01
Semantic layer as ground truth
All metrics defined once in dbt with owners and definitions. The model can only query pre-approved measures and dimensions — no free-form SQL against raw tables.
Products / AI showcase
Analytics
Conversational analytics
Ask business questions in plain English and get trusted charts, cohorts, and forecasts backed by your warehouse's semantic layer.
- Client
- DTC apparel brand
- Industry
- Retail / e-commerce
- Duration
- 8 weeks
94%
of ad-hoc requests self-served
0
conflicting definitions in Q1 review
8 wks
from kickoff to daily use by leadership
01 · Problem
Where the team was stuck.
- 01The analytics team was a 3-person bottleneck for a 90-person company. Ad-hoc requests took 2–5 days.
- 02Non-technical teams built parallel spreadsheets from CSV exports — numbers drifted between teams within a week.
- 03Leadership wanted daily cohort views but couldn't justify pulling analysts off strategic projects.
02 · Approach
How we built it.
Stack
- Gemini 2.5
- dbt semantic layer
- BigQuery
- Cube.dev
02
Show the query, always
Every answer displays the underlying metric definition and filters. Users learn what the numbers mean instead of taking them on faith.
03
Forecasts with guardrails
Forecasting is a specific tool call, not a hallucinated line chart. Confidence intervals and method (Prophet, moving average) are labeled on every projection.
03 · Outcomes
What changed.
The analytics team went from a request queue to a platform team. They now own the semantic layer and coach domain teams, while the business self-serves 90%+ of the questions that used to sit in Jira for a week.
- 94%of ad-hoc requests self-served
- 0conflicting definitions in Q1 review
- 8 wksfrom kickoff to daily use by leadership
04 · Demo preview
Ask a business question
Head of merchandising exploring a return-rate anomaly.
Book a live demosession preview · illustrative
User
Why did returns spike last week for outerwear?
· Measures used: returns.rate, returns.count · Dimensions: category, sku, reason_code · Filter: last 14 days vs prior 14
AI
Outerwear return rate rose from 8.1% to 14.7% (Nov 25–Dec 1). 72% of the increase comes from a single SKU — the merino puffer in size M. Top reason code: 'runs small' (61% of returns on that SKU vs 12% baseline).
User
Show me that SKU's return rate by size over the last 8 weeks.
· Rendered chart · exported to a dashboard tile named 'Merino puffer return rate — sizing watch'.
More AI case studies
All products →Support Assistant
Customer support copilot
A grounded assistant that answers customer queries from product docs, ticket history, and account data — with graceful human handoff.
Read case studyEnterprise RAG
Document intelligence
Search, summarize, and cite across contracts, SOPs, and regulatory filings — grounded in a permissioned knowledge base.
Read case studyAgentic Automation
Operations agent
Multi-step agents that triage inbound email, update records across systems, and orchestrate approvals with clear audit trails.
Read case study