What is Nash AI?
Nash AI is Astrato's built-in AI assistant. It works directly inside the platform so you don't need to switch tools or export data to get help. Nash understands your semantic model, your data, and what you're trying to build - powered by Astrato for free (daily limits may apply), or by your own LLM.
Enterprise analytics is different from generic AI use cases. Revenue, ARR, churn and customer metrics must be calculated using approved business logic, trusted data sources and governed definitions. Nash uses Astrato's semantic layer as the foundation for every dashboard, helping teams move faster without sacrificing trust.
Nash is available in the following places:
Each brings a different kind of help, tuned to what you're doing right now.
| Powered by your semantic model Nash knows your business. Every suggestion, every chart, every measure description it produces is grounded in the data model you have already defined. That means it stays accurate and stays in sync. |
Nash in the Workbook Editor
Building a dashboard used to mean starting from a blank canvas and working through every chart one by one. Nash changes that. Open the AI panel in any workbook and tell Nash what you need. It reads your semantic model, asks a couple of focused questions, then builds the whole thing.
What Nash Can Do in the Workbook Editor
Capability | Detail |
Generate dashboards from scratch | Ask Nash to build a dashboard and it handles layout, charts, filters, and KPI cards based on your semantic model |
Guided intent questions | Nash asks up to two focused questions to understand your goal before it builds |
Transparent step-by-step plan | Every action is listed in the panel so you can see exactly what Nash is doing and why |
Update existing dashboards | Ask Nash to change charts, measures, formatting, or layout on any existing workbook |
Add new chart types | Request specific visuals like Sankey diagrams or heatmaps by name and Nash places and configures them |
Cross-visual formatting changes | Ask Nash to update formatting (e.g., currency) and it applies the change across all relevant components at once |
Semantic model awareness | Nash reads your live semantic model and suggests appropriate measures and dimensions for every visual |
Context-aware relayout | Ask Nash to reorganise the dashboard and it rearranges components to accommodate new additions |
Opening Nash
Click the AI icon at the bottom of the left toolbar to open the Nash panel. It slides in alongside your canvas. You can keep it open while you work.
Nash AI panel open in the Workbook Editor, ready to receive a prompt
Creating a Dashboard from Scratch
Type your goal in plain language. Nash uses your semantic model to understand what data is available, then it asks two short questions to make sure it builds the right thing. It recommends a focus area and shows you the options so you stay in control.
Nash analyses the semantic model and recommends a dashboard focus, with selectable options
Once you confirm your answers, Nash builds a full execution plan and begins placing components. You can watch each step run in the side panel. Filters, KPI cards, trend lines, and detail tables appear on the canvas as Nash works through them.
The completed plan shown alongside the finished dashboard, including filters and all visuals: KPI cards, trend chart, category bar, and detail table.
| From blank canvas to full dashboard in under 3 minutes The example above, using the Bike Sales model, produced a sales performance dashboard with 5 filters, 4 KPI cards, 2 charts, and a detail table without a single drag-and-drop. |
Updating an Existing Dashboard
Nash is not only for new builds. On an existing dashboard you can ask it to change chart types, swap measures, update formatting, or add entirely new components. Type what you want changed and Nash handles the rest.
Example prompts that work today:
Please update $ into EUR, we're a European business
Add a Sankey diagram for category flow and a heatmap for gross profit
Relayout the dashboard and populate the new charts with suggested dimensions and measures
Nash updating currency formatting across all visuals in a single action
Nash building a Sankey diagram and heatmap in parallel, with a clear step-by-step plan.
| Nash understands context When you ask Nash to add a Sankey diagram, it does not ask which data to use. It reads the existing model, picks appropriate dimensions, and builds the chart. You can always adjust after, but most of the time you won't need to. |
Nash in the Semantic Layer
The semantic layer is where data models live. It is also where a lot of time gets lost. Understanding a new model, checking what measures exist, spotting gaps, knowing which tables are connected and how. Nash in the semantic layer makes all of that fast.
This is often the fastest way for a new analyst, BI developer or business stakeholder to understand how an organisation defines and measures its business.
Open Nash inside the semantic layer and you have a conversational interface on top of your data model. Ask it anything about the model and get a clear, structured answer.
