Blog: User research for AI: What we’ve learned about researching and designing AI assistants
The virtual assistant market is set to sore in value to USD 19.6 billion globally by 2025.
Now, we’re no fortune tellers, but it’s inevitable that this will lead to a similarly huge amount of growth in the number of people and businesses seeking to get their slice of the AI Assistant market! Whilst we don’t see the front-runners like Amazon’s Alexa and Google’s Home Assistant going anywhere any time soon, we do suspect that more and more AI assistants will become available in the near future.
Competition is only going to get more fierce. As such, if you’re looking into creating your own AI assistant, it’s imperative that you conduct incredible research that will give you the edge.
With that in mind, we wanted to share some of the biggest lessons that we’ve learned whilst both researching for AI Assistants, and building them based on that research.
What we have learned whilst researching AI assistants
User and market research are completely necessary for the success of any product or service, but especially for those that utilise emerging technologies. Both types of research will allow you to better understand the wants and needs of your users relating to something that is still relatively new to us all.
Related: Snap Out’s user research services
Here are the key questions to keep in mind throughout the research process.
How are people currently solving their problem?
So, you know what problem your product or service is trying to solve. That’s great.
But we’ve learnt that one of the most important things to research is how your users are currently trying to resolve that issue. Answering this question and looking at the existing habits of your potential customers will allow you assess whether an AI assistant will actually help them.
If the assistant doesn’t fit almost seamlessly into their current habits, only requiring slight behavioural changes, adoption rates may not be what you would hope.
Related video: Understanding Behaviour Change
Is this “tech for the sake of tech”?
Again, we hugely recommend researching your user’s current processes. Whilst it may be possible to insert an AI assistant into this process, that doesn’t necessarily mean it will be helpful or needed!
In our experience, you should never start with a focus on the technology. Instead, always start with a focus on the user. This will help you to understand whether technology is actually the best solution to their problem, or if you are just building tech for the sake of tech.
What are the users’ expectations of AI assistants?
As with most emerging technologies, users are still learning how to best use them. We, as users, all make assumptions about the products and services that we use. Therefore, it is crucial to understand what those expectations are for AI assitsants.
Founder of Snap Out and Emma.ai, Aaron Mason says that,
“What has been key for us has been aligning the expectations of the user with the capabilities of the AI assistant.
It is important that the user experience sets realistic expectations. AI performs exceptionally well at some tasks and poorly on other tasks and products need to be designed in such a way that exploits these challenges and opportunities. You can think of this as the end user and the AI assistant collaborating on a task and understanding how best to work together.”
What questions will user research answer?
Having realised the above, we have taken note of the most important questions that user research for AI assistants should be answering.
- Will the AI assistant genuinely add value to the individual and save them time and effort?
- Will an AI assistant enable them to do what they want in the way they want to?
- How can we best align the expectations of the user with the realistic possibilities of the technology?
What we have learned when designing AI assistants
Following successful research, the design process will be a lot less daunting. However, that certainly doesn’t mean it will be easy!
Here’s what we recommend keeping in mind throughout the design phase of your product or service.
It’s all in the data
The amount of data that AI assistants require simply to function is enormous. The amount of data that they need in order to be useful is almost unimaginable!
After all, it’s one thing for a bot to perform a basic task, but it’s an entirely different thing for them to do so in a way that is quicker and/or more effective than the human brain.
Consider Alexa, for example. Amazon’s infamous assistant has in excess of 3,000 “skills”.
Whilst it’s certainly true that not all assistants need to be so advanced, this statistic gives you insight into the huge volumes of data that AI in this form requires.
You might not have a business model
Whilst an AI assistant may help people — meaning that it saves them time and/or effort — that doesn’t necessarily mean that it will be successful. Sometimes the business model just isn’t there and brilliant technology is therefore unable to hit the mainstream.
When it comes to designing an AI Assistant, think about how you would commercialise it and validate it with others. If this isn’t possible, moving forward may be a waste of your time, energy and resources.
Always protect your users
AI assistants have access to a lot of personal, and sometimes extremely sensitive, information. It is of the utmost importance to protect users’ data.
From an ethics point of view, you hold a huge amount of responsibility as the creator of an AI assistant. They are powerful tools that have the potential to change lives for the better, but they could certainly be used to do the opposite. It is your moral obligation to ensure that you are not misusing your users’ data or handling it in an unsecure way.
From a marketing point of view, more and more people are aware of how their data is being used. In the wake of GDPR changes, your customers need to trust that you will protect them before they invest their money and time into you.
If there’s one thing that you should take heed of within this article it’s this: Always start with your user.
Basing your product or service design on the fact that you have a powerful piece of technology, or on your own personal pain, is not a recipe for success. Instead, ask whether your user truly needs an AI assistant and how it will actually help them in their everyday life.
If there is a need and a want for the product, continue to focus on the user throughout the building process. Focus on how you will entice your potential customers through commercialisation, generating trust and effectively utilising data.
What do you think the future of AI assistant looks like? We’d love to know in the comments.
If you enjoyed this article, don’t forget to give it “claps”!
The Snap Out Team 🚀