4 Questions to Help You Call BS on AI

Using artificial intelligence (AI) software is a great idea — if that artificial intelligence software is actually useful. It sounds obvious, but surprisingly it isn’t.

by Urvish Vashi, VP Global Marketing, HighRadius

When implemented thoughtfully, AI-based products can help your company improve customer engagement, restructure back-office processes and even result in better financial performance. A recent Forbes article projects AI to enable profit gains of 38 percent by 2035.

Eager to cash in, many companies find themselves seduced into the idea of AI without actually understanding how it can deliver value. With many executives eager to jump on the AI bandwagon and so many products to choose from, making informed decisions is critical. Instead of falling for the solutions and software with the most hype, sharpen your focus on which ones will offer the biggest return on investment.

After all, not all AI solutions are created equal — here are some questions that decision-makers should be asking before they sign on with a particular solution.

1. How much data has the product been trained on?

Artificial intelligence systems become “intelligent” through machine learning, a process where computers find patterns and draw conclusions by analyzing vast amounts of data. The more data that the program has used in its learning process, the more likely it is to draw accurate, useful and universally applicable conclusions.

With this in mind, make sure to ask your vendors how much real-world data they have trained their AI systems on. The more the better. You can expect to gain more from solutions whose artificial intelligence systems have had significant experience working with large clients in data-heavy industries.

If the product has been trained on relatively little data, you may be giving more than you get. The system may not have learned enough to automate your processes and your teams might end up spending significant amount of their time and energy in actually training the system to get to a basic level of automated output.

2. Are you paying for AI technology or for an actual solution?

You might come across vendors who are willing to sell your company machine learning technology as opposed to a prepackaged, out-of-the-box, AI-based solution. Owning the technology to control your own destiny might sound appealing, but in reality it’s highly risky and time-intensive.

You’ll end up conducting a science experiment, devoting too many resources to figuring out how to actually put that technology to use. Instead of accidentally involving your company in an unclear AI quagmire, choose an AI-based solution that solves specific problems for your business process. Before you buy into a product, the vendor should be able to explain exactly what problem it will solve and how it can save your company time and money.

3. Is the product genuinely intelligent, or is it just a bunch of “if this, then that” rules?

With a recent surge of interest in Robotic Process Automation (RPA), many companies are inquiring about whether an RPA solution might be just as effective as an AI solution. The answer is simple: it won’t.

RPA is very different from AI. RPA relies on hard-coded rules to automate user actions and simple business processes. Essentially, RPA systems are not much more than a series of “if this, then that” algorithms. However, RPA systems are brittle because they will continue to work perfectly only as long as the inputs to the process stay exactly the same. With an RPA-based system, the algorithms need to change every time a business requirement changes. That will either require you to reconfigure and recode the “bot” with each change or else end up with a system that doesn’t function properly.

Make sure you know what you’re buying into when you choose RPA; don’t fall into the trap of thinking that a cheaper product with a fancy name will perform just as well.

4. What responsibilities will fall on your own IT department?

One of the benefits of purchasing software powered by artificial intelligence is that it should require relatively little maintenance. A true artificial intelligence solution should be able to learn and adapt to changing situations.

One-time setup is understandable, but if you constantly need to engage an army of IT experts for system maintenance, then you’re likely not being sold a true AI solution. Before purchasing, take the time to ask your vendor about the responsibilities that your IT department should expect to handle.

Without Return on Investment, AI is BS

The bottom line is that your bottom line matters. No matter how exciting or groundbreaking an AI product is, you should evaluate an AI purchase the same way you would evaluate any major purchase: based on expected returns.


Thanks for reading. We hope you enjoyed another contributed article by hands-on industry experts. Let them know you liked it by clicking the 👏 button — as often as you like.

Visit us on Twitter and don’t miss the current fintech newsletter issue here.