Artificial Intelligence still requires design intelligence

Then I took a deep dive into IBM’s AI: Watson. A colleague of mine had already made a start and a client wanted to experiment with a chatbot as part of the service to its clients. AI at first seemed very complex and daunting, but in an afternoon we had a chatbot up and running in a design prototype that could hold a conversation around a user problem. The interface of IBM’s Watson for conversations is very intuitive and well designed. I learned that AI means you have to train the bot. It can automated certain things based on libraries and machine learning, but it needs a trainer. One of the images that I had in my head was that you could feed the bot a database of FAQ’s and that the AI would just interpret the database and could talk to a human about all his problems. But that’s not how it works. The interaction needs to be designed.

Not that I am avoiding coding at all costs, but I noticed that my first encounter with Watson did not require any coding. I think it’s helpful if you understand the basic logic of coding when training the AI but I could make this first prototype without a line of code. Behind the scenes there is a lot of code and combining the different elements of Watson and the interaction with the other technology of your service through the API’s requires code, but IBM succeeded in creating a great interface for the training.

The complexity of combining different parts of the AI and the integration with your systems is one of the things I learned while talking to the people of IBM. Our local Smart Services Campus organized a workshop with IBM to get started with Watson. The biggest eye opener for me was that AI is a technology that can help you build great services. Next to a technical challenge to make the AI work better, AI is a service design problem. Now this makes perfect sense to me, but my image of AI was totally clouded by science fiction. How well an AI bot can perform is as much a matter of design as it is a matter of the code. At least that is how it seemed with Watson. I have no experience with other AI’s, but the technology seems to work so well that technology is no longer the bottleneck for providing great services with AI. Having a conversation with a bot is a function of how well the conversation was designed: What is the context? Where is the user in his customer journey? What task is he trying to complete? What are the problems his is facing when completing the task? Of course you can trick a bot and they still fail the Turing test, but that is not the question. The question is how good a bot can help a user complete a task as part of a service he is using. The goal of a bot is not to know everything about everything and be as human as possible. A good bot is a specialist that is well designed and trained. That is also how Watson is built up: separate products and services you can link together to enhance the power of the whole. If you have a team of specialist bots, they can work together and the individual bots can be designed and trained better.

The guys from IBM at the workshop were developers. But their message was: talk to a service designer if you want to use bots. The technology modules are there. They are fairly easy to use, intelligent in themselves. They need some connecting with API’s, but the success of this technology will be determined by the quality of the design: understanding of context, user empathy and solution design. I did not see many service designers at the workshop but service designers need to team up with developers to make bots a success. Many people have bad experiences with bots and the hypothesis that I have, based on my experience and talking to the guys from IBM, is that this is more due to bad design than bad technology. The problem is that this is all about the little things of a service that are hard to design. The result is that people get a wrong image of what AI is, what it can do and how you should go about deploying it. AI can be a game changer for Service Design and I wonder how many Service Designers have dived into the subject.

Related posts:

By implementing the strategies outlined in this blog post, you can maximize user engagement and participation in your giveaways. Remember to tailor your approach to your target audience and continually analyze and refine your strategies based on data-driven insights.

Can artificial intelligence be racist? Well, it depends. Let’s say you’re an African-American student at a school that uses facial recognition software. The school uses it to access the building and…