After completing another project, I gathered feedback from 15 users about the sign-up process and their survey-building methods. I observed that many users had pre-set questions but still had to manually build each question, so I asked them about this issue.

Key Insights:

• "You know, creating surveys can be quite a process. First, I have to draft the questions and get them approved by different departments to ensure they align with our standards and goals so often we already have a word doc to begin with that is pre-determined”.
• ”I think thats were a lot of my time goes because once it’s built all you have to do is distribute it which is easy”
• ”I find Creating surveys involves a lot of back-and-forth. Our questions are usually reviewed by various departments for approval, and then build it using SmartSurvey . It’s a bit of a hassle. If there was a way to upload a document or something then perhaps that would be easier - I think Qualtrics used to let me do that.”
• "Deciding on the right question types can be tricky for us. We receive templates for surveys to create, but sometimes we're uncertain about which questions will work best."

Ideation session board

Problem Defenition

The challenge lies in efficiently integrating pre-built surveys into our platform. Users face uncertainty in selecting optimal question types and encounter delays in obtaining necessary approvals from stakeholders. Our goal is to streamline this process, ensuring surveys are user-friendly, align with organisational standards, and meet user expectations seamlessly.

The ideation workshop

During the ideation session, I defined the problem and presented my findings to the team. We then generated ideas and refined HMW (How Might We) statements to ensure our solutions were focused on the user's needs.

Chosen path

We refined and prioritized our ideas, then used an impact/effort matrix to evaluate their value and required effort. We decided to use AI for several reasons:

• It was scalable and consistent - AI can handle a large amount of surveys and is incredibly accurate.
• Cost efficent - This minimised resources and reduced the build time - We also could integrate with our exsisting AI offering.
• It would meet the users pain-point by reducing the time taken and helping suggest appropriate question types.

Analytics

We had recently introduced an AI feature that generates question lists, providing an opportunity to automate the survey process. I analyzed AI usage data, which showed strong user interest, supporting the development of a lightweight MVP and potentially avoiding a lengthy development cycle.

User requirements:

• Implement a text box where users can copy and paste pre-built survey content.
• Develop an AI system that analyzes the pasted text and generates corresponding survey questions and formats.
• Enable the AI to recommend the most suitable question types (e.g., multiple-choice, open-ended) based on the content of the pasted text.
• Include a feature that allows users to regenerate or modify questions if the initial AI-generated output is incorrect or suboptimal.
• Implement a feedback mechanism where users can rate the accuracy of the AI-generated questions.
• Provide a real-time preview of the generated survey questions as users input text.

In order to create an MVP we decided to add onto the current AI flow for prompts as this meant it was easier for users to discover, would not interrupt squad work and meant we could tap into a steady stream of users already using the product.

Having discussed requirements, how we could utilise it in the current platform and establishing the rough flow meant I could start with the initial designs.

Initial Designs

From this point I then reached out to previous users to get their initial thoughts and responses on our proposed solution and was met with overall positive responses altohugh it did give us more to consider which helped mitigate later issues.

• "The import feature looks like it would be really efficient. Being able to paste questions and see them instantly in the survey preview seems like a huge time-saver."
• "The design looks straightforward, but I wonder how well the AI handles more complex questions. If it struggles, users might need to spend extra time adjusting things manually."
• "The design seems straightforward to me, and the tips provided on the side add some support. They seem like they could be useful and I like having pointers on how to get the most out of my questions"
• "It’s great that the AI can format the questions, but I’d be concerned about it changing the intent. Maybe an option to preserve the original question type would be useful."