Adobe AI Support Chatbot Redesign
Redefining the implementation of AI within Adobe’s support chatbot feature
Date
Summer 2024
Project Overview
Redesign Adobe’s AI support chatbot interface to accommodate for less technology-inclined users and make simple tasks a breeze — as they should be. Utilized Figma and UXPressia.
The Problem
Artificial Intelligence chatbots are becoming more common on most websites — especially on websites that offer subscription services. 35% of people use chatbots to resolve complaints or obtain detailed information (Backlinko). This number is surprisingly low, proving that people either don’t trust or don’t have positive experiences with AI assistants.
Despite their noble purpose, most of these chatbots are poorly integrated and more hurtful than helpful. They almost always direct users to a support team member and waste users’ time. More often than not, this is due to their poorly designed interfaces that do not lower the barrier of entry for users unfamiliar with utilizing and prompting AI. For being a leader in the technology and design space, Adobe’s integration of their AI support chatbot is shockingly poor. This project focused specifically on the touchpoint of canceling a subscription — a long and convoluted process in its current state. The current chatbot simply provides the user a link to text instructions on how they can go about canceling, doesn’t mention any specific charges or restrictions that may apply, and is overall far too broad in the responses it provides.
Let’s take a look at the old user experience journey map to identify some of the biggest pain points.
The Solution
As you can see from the journey map above, the old implementation leaves a lot to be desired. It leaves users frustrated with unsolved problems. Oftentimes, they have to resort to contacting Adobe support through email or requesting a live assistant which could take a lifetime to receive a response. This redesign is meant to fill these gaps, making the tool easier to use for users at all experience levels, more intuitive, and much faster for advanced users.
Humanized AI Personas
Firstly, to lower the barrier of entry for using the feature, I had the idea of using faces and building personas for a few different support bots, allowing users to choose their favorite and/or most specialized assistant for their task. The most off-putting aspect of AI technology for new users is the idea of communicating with a robot. By attaching human faces, names, and personalities to each virtual assistant, these users will be more inclined to interact with the support bot in a more conversational manner, ultimately making their experience smoother and far more natural.
Additionally, by assigning different specialties to each persona, users will feel more confident with the assistance the AI support bots provide. Training each model on specific tasks will also improve the usability and helpfulness of each bot for certain tasks.
To implement this into the experience, upon opening the feature, users are presented with 4 choices of assistants and prompted to choose the one that best fits their needs. For edge cases, where users select one that is not best suited for their issue, they will be redirected to the best fit by the support bot after it assesses their request.
Opening Message & Response
To highlight some of the exact points where the current feature falls short, I have included some images of the old user interface below followed by the new redesigned interface, starting with the opening message and response.
As you can see, the feature prompts the user, but fails to give them any guidance in how to answer or what to ask. This could leave new users in the dark and make advanced users provide unnecessary depth, wasting time.
To combat this, the redesigned feature includes quick prompts underneath the original message to guide new users and quickly assist advanced users. Users also have the option to type a message response, making sure they can resolve their issue if it is not one of the generated options listed.
Cancel Redirection vs. Walkthrough
One of the most beneficial parts of this redesign, is the confinement of task completion within the chat interface itself. The old support bot directs the user to a link where they can manually cancel their subscription. The chatbot does not provide the user with any additional assistance, tips, or important information they may need to know before they begin this long process.
This is one of the biggest downfalls of the old implementation. Rather than assisting a user through the entirety of the task, it simply redirects them to another page where they will most likely encounter more issues.
By actually walking a user through the cancellation process and doing the work for them, the burden on the user is lessened and is placed on the software instead — simultaneously reducing the user’s short-term memory load.
By providing users with alternate options and multiple stages of confirmation, it ensures they will always reach their desired goal and understand all of the policies and restrictions that may apply to them. Providing clarity and being transparent with users is imperative, especially when paid subscriptions are involved.
Additional Closing Prompts
Finally, before the user exits, the chatbot provides more generated options in the case a user needs more assistance with anything. This makes sure the user is actually finished with all of the tasks they intended to complete and opens up space for them to ask additional questions if desired.
Journey Maps
The journey map for the new experience alongside the respective map for the old experience are presented below, side-by-side, to highlight the improvements made to the user experience through this redesign.
Results
Through this project, I gained a new understanding and appreciation for designing with Artificial Intelligence in mind. AI can do so much, but can also confuse users if not implemented properly. Prompting users and providing options to keep them on the right track is crucial.
On a separate note, this was my second time trying to design a chat-based interface and I definitely feel much more confident and excited to create these types of interfaces in the future!
What’s next for this feature? As I mentioned earlier, Adobe is one of the biggest players in the design and technology field, therefore I hope this redesign of their AI chatbot window will prompt necessary change within the space. I think it is a great place to start and sets a precedent for the redesigned experience of the feature as a whole as well as other companies’ AI support chatbots.
