Team
1 Product Designer (me)
1 Product Manager
2 Backend Developers
1 Frontend Developer
Stakeholders (Analytics team)
Contribution
UX Research
User testing
High-fidelity UI
Developer Handoff
PROJECT OVERVIEW
Innovaccer is a US healthcare data unicorn providing better care delivery through their SaaS data platform with analytics at the heart.
I worked with their analytics product team as a solo product designer to ship an AI conversational interface that allowed non-technical users to ask questions in plain English and instantly get insights - reducing turnaround times from 2 weeks to 3 minutes and enabling faster, data-driven decisions.
PROBLEM CONTEXT
Innovaccer had powerful analytics capabilities, but the analytical complexity limited adoption. The main users of dashboards were non-technical healthcare leaders who struggled with widget-heavy dashboards and required support from analysts to interpret.
72%
of non-technical users depended on analysts to extract insights
3 weeks
Average turnaround time for a custom request.
65%
customers reported needing customized assets
OPPORTUNITY
How might we empower non-technical healthcare leaders to access and act on data insights instantly, without relying on analysts or navigating complex dashboards by leveraging AI?
DESIGNING FOR AI
The design team had set up principles for incorporating AI into the products which guided me along the way, not adding AI for the sake of it, but to actually add value for the users and the business.
Problem first approach
AI shouldn’t be used just because it is trendy. It should be solving a problem first.
Empower, not replace
AI should not give clinical advice/offering but empower the trained healthcare professionals and not replace them.
Automation
AI can significantly increase the productivity by automating manual tasks.
Insights
AI can analyze a vast amount of data in short duration which is humanly not possible.
RESEARCH
Tools like Mixpanel helped me understand the usage of different features and drop-off points but it only took me so far. I needed to understand what was missing through the people closed to the customer, the customer success team.
I conducted 1:1 interviews with the customer success team to uncover their painpoints as well as about the most asked customizations and what caused the most delays in their journey.
Further collaborating with product managers and data scientists to map the workflow helped highlight where bottlenecks occurred and how long decision-making cycles stretched as a result.
MAPPING INSIGHTS TO INTERVENTIONS
Based on these insights, we shifted the analytics experience from a dashboard-first model to a conversational interface. This gave non-technical users a simple way to ask questions in plain English and receive immediate, contextual insights.
#1
High barrier to entry
Non-technical users found dashboards too complex and often relied on analysts for reports.
arrow_right_alt
Conversational querying
Natural language interface for questions
without tackling the complex data directly
Overwhelming interface
The dashboard was cluttered with widgets and charts, with no clear starting point for actionable insights.
arrow_right_alt
Guided prompts
Suggested questions to help users get started
and discover possibilities.
#2
Slow turnaround
Customization requests often took 3 to 4 weeks to complete, making insights irrelevant by the time they were delivered.
arrow_right_alt
Contextual customizations
Dynamic data visualizations - charts, KPIs, tables generated directly from queries, with summaries
KPIs
KPIs help track important metrics
Charts
Charts help understand trends over time, category comparisons
Tables (Cohorts)
Cohorts are targeted patient lists which can be filtered
#3
"I would like to confirm and validate if these numbers are correct…"
arrow_right_alt
Required transparency to build trust with AI
Show the working steps behind the scenes (also helps the delay seem shorter)
Educate the user on how the model is powered
Different levels of explanations for different users
COMPONENTS
IMPACT