Top 3 AI Experience Mistakes to Avoid

By Michelle Gaseor, Experience Designer
March 15, 2019 • 4 minute read

Designing AI to solve problems, answer questions and provide efficient, user-friendly interactions requires close collaboration between business stakeholders, designers and engineers.

The start of 2019 marked the one-year anniversary of my career in AI Conversational Experience Design, a job that has only existed for a year. The first conversationally-equipped member of our Experience Design team was hired back in late 2016 — as a “Linguist.” Before the first Linguist, we had engineers writing dialogue for Amelia. Now we’re a global team teaching Amelia how to help our clients in various industries ranging from insurance, healthcare, banking and investments, to hospitality and IT support. When 2018 ended, we reflected on our work, our pain points and what we’ve learned from it all.

Designing AI to solve problems, answer questions and provide efficient, user-friendly interactions requires close collaboration between business stakeholders, designers and engineers. As we detail in this post, misalignment between these three groups will jeopardize any conversational AI project and will result in at least one of three main challenges that we encountered.

We wanted to share with you how we at IPsoft are overcoming these hurdles. We invite you to put our strategies into practice as you think about your conversational AI projects for 2019 – because the time to realize them is now.

Your chatbots’ skills should make sense to your customer too.

At the beginning of any conversational AI project, business stakeholders typically share a report of the highest-value use cases across their organization. The temptation is to crank out these top skills immediately to prove the value of your business investment.

Your ambition is admirable, but consider this: The only way to achieve that value is through your target users. For example, let’s say your analysis suggests you can introduce the highest ROI by starting with the following three skills: reset passwords, check order status on office supplies and log sick/vacation days. These three skills are completely unrelated from a user’s perspective — you wouldn’t request vacation days from an IT helpdesk or ask for an order status on your toner cartridge from HR, right? Plus, what if your user has a general question about your password guidelines or how frequently they will need to reset their password? These related questions or skills may not drive the same direct business value but they are key to making your conversational AI effective.

You wouldn’t train IT service desk agents to only reset passwords — you would make sure they had a well-rounded set of training on all closely related issues so they could dynamically pivot and engage to solve related problems. Empowering your conversational AI to interact just as cohesively with users is critical to realizing lasting business value by converting first-time users into satisfied repeat customers. To make your product more usable and valuable, reinforce your analysis with user research.

Pick a role. Create a blueprint.

You’ve decided to train your conversational AI as a customer service support specialist for your website that sells yoga pants. You have a general idea of the skills you want to have:

  • Check size availability
  • Check order status
  • Upsell matching tops
  • Check for discounts

This is a good start, and the instinct might be to start building each agreed-upon skill to start getting your product in front of users.

Here’s why you should start slow to go fast. Before any work begins, bring your design, engineering and business teams together to plan out all of the skills you want your conversational AI to possess. A role blueprint — the job description for your digital colleague — allows you to identify engineering and design efficiencies.

The human brain isn’t a linear entity and neither is the brain we’re building for your new digital employee. The 30 different things your user can do with your agent isn’t built as 30 different “things” —  foresight gives your teams the chance to accomplish 30 use cases with a few intents, a handful of entities and a few business process networks (BPNs) on the back end with integrations. Finding these efficiencies will help you build and design more quickly with less long-term rework — all while providing a more dynamic experience for users.

Sample Blueprint: Customer Service Support Specialist Role

Usability testing is your insurance policy.

The process of getting skills and roles through the concept-design-development pipeline can be like a game of telephone. You’re probably working with a lot of really smart people with great intentions and good ideas about how to build this stuff. When a project is under time pressure, the gut reaction is to push as quickly as possible toward development and see how things work when they’re built.

Here’s why you shouldn’t forget to test with your most important audience. What we (business stakeholders/designers/engineers) think users will do or interpret, and what they actually will do or interpret, are often very different. It’s cheaper to learn from a week’s worth of simulations with an image-based prototype than it is to have a failed launch that can damage your product’s brand.

Therefore, we strongly recommend you begin any project with user research and prototyping, and conduct usability testing before you start engineering. Your user research will help you determine how many skills are needed for a role to be truly helpful. Make sure your users interact positively with key design elements and learn from areas where you could make improvements. Finally, always wait to release a role to users (test users included!) until it actually solves real problems.

Have we convinced you that you can do better? We would welcome your feedback and further discussions.

Already an IPsoft customer? Feel free to reach out for an experience audit — we can help you get on track if you’ve encountered these or other pain points.


Previous Next

A Beginner’s Guide to Conversational AI

Crossing the bridge between digital assistants/chatbots and real Conversational AI requires a fuller understanding of how the technology works and its potential business value.

In our latest white paper, A Beginner's Guide to Conversational AI, we explore these subjects for companies pursuing a near- or long-term technology strategy that includes Conversational AI solutions and Digital Employees.

Download our white paper to learn how to generate business value with Conversational AI.