What Nash Can Do in the Semantic Layer
Capability | Detail |
Model description | Full structured summary of tables, joins, dimensions, measures, and relationships in plain language |
Relationship mapping | Nash reads every join and explains how tables connect, including where the join web is incomplete |
Measure catalogue | Asks Nash to list all measures with their formula, source table, and what the metric means |
Deep measure detail | Ask for full detail on any measure or set of measures and Nash explains the logic, filters, and time intelligence applied |
Quick observations | Nash flags potential issues, such as duplicate tables, non-standard aggregations, or incomplete joins |
Conversational follow-up | Ask follow-up questions in plain language. Nash keeps context across the conversation |
Model health check | Nash assesses overall model completeness and suggests where further work might be needed |
AI suggestions | Measure Suggestions, Dimension Suggestions, Field Alias Suggestions, and Join Suggestions available from the model canvas |
Describe My Semantic Layer
Click "Describe my semantic layer" and Nash reads the entire model. It produces a structured summary covering tables, joins, relationships, dimensions, measures, and observations about the model's health and completeness.
Nash summary showing table groups, key relationships, and full measure catalogue
The description is not just a list. Nash reads the relationships and highlights things worth knowing, such as which tables connect to which, where the join web has gaps, and which measures use time intelligence.
Nash surface quick observations and recommendations, including a suggestion for further queries
| Onboarding made fast A new team member or analyst joining a project can ask Nash to describe the model and have a working understanding of the data structure in under a minute. |
Deep Measure Detail
Ask Nash to explain a specific measure or give you detail on all measures. It breaks down the formula, the source table, the logic, and flags anything worth knowing.
Nash listing measures with formulas, source tables, and key observations about each
Nash also surfaces patterns across the measure set, grouping them by type: revenue metrics, product metrics, order metrics, time intelligence measures, and more. This makes it easy to spot what you have and what might be missing.
Advanced Use Cases
Nash AI is designed to handle more than day-to-day dashboard and model work. The following use cases show how Nash can support larger data modernisation, governance, and migration initiatives across an enterprise. Nash can create complex measures such as YTD, MTD and YoY measures.
Cross-Platform BI Migration and Logic Translation
Moving from one BI platform to another is rarely straightforward. Years of logic built in Qlik Set Analysis, Power BI DAX measures, Tableau calculations, SQL business logic, or Excel-based reporting do not translate automatically. Teams are often left rebuilding from scratch, which is slow and introduces inconsistency.
Nash AI can analyse existing BI logic and translate it into reusable, governed semantic assets inside Astrato. Instead of manually recreating every measure and calculation, teams can describe or paste existing logic and Nash will produce the equivalent semantic definition. Those definitions become part of the governed layer, not buried in a single workbook.
This accelerates migration and modernisation initiatives while preserving the trusted business definitions teams have built over years.
Example prompts:
“Translate this DAX measure into a semantic metric”
“Here is a Qlik Set Analysis expression. What is the equivalent in Astrato?”
“Convert this Excel formula into a governed calculation”
Looker's LookML | Qlik Set Analysis | PowerBI DAX |
measure: ticket_count { | Count({<CUSTOMER=>} ID) | Share of Tickets by Customer = |
Example of Nash creating measures from Power BI's Dash Expressions.
Turn Warehouse Tables Into Governed Business Models
Raw cloud warehouse schemas rarely arrive ready for business use. Tables need to be understood, relationships need to be defined, and measures need to be created before analysts can start answering questions. The dashboard is usually the easy part. Defining trusted metrics, business entities, relationships and reusable calculations is the hard part. Nash starts there.
Nash AI can analyse enterprise warehouse schemas and help generate the semantic foundations needed for business-ready analytics. This includes:
Dimensions - what business entities are represented
Metrics - what the business wants to measure
Relationships - how tables connect to each other
Hierarchies - how data rolls up (region > country > city, for example)
Drill paths - the ability to click into a number and see what is behind it
Aliases - cleaner, business-friendly names for technical field names
Reusable calculations - common expressions defined once and shared across the model
This dramatically reduces the time required to move from raw cloud warehouse structures into business-ready analytics systems.
Continuously Improve Semantic Consistency
As semantic layers grow over time, they accumulate inconsistencies. Duplicated metrics appear with slightly different definitions. KPIs are named differently across teams. Business logic fragments across multiple models. Standards drift.
Nash AI helps identify and fix these issues before they cause problems downstream. It detects overlapping logic, recommends reusable KPI structures, and highlights opportunities for standardisation. Teams can use Nash to run a consistency review across the full semantic layer and surface exactly where definitions conflict or overlap.
This is particularly valuable at scale, where a manual review of hundreds of measures and dimensions would take weeks.
