GitLab Duo Acceleration Designer Resources

What is this issue about?

This issue is a resource for team members assigned temporarily or permanently to a focus on AI. We’ve gathered information for you here so that you know who to work with, how to get help, what research support is available to you.

In the AI space, the product design team started off strong (more than one year ago) by collaborating on design patterns. Since then, we haven’t had the opportunity to take this work to the next step and give our AI tools the attention and refinement they deserve. 

Now we are ready to change this and ensure our next releases meet user needs and expectations. This is an opportunity for you to do work you are really proud of and excited about. Please let your manager know right away if you encounter blockers that would impact your ability to do your best work. Since we can’t anticipate exactly what each capability will need in terms of UX support, you should also let us know if you need something different from what’s in this issue, so that we can support you. 

The importance of collaboration: to achieve a cohesive experience

Currently, our AI features are a bit disjointed, since different teams worked on them at different times. As a UX team, we have the opportunity to lead to better user experience by taking a holistic approach. So please over-communicate your designs and work with fellow product designers to identify and document patterns. Share the questions and challenges you have to see how others are solving them. Please share articles you learn from and examples of design inspiration from other products in this space. AI is still very new and the interaction design patterns change all the time - so we want you to learn and have fun with it!

Getting Started

1. Get to know the AI capability you are working on and how it fits in the bigger picture by reviewing: 
   1. The AI capability you are working on, as well as those that are similar to yours. For example, if you are working on a feature that helps users troubleshoot, inquire about other features that enable users to troubleshoot.
   2. GitLab Duo docs
   3. The heuristic evaluation and Figma walkthrough for your capability and similar ones
   4. Pajamas AI documentation
2. Get to know your team if you haven’t worked with them before. 
   1. Coffee chat and/or working session with the PM as well as team introductions.
   2. Collaborate with PM to triage the UX issues to work on, including group issues and issues created as a result of the heuristic evaluation. Feel free to create or suggest additional issues to improve the area you are working on. You and your PM should use your best judgment and expertise to select the most impactful things to work on.
   3. Add UX weights to issues and level set on what you can deliver. If you and your PM have any concerns regarding capacity and expectations (for instance if there is a delivery date you feel you can’t meet), let your manager know right away. 
3. After this, you can get started!
   1. Formulate your viewpoint for what your AI capability offers users in terms of value and how to maximize that. 
   2. You can use the Competitor Eval project and this AI Inspiration Figma file, to find and share information about how others are solving similar problems.
   3. As you identify reusable patterns or components, ensure they have corresponding issues in Pajamas. Work with Katie and other designers on refining and validating these patterns so others can use them after you.
   4. Please ensure your design issues have the label ~”AI-UX” and that you are assigned to them, so that we can find them. Also, please save your Figma files in the AI Integration directory.
   5. As early as possible, identify research needs you anticipate having. See below for info on doing UX research.

Communication

1. If you haven’t already, join #ux_ai_integration. This channel can be used similar to ux-coworking for async design feedback, design inspiration and any other AI related UX conversation. 
2. Check the UX calendar for AI UX feature maturity: office hours and plan to participate either sync or async. This is an optional meeting but it’s an excellent place to get visibility into what others are working on, and see where we are working on similar problems. If the time doesn’t work for you, please reach out to Jacki. We can add additional times. 

Additional Information

• The UX Research team has put together a plan to support your research needs. We want to help you get the feedback you need from users as efficiently as possible. In turn, we ask that you start creating and filling in research issues as soon as you can. Highlights of the plan:
  • Strategic research: You can add your research questions to FigJam via this issue.
  • Usability testing: We’re working on accelerating recruiting and test environments so you can get feedback on prototypes or current functionality in production.
  • Gen AI evaluation: A chat bash is a user research method we can use to evaluate AI experiences, that is, where the outcome is indeterminate because we can’t predict how the AI will answer the user. In a chat bash, we ask participants to do a task several times over a span of time, and add a diary entry each time so that we can capture multiple data pointes. 
• UX Metrics: We should measure our output to ensure it meets user needs. Additionally, AI capabilities should meet these expectations prior to being GA. Some AI capabilities do have target dates, so work with your PM to understand if a target date will impact your ability to meet baseline UX metrics.
• Definition of Done for GA features
• Release schedule for Duo Enterprise: Bookmark this epic as a starting place to see if and how your AI capability fits in to plans for Duo Enterprise.
• GDK Setup Help:
  • The instructions can be found here -https://docs.gitlab.com/ee/development/ai_features/#get-started
  • If you get stuck, you can ask questions in the Slack channel #g_ai_framework_lounge
• Design resources and inspiration
  • Pajamas AI documentation links to two resources that can help designers up-level their knowledge very quickly about designing for AI:
    1. People + AI Guidebook, Google
    2. Human-AI eXperience (HAX) Toolkit, Microsoft
  • Check out our AI Inspiration Figma file and The Shape of AI.

