Key findings
- 24 accounts discuss the semantic layer, 22 of them positively, and 7 say Looker put their teams on one definition per metric.
- 17 accounts discuss self-serve exploration, 12 positively and 5 critically. The 7 that take a side on business users split 4 to 3, with the larger side saying business users can serve themselves.
- 23 accounts discuss the LookML learning curve and 20 raise it as a concern, most often from people arriving from Power BI.
- 10 accounts raise limits on charts and dashboard design, and 7 say Google has let the product stall or has no clear roadmap.
- 14 accounts discuss AI on Looker, 7 positively and 7 critically. Two of the critical accounts describe a pre-GA MCP server that would not connect to Claude's web connector until Google fixed it.
Choose Looker when one governed definition per metric matters more than fast dashboard building, and name LookML owners before rollout. In public Looker accounts from 2025 to 2026, the semantic layer is the main reason teams pick Looker, and the LookML learning curve is the most common complaint. The newest critical accounts come from data platform teams moving dashboards into tools that agents can read and edit.
Why teams choose it
Teams stay because metrics, joins, and access live in LookML, so dashboards and Explores agree. A reviewer at a small company says modeling metrics once ended the arguments about which report had the right number, and product and marketing staff now build their own dashboards. A BI architect at a mid-market company says power users get answers without going through the BI architect or a new analyst hire.
That consistency needs owners. One leader says Looker took a year to gain buy-in. A company moved business-user dashboards to Domo because only a few people could safely change LookML. A team at a large organization found KPIs modeled by different teams hard to reuse in a new Explore.
Why some leave
Teams that leave cite price and charts. They report a license cost that looked shocking next to Power BI, dashboards that drove up warehouse compute, and charts they could not build. An agency moved to Sigma, a company moved its business-user dashboards to Domo, and several commenters say Google has let Looker stall. At one 500-person company, dashboards Claude writes now draw as many weekly users as Looker.
The first AI agents on LookML
Teams praise AI on Looker for answering business questions and fault it for writing LookML. A G2 reviewer credits natural-language queries on company-wide LookML definitions, and a data engineer finds the AI question answering useful for non-technical colleagues. On the critical side, one analyst found AI tools wrote invalid LookML. Two teams found the pre-GA managed MCP server would not connect to Claude's web connector until Google fixed their instances.
Engineers building agents on Looker's MCP tools file the sharpest reports on GitHub. When a Looker MCP query returned no rows, Gemini CLI answered with invented data. An Amazon Music engineer's agent cannot ask Looker for users who did both of two things, so it rewrites Looker's SQL. An MCP Toolbox upgrade cut a data engineer's agent off from the development-mode views it had just created.
Methodology
GILD read 65 public Looker accounts published from Feb 13, 2025 to Oct 26, 2026, each in full, and tagged the topics each one discusses. They come from Hacker News, Reddit through the Arctic Shift archive, G2, and GitHub issues. The rest are practitioner and company blogs, a podcast, and a conference talk recap. Counts tally public accounts and do not stand in for a survey of Looker customers. Threads that compare BI tools draw critics and switchers. One account measures cost: a data platform team's eight-month bill for replacing part of its Looker use. Every quote and count links to its post in Sources.
Products in this piece
Sources
- Official documentation
- 4
- Public source
- 11
- Looker
- Looker docs: Conversational Analytics in Looker overview
- Looker docs: Looker-managed MCP server
- Looker release notes
- G2 review on modeling metrics once in LookML (Nov 15, 2025)
- G2 review from a BI architect on self-serve and price (Jun 27, 2025)
- Reddit comment on a year to gain buy-in for Looker (Nov 22, 2025)
- Reddit comment on a LookML bottleneck and a move to Domo (Nov 30, 2025)
- Reddit post on reusing metrics across Explores at scale (Apr 17, 2025)
- Reddit post on rolling Looker out to finance teams that work in Excel (Oct 7, 2026)
- Reddit thread on moving from Power BI to Looker (Feb 17, 2026)
- Reddit comment on moving from Looker to Sigma (Apr 16, 2026)
- Hacker News comment on Looker as a semantic layer with no roadmap (Mar 13, 2025)
- G2 review on company-wide LookML metrics and natural-language queries (Jan 15, 2026)
- Reddit post on the pre-GA Looker-managed MCP server and Claude's web connector (Jun 17, 2026)