AI Generative

An advanced tool for auto-generating support content, allowing businesses to create and launch manuals in minutes effortlessly.

MY ROLE
Product strategy
User experience
Design thinking
UI Design
TEAM
Product managers
Software engineers
ML engineers
Design team
TIMELINE
Sep 2023 - Jan 2024 Released
CONTEXT
Mavenoid leverages generative AI to automate the creation of product support content, allowing businesses to quickly scale and keep up with growing product lines, all while staying competitive in an evolving market.
IMPACT
-Churn
This solution met the demand for faster modeling, reduced churn, and provided long-term value.
+Usage
The adoption rate of AI features increased from 7% to 33% of active users.
+Insights
We gained key insights and identified value opportunities for future generative answer features through this project.
Overview
With Mavenoid, you can automate any support process for any product, moving beyond basic FAQs to advanced self-service and troubleshooting. Our nodal canvas empowers users to design these flows effortlessly.
Mavenoids flow builder page
Problem
What is the problem?Previous attempts since 2021 to use GPT for auto-generating flows and flow suggestions failed to deliver useful or usable results in production. Both generated some excitement during the sales/demo process, but in reality, their output was so poor that no one used them in production. The company is now facing a challenge of scalability, as their current mostly manual process is not feasible for large enterprise customers.
‍
Why is it a problem?The manual flow modeling process is unable to scale for large enterprises with thousands of product customers, making it difficult to meet their needs effectively.
‍
Growing competition from GPT-based bots offering faster, no-modeling solutions is increasing pressure, with promises like “launch in 1 day” that appeal to customers seeking quicker deployment.
‍
As customer expectations rise for more innovative, efficient, and scalable options, failure to meet these needs risks losing market position to competitors offering better solutions.
Solution
The solution involves revisiting the problem with a more promising approach, leveraging recent advancements with GPT-4. The company intends to:
‍
Use GPT-4 to generate flow drafts, as early experiments have shown promising results.
‍
Reintroduce AI-driven auto-generation with improvements, focusing on better scalability, efficiency, and innovation to meet customer demands and stay competitive.
Research
Through a series of interviews, users revealed valuable insights into their workflows, highlighting the need for AI to enhance efficiency while maintaining control over content, which included the following key user stories:
Thomas
Implementation Specialist

"I see AI enhancing my workflow by offering numerous solutions, user-friendly text, and well-structured flows. I would find it useful to use AI to gather content, but always with expert supervision."

Torsti
Project Manager

"AI tends to generate too much text. We follow a simplified technical English standard, which includes many rules for writing technical content for technical audiences."

Hugo
Customer Success

"I look into user feedback, usually when I identify the symptom. I rely on user feedback overall when I notice the conversations declining."

Key features
The design process unfolded in two phases. The first phase involved creating a user-friendly interface that guides users through simple steps to effortlessly generate automated flows in minutes. The second phase enabled users to update existing flows with AI-generated content directly within the canvas.
The goal was to provide easy access to the AI Generator, enabling smooth content updates.
Initial user testing revealed friction in identifying and reviewing generated content. To address this, I designed a table list with one-click access to all generated nodes.
Customers with thousands of products found manual content entry time-consuming. To address this, we designed a guided solution with a short-form modal and step-by-step instructions for auto-generating flows with no manual effort.
To guide the AI, we provided options to address common issues, allowing it to generate solutions in a controlled manner.
Phase 1: Flow auto-generation setup
Refinement options were provided to build controlled flows, including escalation paths and the ability to define which parts to generate, such as troubleshooting or FAQs.
Customers with thousands of products found manual content entry time-consuming. To address this, we designed a guided solution with a short-form modal and step-by-step instructions for auto-generating flows with no manual effort.
Ideation
It was essential to create a solution that could be easily translated across different areas of the app, seamlessly integrated, and require minimal effort for engineers to implement quickly, given the project’s complexity, while focusing on the MVP (Minimum Viable Product).
Results
The adoption rate of AI features increased from 7% to 33% of active users. For the AI Copyeditor, 18% of users engage with it, and 17% use AI Gap Analysis (flow and org-level).

While adoption of AI flow/node generation didn’t meet expectations, the functionality provided valuable insights into customer needs and paved the way for innovations like Generative Answers, coming soon.

Despite the adoption not reaching our targets, we successfully built a feature that incorporates GPT-4, offering a scalable solution that meets the growing demand for faster, more efficient modeling, while also reducing customer churn and providing long-term value.
NEXT PROJECT

Built-in Translations Localization System

See project
Get in touch!
leonkibuka@gmail.com
Designed by León Kibuka