their business, will most likely disappear.
So, when investing in AI, where can companies find the ROI? Calculating the ROI in AI projects is not straightforward, mainly because it affects the companies transversally, and oftentimes we are just not measuring the right metrics at the right time or at the right place. I like to think the benefits of AI are like an iceberg, there direct and measurable metrics, and indirects and harder to measure metrics, but that have a profound impact on the company.?
Let’s go through 5 metrics to measure the impact of AI projects, with an example in each one so it's easier to understand.
- The classic costs savings or increase of sales. This is a great place to start, as it is the most frequent and most wanted metric for any business. For example, an insurance company receives hundreds of requests per month from customers who need to send a copy of a telephone bill, or an electricity bill to prove they live in a certain address. Internal customer care teams have to open that email, fill the gaps in case of small typography mistakes, classify the request and channel it through the proper internal processes. An NLP engine can understand the intent and content of the email including attachments, and will save thousands of hours to the company. Or a recommendations engine in an e-commerce site, that can recommend the products the customer is looking for and make an enjoyable purchasing process. When Amazon integrated recommendations, it registered almost a 30% increase in sales compared to the same period last year. These AI tools work, and these metrics are easier to calculate.
- Talent retention. It is a huge problem nowadays, companies are really struggling to retain their talent because it takes a big amount of internal effort and hours to find and hire the right talent, and if they leave, they create a big problem. Projects are underserved, the rest of the teams have to work more so they also increase the risk of burnout, NPS scores plummet, HR teams struggle to deliver or external agencies increase their fees and so on and so forth. AI tools can play a huge role in this. They can take care of the most time consuming tasks, allowing employees to focus on more value-added tasks. We humans should work on more creative, value-added tasks, and companies should benefit from the employees’ know-how, experience and creativity, rather than from “box-moving”. Take for example a risk analyst. Her job is to assess the risk when onboarding clients for a big law firm. She wants to spend her time using her expertise, know-how and make good decisions for the law firm. But in reality, she spends 50% of her time filling internal forms, and using a software that is not able to even remember her previous actions. This will start growing on her, and chances are, she would rather work for a firm with a tool that can easily find companies’ names, and that can automate the most obvious cases and pint her attention to the more complex cases that require her attention. This is exactly what an AI tool could do, find anomalies in large databases of transactions, “fill the gaps” and automate the easiest cases so she can spend her time in more value-added tasks,
- Risk management. Have you ever thought that, when companies pay employees, they rent their know-how, they don’t own it? If the employee leaves, the know-how also leaves the company. Take for example an expert car mechanic. He was the best at diagnosing cars, faster and more accurate than any other technician. But retirement time came, and he left, leaving a big gap in the company. Image that know-how, expertise and decision making could be “digitalized”, then the know-how would stay in the company. AI tools can actually do that. Especially machine learning algorithms are built on historical data, so if the company kept a record of the results of the employees’ decisions, it’s possible to build an ML model that would essentially mimic the decision making of those employees. Imagine an AI tool that listens to a car engine and can diagnose what is the problem with that car just by hearing, like the expert mechanic would do. Or a diamond expert who is able to assess if a diamond has any inclusions with a quick look at it, while other employees need a second or even third opinion before making a decision. If those expert employees would leave, the know-how stays inside the company, and the company owns it.
- Overhead costs reduction. When making plans for the upcoming year, companies use forecasts as accurately as they possibly can so they can plan their budgets and business results properly.The accuracy of those forecasts will make a big difference in “winning” or “losing” that year. Still nowadays, companies and teams base their forecasts on their own experience, but of course that is prone to error. Let’s take? the example of a company that needs to place their orders as early as possible to secure on time deliveries and negotiate the best possible business terms. Without an accurate prediction, there will be inaccuracies in the ordering process, which can cause not only contract penalties, but also all kinds of internal “fires” to extinguish that only add overhead costs and reduce company profits. However, a ML prediction tool can accurately make supply/demand predictions taking into account all kinds of “out-of-the-box” factors (e.g. weather prediction in a retail store flagging upcoming bad weather hence the need to stock up on umbrellas). The ML model can make a prediction of the orders needed to satisfy the demand in the following 6 months, and optimize the logistics of the shipments, so the purchasing department could ask their supplier to place the exact orders until of the year, ensuring best commercial terms and reducing the amount of overhead costs throughout the company to extinguish those internal fires.
- Company valuation. For me, this is the most important one. And it’s a bit controversial. Why? I think if you buy ready-made AI products you are essentially giving away your precious data, and empowering those AI startups which will be able to improve their models thanks to YOUR data, and ultimately, you may even be empowering your competitors. Let me explain. According to Andrew NG from Landing AI, to which I couldn’t be more agree with, we are in a data-centric stage of AI. This means, the models’ performances have more or less reached a uniform performance, and all smart AI engineers, with the same data, will reach similar results. This means, the difference is in the data. Data is the key for success, and those companies with better data will achieve better results than other companies with a poor data strategy. So, if data makes such a huge difference, when you purchase AI related ready-made products, you may consider if you are not just empowering that company, and their clients, which may very well be your competitors. What is the solution? In my opinion, if it’s a “secret-sauce” of your business, like how to make better risk assessments than your competitors, or sell more than your competitors, anything that is related to your competitive advantage, my advice is clear: build, not buy. Build your in-house solutions and custom-develop tools where you own the IP, or at least share it. If you don’t have the skills inhouse to build such a thing, you may want to consider partnering up with an AI solutions provider, like Etnetera AI, with flexible business frameworks. You can even create a new tech division, or company, pr product, expanding your portfolio and increasing your company valuation.
Ok folks, this has been my view on the ROI in AI. I hope you found it interesting, and if you have any questions regarding what, why and how to implement AI in your company, please don’t hesitate to reach out and we can have a chat and share my two cents.
