Overview

Zoro is a B2B eCommerce company in the MRO space. They ran analytics on BigQuery and used Looker to publish human-reviewed official dashboards. The problem was everything that fell between official reporting and a full SQL request.

Roughly 20 to 25 analytics questions came in every day, mostly through Slack. Many became one-off queries that created backlog. Others got answered differently by different people, producing slightly inconsistent versions of the truth. Mid-level leaders still relied on analysts for everyday exploration and validation that should not have required analyst involvement at all.

Zoro wanted a governed natural-language analytics layer so teams could self-serve reliable answers faster, without replacing Looker or opening broad access to the full warehouse.

The Challenge

High-volume ad hoc questions created churn and inconsistency: With 20 to 25 analytics questions per day flowing through Slack, the volume alone was a problem. Questions either became one-off SQL requests that added to backlog, or were answered independently by different people surfacing slightly different answers or results. Neither outcome was acceptable for a team trying to make consistent, data-driven decisions.

Official reporting had to stay governed and reviewable: Zoro had invested in Looker as their home for human-reviewed dashboards and were not looking to replace it. Any natural-language layer had to work alongside that system, with SQL visibility and exportable outputs so teams could validate answers, and have the ability to fail safely when context was missing rather than producing a confident but incorrect result.

Security and data scope had to be tightly controlled: Zoro wanted a clean, controlled pilot scoped to a small set of approved BigQuery tables, a data dictionary, and curated definitions. Deployment options also needed to align with SOC 2 Type 2 expectations and support either SaaS or Zoro's own GCP environment.

The Solution

WisdomAI connected to Zoro's BigQuery environment and grounded natural-language answers in Zoro's existing definitions and documentation, including a data dictionary, selected LookML, and curated query examples.

The approach mattered as much as the technology. Rather than building a parallel analytics system, WisdomAI was layered on top of what Zoro already had. Metric definitions came from Zoro's own LookML and data dictionary, so answers were consistent with the official reporting teams already trusted. SQL was visible and outputs were exportable, so anyone could validate what they got. When context was missing, the system failed safely rather than generating a plausible-sounding wrong answer.

Looker remained the source for published dashboards. WisdomAI handled the questions that never belonged in Looker in the first place.

What this enabled:

Self-serve natural-language Q&A scoped to approved BigQuery tables, with SQL visibility and exportable outputs for validation. Consistent metric definitions drawn from curated LookML and data dictionary context, so answers aligned with official reporting. Admin controls, feedback workflows, and evaluation tests that improved accuracy over time. Product and business leaders got faster answers without creating BI backlog, and the pilot was designed from the start to scale from a scoped domain to broader team adoption.

The Results

With WisdomAI, product and business leaders could get answers from day one without submitting a request or waiting on an analyst. The governed setup meant those answers were trustworthy, consistent with official reporting, and auditable. The volume of questions flowing through analyst queues dropped, and teams could move faster on the decisions that mattered.

"WisdomAI's ability to inspect SQL and validate outputs made it easier for us to trust the answers and bring more teams onto the platform."

Want to learn how WisdomAI can help you get trusted answers from day one?

Book your demo today!