Business Metric Creation in Natural Language
Not everyone who needs to define a metric knows SQL or BI-specific calculation languages. Business users and analysts often have a clear idea of what they want to measure but need a technical resource to translate it into something the platform can use.
Nash AI bridges the gap between business intent and governed analytics. Users can define governed metrics using plain language and Nash will produce the semantic definition. The result is a reusable semantic asset that can be trusted across dashboards, reports and applications.
Example prompts:
“Create a churn rate metric”
“Define active customers as customers purchasing within the last 90 days”
“Create gross margin by product category”
Nash translates business intent into reusable semantic definitions. This means metric creation becomes a conversation, not a development ticket.
Explain and Discover Enterprise Data
Enterprise data environments are large, complex, and difficult to navigate even for experienced analysts. New team members spend days trying to understand what data is available, how it is structured, and which metrics are trusted. Even experienced analysts waste time navigating schemas they rarely use.
Nash AI makes enterprise data navigable for everyone. Ask Nash about the data model in plain language and get clear, structured answers about:
Data models and how they are structured
Relationships between tables and entities
Trusted metrics and what they measure
Semantic definitions and how fields are defined
Available business entities and what each one represents
Example prompts:
“What tables contain customer lifecycle data?”
“Which metrics are finance-approved?”
“What is the difference between bookings and revenue?”
This reduces onboarding friction, speeds up self-service analytics, and makes enterprise data more accessible across the organisation.
Prompts That Work Well
Nash works best when you are direct about what you want. You do not need special syntax. Write it the way you would explain it to a colleague.
Workbook Editor Prompts
Build a dashboard based on the current semantic model, with the most important KPIs and filters at the top
Add a Sankey diagram and a heatmap, relayout the dashboard, and populate with suggested dimensions and measures
Please update $ into EUR, we're a European business
Show me sales performance by region over time
Create an executive summary sheet with a revenue trend and top 10 customers
Semantic Layer Prompts
Describe my semantic layer
Give me an overview of all measures with all necessary detail
Which tables are connected to Orders and how?
What time intelligence measures do I have?
Are there any gaps in my join web?
Explain the Revenue MTD measure
| Use follow-up questions Nash keeps context within a session. After it describes the model, ask "what measures are missing for a sales pipeline view?" and Nash will answer in context. You do not need to re-explain the model. |
Setting up & configuring Nash
Nash uses one of the AI providers configured within Astrato. Organisations can use Astrato's managed provider or connect their own approved LLM provider.
Nash can be enabled disabled on this page, and you can choose the default AI Provider to use with Nash here too.
Using Nash with Snowflake Cortex API
The API version of Snowflake Cortex is available for use with Nash AI.
What you need before starting:
PAT Token for your Snowflake user
Account name:
https://<account-identifier>.snowflakecomputing.com/api/v1/openai
To configure Snowflake Cortex API:
Select OpenAI
Select the custom model name option in Astrato. You will type in the model name you wish to use from Snowflake Cortex. Non-OpenAI models are compatible such as
claude-3-5-sonnet.Clearly name the connection.
We suggest:
Snowflake Cortex API {ModelName} {Account Name}
Known Limitations
General
Nash does not yet access underlying data values when:
Building content in the Workbook Editor
Managing or updating the Semantic Layer
As a result, filters and filter values may be suggested based on metadata and context rather than actual data. These should always be reviewed by a user.
ℹ️ Nash currently supports only a subset of LLM providers that perform well with tool-calling workflows.
To maintain responsiveness, Nash compresses conversation and tool context. This means it may lose awareness of older messages, actions, or generated content during longer sessions.
💾 Chat history is stored locally in the browser for privacy reasons. Chats are not shared across browsers, devices, or computers.
Workbook Editor
Nash only works on the currently open sheet.
If you leave the sheet, Nash stops building. You can resume when you return.
Nash cannot create new measures yet, this currently only works in the Semantic Layer.
Some chart properties and configuration options not yet supported will be added successively.
Nash may suggest filters and filter values that should be reviewed before use.
Maps are not currently supported.
Custom Report is currently not yet supported.
Semantic Layer
Nash does not query data values when working in the Semantic Layer.
Nash cannot validate suggested filter values against actual data.
Filters and filter values suggested by Nash should be reviewed before publishing changes.
Nash does not yet allow you to delete fields or remove tables.
Workbook Actions
Nash AI for Workbook actions is coming next.