Artificial intelligence (AI) is everywhere now. It’s in the products that we use every day, and it’s at the forefront of any innovation in every industry we can think of. SaaS is not an exception. HubSpot, Adobe, Google, Amazon, Slack, GitHub, Zoom, and other top-notch companies implement AI technology in their products.
Still, one of the main problems with AI-powered products is that very often people don’t trust them, or at least feel skeptical if such software is reliable enough to help them solve important tasks.
Like any other technology, AI-powered products exist to make people’s lives easier. And as a team of product designers, at Eleken we believe that carefully thought-out AI UX can help us perceive artificial intelligence as a technology that helps people, not replaces them. Machines can handle complex tasks that require attention to detail with ease. In turn, UX designers can give the product human empathy, a sense of trust, and clarity.
So below we’re going to talk about ways to create UI/UX design for AI and produce products that work well, loved by people and that enable entrepreneurs to build successful businesses around them.
To begin with, let’s see in what forms we can find AI technology in SaaS products.
Common use cases of AI in SaaS
Investing into artificial intelligence and machine learning is now definitely one of the greatest SaaS trends that allow advanced SaaS companies to provide their customers with hyper-personalization, and stay at the top of the market.
An AI-driven system can:
- Encourage users to interact more with your SaaS.
- Shorten the time needed to find specific content.
- Increase the purchase rate.
- Help an undecided customer to make a choice.
- Make the search results more relevant, and so on.
Here are several popular ways of integrating AI into cloud solutions.
An AI recommender system is an algorithm that developers use to predict a customer’s choice and based on it give relevant recommendations to this particular user. In other words, a recommender system is your highly professional shop assistant who knows their customers well and can offer exactly what they need. It’s often used to personalize the user experience of a product.
For example, I still remember that feeling of aha moment when I first discovered the list of suggested artists created by Spotify for me personally and understood what a match with my tastes it was.
Conversational interfaces (online chatbots) are taking the place of human customer service representatives, helping businesses improve customer engagement and increase work efficiency by communicating with customers in a human-like manner. Chatbots can give tailored advice, cross-sell items, respond to frequently asked questions (FAQs), or even become users’ virtual assistants.
For example, a Slack chatbot can help users manage emails, set reminders or set notifications to team members, collect and analyze data, and more.
Banks and other financial organizations often use AI technology to identify fraudulent transactions. They train a model using data with already recognized fraudulent transactions so that when an activity seems out of the ordinary and demands more examination the system detects it as an anomaly.
Shift Technology, a cybersecurity software, uses AI analysis to automatically find an individual or network fraud schemes in a variety of industries. The solutions are then made available through their cloud platform.
Businesses may use AI to analyze great volumes of customer data and provide tailored experiences for each client. AI-driven segmentation helps them to learn customer behavior, identify their preferences, and, based on collected insights, effectively target marketing campaigns to reach the audience they want.
For instance, Salesforce uses artificial intelligence for customer segmentation by enabling sales teams to create personalized marketing campaigns using its CRM Marketing Cloud.
AI is excellent at spotting patterns and insights in vast datasets that are just invisible to human eyes. AI-powered systems are able to analyze data from a large number of sources and make predictions about what will work and what won't. They can also combine all of the company’s customer data into a single, unified view or use analytics data to predict product demand based on seasonal trends, previous purchasing patterns, and other factors.
For instance, Apptio uses AI-driven insights to help other SaaS businesses with licensing, planning, and investment. The Apptio SaaS model incorporates AI-generated analytics that helps deploy and use the financial systems of other businesses. This way, Apptio uses AI to help other firms make wise investment decisions.
These were only some of the cases that illustrate how SaaS businesses can benefit from utilizing artificial intelligence. However, to make it work for your product, it’s essential to create an AI user experience that makes people feel that they are in control of everything, that you’re being transparent with them, and most importantly that they can trust you and your product.
Best practices of designing UI/UX for AI features
If you incorporate a great AI feature into your product but people don’t understand it, then you risk them using your product incorrectly in general, or simply ignoring AI functionality. So you have to work hard to create the user interface and the user experience that help people trust artificial intelligence in your product.
But how do UX and AI go together? Here are some AI UX design principles to help you figure it out.
1. User problem first: do not use AI for the sake of AI.
No matter what technology you use in your SaaS platform, first of all, you have to remember that you are designing for people. And your main concern should be finding the right problem. Designers need to identify the right challenges to solve and recognize what solution would bring the maximum impact and benefit.
