Hevo Answers - Conversational Analytics for Everyone
Empowered business teams with instant, natural-language access to their company’s data → +38% adoption in just 6 weeks
Company
Hevo Data
Team
UX Lead (My Role), EIR, UI Developer, Backend Developer
Stage
0→1 MVP
Platform
Web Dashboard (Internal Pilot)
Background
Hevo is a modern ETL platform that helps companies sync data from 150+ sources into their data warehouse in real-time. It enables data teams to centralize scattered data but insight delivery still remained a bottleneck for business users.
This gap between data centralization and data accessibility became more pronounced as the org scaled.
📊 Impact
+38%
Increase in weekly active users across Sales, Ops, and Marketing teams
2.8
Average chat sessions per user per week during pilot
3m 30s
Average time to first insight from question to complete answer
25%
Reduction in analyst hours spent on repeat queries
❓ The Challenge
Business users especially non-technical teams struggled to get quick insights from data. They relied heavily on analysts for even basic metrics, slowing down decisions and adding frustration.
Key signals:
38% of users reported delays >2 days for basic metrics
60%+ of business queries were repeatable and simple
Data team was overwhelmed with ad-hoc asks
🧱 Starting Point
There was no existing interface. Users had to rely on:
SQL dashboards → overly technical
Static reports → stale and delayed
Slack/Data team → inefficient handoffs
The Current User Journey
Meet Priya, a marketing manager. She wants to understand how her recent campaign impacted revenue.
Using the Hook Model (Trigger → Action → Variable Reward → Investment),
here's how Priya’s experience unfolds
Personas Involved
The Opportunity
Hevo already pipes structured data into warehouses.
We asked: What if users like Priya could just ask questions in plain English and get back answers, charts, and explanations instantly?
That was the seed of Hevo Answers.
🧠 Hypothesis
Can we build an AI-powered data analyst that enables business users to...
🔍 Ask questions in plain English
⚡ Get real-time answers from their warehouse
🧾 View the result, SQL logic, and summary all in one place
This became the core of the MVP
🎯 The Vision
To enable business teams to explore their data as naturally as they think.
Hevo Answers was envisioned as a trust-first, explainable AI experience focused not just on speed, but on making data feel accessible, auditable, and collaborative. Our long-term goal: empower every CX, Ops, or Marketing lead to self-serve insights with confidence.
Design principles I followed:
Speed First
Reduce the time it takes to get insights from days to minutes by eliminating dependencies on analysts.
Build for Trust
Ensure AI interactions are interpretable and safe, so users feel confident relying on the system.
Explain Everything
Provide full transparency into how results are derived show the interpreted question, SQL query, and output.
Enable Self-Serve
Empower non-technical business users across functions to independently ask questions and get answers from warehouse data.
🔍 Process Highlights
Research & Discovery
I conducted the following activities:
Spoke with 12 business users across Ops, Sales, and Marketing
Mapped repeat queries like “Top sources by revenue last 7 days”
Analyzed 300+ Slack tickets to spot query patterns
I uncovered the following insights:
Users asked the same questions repeatedly
Analysts were overburdened by non-complex requests
Business users lacked awareness of what data was available
Competitive Analysis
Iteration, Refinement & Testing
What I validated (through internal pilot testing)
Business users were able to self-serve without analyst support
Prompt suggestions improved engagement and repeat queries
Clarification prompts reduced user frustration during failed queries
Users rated onboarding clarity 4.6/5 in internal feedback sessions
What I iterated on (based on feedback)
Introduced fallback prompts when AI couldn’t interpret questions
Added prompt suggestions to reduce blank-state friction
Designed a YAML-based model browser for transparency
Tuned visual hierarchy in answer cards to make SQL more scannable
🌍 Strategy & Differentiation
Hevo Answers wasn’t just another AI overlay:
B2B Schema Awareness (unlike generic tools)
Explainability-first UX (unlike black box models)
Plug-and-Play Simplicity
Positioned against Seek AI, PetaVue, and Cortex AI, we focused on transparency over magic.
✨ UX Highlights (Solving Real User Frictions)
Onboarding Assistant
Solves: Blank state anxiety
Delivers: Confidence from the first interaction
Impact for Priya: Reduces hesitation and gives a clear starting point
Prompt Suggestions
Solves: Blank state anxiety
Delivers: Confidence from the first interaction
Impact for Priya: Helps frame useful questions without needing SQL knowledge
Explainable Answer Cards & Foot Note
Solves: AI trust issues
Delivers: Transparency by showing SQL + reasoning
Impact for Priya: Understands and validates results with confidence
Semantic Model Browser (YAML)
Solves: Schema misunderstandings and logic disconnects
Delivers: A source of truth between business and data teams
What is YAML?
YAML (YAML Ain’t Markup Language) is a human-readable configuration language often used to structure complex data. For Hevo Answers, we chose YAML to define semantic models because it acts as a bridge between technical setup and user-facing design decisions.
Impact for Priya: Makes data structure and availability visible
Fallback Clarifications
Solves: AI hallucinations from vague input
Delivers: Resilience and learning through guided clarification
Impact for Priya: Learns how to ask better questions without feeling lost
Some Customer Testimonials
"I used to wait two days to get answers. Now I can ask a question and get a clear breakdown in seconds. It’s like having an analyst on call."
Senior Marketing Manager
"Our analysts spend less time pulling repeat reports and more time doing real analysis. That’s a huge win."
Senior Marketing Manager
"The fact that I can see how the AI built the SQL behind my question builds so much trust. I finally feel empowered to explore data myself."
CX Lead
✅ Outcome & Learnings
Internal MVP shipped within 5 weeks.
Onboarding tested with internal business teams scored 4.6/5 on ease-of-use.
The model browser saw strong engagement, users explored models before chatting
We learned that speed alone isn’t enough. Trust and clarity are what drive adoption.
"Explainability isn't a feature, it's the product."
🚀 What’s Next (Post-MVP)
These learnings directly influenced our next steps
Implementing a schema change tracker within the semantic model editor to prevent outdated models from causing errors.
Introducing alerts for stale or broken models based on underlying schema drift.
Enabling inline feedback on individual AI responses to help analysts correct and improve response quality.Prioritizing error-resilient design and fallback behavior for unpredictable queries in the upcoming release cycle.
Adding version test environments for analysts to simulate how the AI will behave before publishing changes.
📝 Final Reflection
Hevo Answers fundamentally reshaped how I approach AI-driven UX design. It taught me that speed means nothing if users can’t understand or trust what they’re seeing. In a world of black-box automation, our responsibility as designers is to build clarity into every interaction and empower users with transparency not just outputs.
This project wasn't just about crafting interfaces it was about shifting mental models. Designing for ambiguity, handling failure gracefully, and illuminating the "why" behind each answer became non-negotiable. And through this, we created a foundation where trust could scale as fast as the technology behind it.
"Now I ask a question, get an answer in minutes, and I actually understand how it was generated."
Priya’s journey transformed
From zero to MVP in five weeks, Hevo Answers became a proof point that thoughtful AI UX can unlock true self-serve analytics one question at a time.