Managing Customer Expectations in the Age of AI
Chief AI and Business at Nodeflux
Customers Having Unrealistic Expectations From AI
Working at an AI startup gives your clients the impression that you and your engineers are wizards, and can make the impossible possible. After all, most people are getting their AI education from Sci-fi movies. Your clients may have over and above expectations, which may truly become a challenge. There’s a huge gap in communication as to what the AI can do and what to expect from it. As most consumers are learning about the competencies of AI from movies and science-friction videos, and thus, their expectations keep going higher. Even in a thousand years, AI will not have the possibility of reaching the intelligence of human beings. So, the question arises: how can AI experts manage the expectations of clients?
How to Set Reasonable Expectations on AI Engineering?
First of all, the only thing that would save AI experts on managing their client’s expectations is putting a thorough and comprehensive disclaimer. For instance, if you’re working on an AI system that can detect an object in front of you, you have to tell them about its limitations, and place your warnings on where or in what circumstances it may not work. In other words, you have to be able to break your client’s bubbles in order to come up with realistic forecasts.
Another example would be if you’re building a fraud detection system, where the setup should not be able to surpass the authentication process, discuss its possibilities with the client. In many attempts, people try to wear a mask or use a video to resemble themselves in front of the system, so there has to be some foundation for the system to detect it.
As AI technology keeps evolving, nowadays, we don’t need to go to the banks to open a bank account. We can use our smartphones or a laptop to proof our identity via uploaded documents and showing ID to the system. Sometimes a low-end smartphone can have a very different result as opposed to a high-end device. In that regard, we need to create an SDK.
We need to have an explainable AI mechanism. No matter how perfectly you design your system, unexpected situations are likely to happen. Developing an AI system calls for being ready for unrealistic circumstances to arise. With explainable AI systems, you can actually predict some of the incidents or unexpected situations that can happen.
The advantage of having an explainable AI system is that, in any event, if your system has a failure and there is a legal complexity that has arisen, you can have an explanation to justify. If you can clarify and prove that explainable unexpected events are likely to happen, it can eliminate the legal complications.
Navigate the Choppy Waters
- Working with AI and other newer concepts every day calls for you to explain and educate your customers accordingly. Answer questions like how the AI works, what the limitations are, and what they should and should not do with the system.
- Your system needs to be protected from an unlimited range of inputs beyond proper operational conditions from users, and therefore, you have to create an SDK for them. For instance, if you have an AI chatbot or speech recognition, your users or customers might say something explicitly, or in an extremely noisy place and if you have an SDK, it will prevent wrong inputs that will create strange situations from arising.
- Embed an explainable AI into your system so that there’s always logical reasoning behind what caused something to happen. You can also eliminate the gap between the system and the user.
Be notified about next articles from Adhiguna Mahendra
Chief AI and Business at Nodeflux
Connect and Learn with the Best Eng Leaders
We will send you a weekly newsletter with new mentors, circles, peer groups, content, webinars,bounties and free events.