So, instead of jumping right into algorithms, consider how individuals currently cope with the issue and how you can make their experience better. Sometimes you can make your app feel smart and personalized without implementing machine learning.
For instance, Gmail looks for emails with the words "attachment" or "attached" and shows users a reminder not to forget to attach a file. This method works well enough to cope with a problem. Even though AI technology would be much more effective in identifying potential errors in this case, Google saves money by opting for another solution.
2. Distinguish AI-driven content.
Users need to be able to discern AI-based data so they can choose how much they want to depend on it. In other words, people will feel better using your product if you define AI explicitly and explain what it is not.
That’s why our advice is to make sure users are aware of when the system delivers information that was created by AI and when not. This way you prevent people from misunderstanding your product.
For example, Zendesk uses AI to analyze support requests and identify the most typical customer issues and their degree of satisfaction. And Zendesk clearly states where it’s a prediction made by AI, so that users can form some expectations about a certain support ticket.
3. Tell users how AI works within your product.
AI performs a huge number of tasks that can be overwhelming for one person. However, users need to understand how machines operate. In this case, they will be able to trust them and use them more effectively.
Usually, people don’t need to know everything. Sometimes full disclosure is even less helpful than just giving the key pieces of information that are most important. So, through doing research, testing, and iterating you will come to know what explanations about AI functions your customers need to trust your product.
In most cases, those pieces of information should include sources of data used by AI: inputs, outputs, capabilities, and limitations.
Here’s how Netflix explains to its users why they recommend certain videos.
4. Design for failure.
Even if you are 99% sure what your customer’s choice is gonna be, give them extra variants. Make some room for a mistake. Always include an option to reject or reverse any recommendations or modifications the system makes.
A simple example may be that when using chatbots for your business, you must provide a possibility for users to switch to humans at any time. Sentiment analysis should automatically connect the user to a human when it notices that they are furious. Forcing the user to interact with the AI when they are not satisfied with the answers it gives, will simply make them abandon your solution.
5. Set the right expectations.
People will anticipate that your AI will be either more intelligent or less intelligent than it actually is. That’s why it’s important to set proper expectations about what the machine can do and how accurate its work can be from the outset.
Users will eventually discover the ideal ways to incorporate AI into their daily tasks. But you have to be as clear as possible when describing the capabilities and restrictions of your AI to gain users' confidence from the very beginning.
Here’s an example from our experience. Stradigi AI is a SaaS AI business platform provider that helps companies automate their business processes using the power of machine learning (ML). As a part of our work on this project, we had to design UI and UX for one of its new features that allow to create and approve predictions generated by the ML model.
To help users understand the quality of models they create, we added a recommendations block that explains how accurate the prediction is and how users can improve the result they got.
6. Traditional user testing doesn't work.
It can be significantly more challenging to test the UX of AI products than it is for other types of applications. The key selling point of these products is that they offer an experience tailored to specific needs, but you can scarcely simulate that at the early stage of design when the AI system is not fully built and you’ve got only rough prototypes.
Currently, Wizard of Oz and personal examples are two methods that have proven to be the most effective in testing AI products.
Personal examples testing involves asking participants for some personal data, for example, movies they like (don’t forget to explain to the individual why you need this information). You can then use this information to simulate right and wrong recommendations from the system and see how users react to them.
Participants in the Wizard of Oz testing engage with what they think is an autonomous system, but is actually being managed by a person. It’s widely used to test chatbots: the user thinks they are communicating with the chatbot, while it’s your colleague who gives responses.
7. Don't forget about the feedback.
The user experience in AI products is improving as we continue to feed more data to the algorithms. UX designers can't test every bad experience. Therefore, it is necessary to ask users what they think about the content generated by AI in order to improve the performance of machine learning.
Give users the option to quickly provide feedback on any screen where the app shows AI content. Usually, it’s enough to place a single-tap feedback option right next to an AI-generated recommendation or prediction.
To sum up
Design is evolving together with new technologies. And these innovations bring new challenges that we have to learn to overcome. As designers, we know that all these problems are contextual, and may differ from project to project. Still, there are certain questions we have to address again and again when talking about AI for UX. That’s why we hope that UI/UX principles we’ve covered in this article will help you create better user experiences. And if you need help from professional UI/UX designers, don’t hesitate to contact Eleken.