Who’s Who

@katiemacoy is the UX DRI for AI Framework. Katie will organize and lead work in this space, which is currently defined as:

• Assisting designers with maturing patterns common across AI Workflows
• Identifying new components and guidelines needed
• Providing support to designers on how to identify AI use cases
• Collaborating with UX research to determine best practices for evaluating chat-like experiences

@NickHertz is the point of contact for guidance on UX research methods. Nick has experience evaluating Duo Chat and how users are finding the answer quality and usefulness. He can help you decide the best way to evaluate your chat-like experiences. 

@jackib has the overview of GitLab Duo UX efforts. If you need any help getting started, finding issues, understanding scope, etc, you can ask Jacki.

This table shows the team working on this platform wide effort. The "AI Themes" column can be used to identify designers who might be solving similar problems.

Design DRI	Group	PM	Responsibility/Scope	Notes/Updates	AI Themes
@emilybauman	Multiple	@tmccaslin	Unified AI Settings - turning Duo off completely - experiment/beta toggles - closer workflow with seat assignment - IA scalable for custom models	Katie has been working on this and will work with Emily to transition	AI Settings Adoption
@aregnery	Multiple	@rogerwoo @dashaadu +TBD	Adoption paths, including End user - is AI working in my editor? Admins of SM - is AI set up for my instance? addition onboarding at individual feature or project level Support growth team work related to GitLab Duo trials Support work related to GitLab Duo post-purchase adoption?(TBD between Austin and Tim)	Added to scope Supporting the trial widget	Adoption & AI Settings
@timnoah	Fulfillment Custom Models	@susie.bee +TBD	Adoption scope - Address gaps in full adoption path from purchasing to usage, especially Fulfillment related, seat management, and model selection (TBD here between Austin and Tim) Support Custom Models, especially self-hosted models to unblock the team	Confirmed that we're confident in designs for bulk seat management	Adoption & Custom Models
@katiemacoy	Multiple	NA	Design work to provide patterns common to all Duo Enterprise features - Ensure that similar workflows are consistent across different teams doing implementation - Figure out how this works with rollout approach	Assisting with design of patterns common across AI Workflows Identifying new components and guidelines needed Providing support to designers on how to identify AI use cases Collaborating with UX research to determine best practices for evaluation chat-like experiences	AI Design Patterns
New hire (starting June 10)	Duo Chat	@tlinz	Duo Chat		Chat
@beckalippert	Threat Insights	@abellucci	Vulnerability explanation and resolution		Troubleshoot Generate code Suggest content Chat
@veethika	Pipeline Execution	@rutshah	Root cause analysis		Troubleshoot Generate code Suggest content Chat
@mle	Code Review	@phikai	Code review experience Automated Merge/Squash Commit Messages Code explanation		Summarize Generate code Suggest content
@nickleonard	Project Management	@gweaver	Discussion Summary (or, just summaries) Description Generation potentially, templates		Summarize Suggest Content Chat
@lvanc	Optimize	@hsnir1	AI Impact Dashboard MVC and next steps		Analytics
@tvanderhelm	Editor Extensions	@dashaadu	Code Suggestions and Editor Extensions Test generation	Specific goals/starting point: Duo parity between VS Code and JetBrains (code suggestions + chat) decide on best design for "take something from chat and insert it to code"	Generate code Chat
@pedroms	AI Framework	@pwietchner	Duo Workflows		Actions
@mnichols1	Duo Chat	@tlinz	Context patterns for Gitlab Duo + Amazon Q POC for Duo Chat while Nicolle onboards	As a first step, this area needs to be defined and scoped more clearly.	Actions
@grrbam	Model Ops	@kbychu	Model registry	We have a role open for this space, so Graham will cover it until we hire this person.	Settings/Other