01
KPIs and metrics are untrustworthy
Revenue should mean one thing. But finance, marketing, and sales can still return different figures.
Looker conversational analytics: ask in plain English, get answers you can trust
SquareShift builds Looker conversational analytics. Every answer resolves to your KPI definitions and stays inside your access controls, so the number in a chat matches the number on your dashboard.
Try A Demo
30 minutes, on your own data model, not a slide deck.
- Looker
- BigQuery
- Vertex AI
- Google Cloud
Executives still get inconsistent, scattered and backward-looking analytics
02
Insights are scattered across multiple reports
The evidence sits in reports, documents, and systems. You have to find it, compare it, and collate it yourself.
03
Analytics explains the past but not how to respond
You can explain what happened. You still cannot see the action that will prevent the next miss.
Decision intelligence closes the gap between the number and the next move.
How decision intelligence connects answers to action
The user asks one business question. The system finds the right data, applies the right definitions, and returns an answer with sources, calculations, and next actions.
Metrics are consistent
Evidence is pulled together
The answer leads to action
Business information
- 01PPT
- 02XLS
- 03Doc
- 04Data warehouse
- 05Tables
- 06Reports
- 07Dashboards
- 08Proposals
Query
Revenue is down this month. What caused it, and where can we regain momentum?
Semantic Intelligence
The meaning layer
- AMetric definitionsOne definition per metric
- BSource mappingEvery number traced to a file
- CBusiness rulesSegments, periods, hierarchies
- DEvidence checksNo claim without a citation
Agents & Orchestration
- 01PlannerDecompose
- 02RetrieverFind sources
- 03CalculatorDriver math
- 04WriterCompose
Answer and Report
Answer
Revenue fell 8.7% MoM
Price −6.1% · Volume −2.6% · Mix +0.8%
Evidence
Sales_2024.xlsxTable
Pricing_Report.pptxSlide 7
Finance_Monthly.pdfPage 3
Calculation
Driver
Impact ($M)
Impact (%)
Price
−12.4
−6.1%
Volume
−5.2
−2.6%
Mix
+1.7
+0.8%
Total
−15.9
−7.9%
Next action
Conversational analytics
Ask a question. Get trustworthy answers.
Ask in plain English, not SQL. Get back an answer resolved against your semantic layer, so the response is ready to act on, not just ready to read.
-
Reasoning shown in the open
Every answer shows the exact view it queried and the measure it used.
-
A complete answer
A short summary first, then the numbers as a table, then the same numbers as a chart.
-
Every number traced
The governed view sits behind every answer, so any figure traces straight back to its source.
Governed dashboards
Build the dashboard you need. Just chat.
No second tool and no query builder. An empty slot invites a question, and the answer lands as a tile on the board.
-
Ask once, add a tile
A KPI, a chart, a comparison, or a table, added exactly where you asked for it.
-
Governed like the chat
Tiles resolve through the same semantic layer, so the board and the chat always agree.
-
Store the question, not the numbers
Each tile keeps its question and re-runs it live, so the board stays current, never a snapshot.
Decision records
Get professional reports instantly or schedule them.
When a number has to be defended rather than glanced at, open the workbook underneath it. The same governed figures, as a formal document.
-
One step from the answer
Every governed answer carries a doorway straight into its workbook.
-
Sections, not a transcript
Headed sections with their own charts and tables, built to be quoted from, not scrolled through.
-
Same definitions throughout
Every figure is computed by the same definitions the assistant already queries in chat.
Chat and dashboards together
Ask a new question in plain English. Save its answer as a dashboard tile when others need to track it. Both use the same metric definitions.
Live agents
Pick a domain. Try a live demo.
- 01
Retail & E-commerce
“How have chain sales tracked month by month this year?”
Open demo → - 02
B2B Sales
“Walk me through our FY26 bookings against quota - where did we win and where did we lose?”
Open demo → - 03
Procurement
“Walk me through FY26 spend - where did we save and where did we leak?”
Open demo → - 04
Diagnostic Labs
“How many adults have poorly controlled diabetes, using an HbA1c threshold of 8 percent?”
Open demo → - 05
REITs & Hospitality
“Walk me through the first half of 2026 - what hit us, and how did we recover?”
Open demo → - 06
Manufacturing
“Which lines drove scrap and downtime this quarter?”
In build - 07
Banking & Lending
“Where is portfolio risk concentrating against last year?”
In build - 08
Logistics
“Which lanes missed service levels, and what did it cost?”
In build
Credentials
Partner proof behind the work.
SquareShift brings governed BI, Google Cloud AI, Looker delivery, and Elastic GenAI credentials into the same applied-AI delivery system. See our Looker consulting and how a LookML semantic layer defines metrics. See Looker delivery in our case studies.
-
Google Cloud Premier Partner
-
Data Analytics Specialization
-
Machine Learning Specialization
-
Looker Consulting Partner
-
Google Cloud Generative AI
-
Elastic GenAI credential
- Elastic Select Partner Search, observability, and Elastic delivery credential.
- Anthropic-ready advisory Applied AI and Claude implementation guidance.
Common questions
What is conversational analytics?
Ask a question in plain English, not SQL. You get an answer resolved against your semantic layer, with the numbers, the calculation and the sources.
What is conversational BI?
Conversational BI is business intelligence you use by asking questions in chat. At SquareShift it runs on Looker, so answers and dashboards use the same governed definitions.
What is decision intelligence?
Decision intelligence closes the gap between the number and the next move. Each answer comes with its evidence and a next action.
Does it work with our data and metric definitions?
Yes. Answers resolve to your KPI definitions through your semantic layer, and the assistant queries your warehouse where it already lives.
Do we have to move our data?
No. The assistant queries your warehouse where it already lives, and reads your documents from your own store. Nothing is copied into a separate system to make it work, so there is no second copy of the numbers to keep in step with the first.
What stops it from inventing a number?
There's only one path to the data. When a semantic layer is present, the shortcut around it is switched off completely, so there's no way to bypass it. If the layer can't answer something, it says so rather than guessing, and every figure names its source, so any number can be checked, not just trusted.
Does it respect our access controls?
Answers are compiled by your semantic layer, so the permissions that layer already enforces apply to every question asked through it. Governance is not re-implemented inside the assistant, which is the thing that would let the two drift apart.
How long does setup take on Looker?
Onboarding is a content exercise rather than an engineering one: your definitions, your entities, your document corpus and your golden questions are authored as configuration, and the engine itself is unchanged. That is the difference between a deployment and a rebuild.
Start with the data you have
Want trusted answers from fragmented data without a costly consolidation program?
Bring one decision-critical question. SquareShift connects the sources behind it, applies governed definitions, and shows leaders the answer they can act on.
Talk to SquareShift
Just want to talk to your data? Start there.