Creating an omnichannel conversation builder that’s PCI and HIPAA compliant

Introduction

Inference had been a long-time provider of automated call software, specifically, traditional Interactive Voice Response (IVR) systems (aka archaic telephone systems, Press 1 for Sales, Press 2 for Support...) But as technology evolved, the product evolved and it was important to demonstrate this to stand out as a competitor in the space.

The goal was a complete redesign to showcase the platform's evolution from a standard voice response system to an Intelligent Virtual Agent (Agentic) platform. This redesign would allow sophisticated technologies, such as Natural Language Processing and Speech to Text to be more accessible to our enterprise customers; all while upholding industry compliance standards such as PCI and HIPAA and at the same time maintaining a 99.99% platform uptime.

As the founding designer at Inference, I was tasked with leading this significant transformation from the ground up. With no prior design framework to follow, I took full responsibility for the user experience, user interface as well as myriad other responsibilies ↗...

Working around research limitations

Because Inference's software was often sold through resellers and white-labeled, traditional user research methods weren't feasible as I didn't have direct access to end-users or customers.

To overcome this, I collaborated closely with our customer success team. We were able to collect all Studio users and their job roles and then cross-reference this with how they actually traversed the Studio platform via a sankey diagram. From here I was able to create three proto-personas representing our key user types.

Clunky and disorganized navigation

One of the biggest hurdles was untangling the complicated and confusing navigation of the existing platform. Until then, the information architecture hadn't been formally addressed, leading to a fragmented user experience.

Additionally, the redesign needed to reflect not just technological changes but also a shift in mindset. Inference was showcasing that virtual agents were analogous to human agents (and superior in many ways), and it was important to represent this evolution in the interface.

Information architecture

To address the navigation issues, I organized a collaborative card-sorting session with the customer success team and key stakeholders.

Together, we sorted the existing navigation elements and agreed on a new, more intuitive information architecture. I then referenced these categories with the three proto-personas to ensure the final navigation structure aligned with the specific workflows and needs of each user type.

The navigation layout was streamlined, reducing the number of sections from 12 to just 3. These main sections—Build, Analyze, and Manage—were directly mapped to the three previously identified personas. This design allowed each persona to access all the necessary information and tools within the platform to perform their jobs, without needing to navigate to or learn other areas of the application.

Delivering a Better Experience

The final design introduced a cleaner and more intuitive navigation system, directly tackling the confusion of the previous version. By aligning the new structure with the needs of our main user personas, users could easily find the tools and information relevant to their roles. The redesign also effectively showcased the platform's evolution to a multi-channel IVA system, allowing users to move smoothly between voice, SMS, chatbots, and instant messaging. This seamless experience empowered users to fully utilize the IVA system's capabilities while keeping their daily tasks straightforward.

Reflecting on the Results

Although we didn't have precise metrics, feedback from the customer success team indicated that as the customer base grew and platform traffic increased, the time required to build conversational flows decreased. This showed the effectiveness of the redesigned navigation and improved workflows. Moreover, the success of this redesign played a significant role in Inference being acquired by Five9 for $170+ million.

Lessons Learned

This project was a valuable learning experience, especially in dealing with the limitations of not having direct access to end-users. It pushed me to think creatively, relying on indirect data sources like customer success feedback and developing proto-personas to guide the design. The project highlighted the importance of collaboration across teams, with insights from key stakeholders being essential in shaping the final product.

Additionally, the project emphasized the need to balance ideal design goals with practical resource limitations. Some design decisions were made knowing they were "good enough for now," with plans to revisit and improve them as more detailed user feedback became available. This iterative approach ensured the design remained flexible and responsive to future needs, while still providing significant improvements.