Setup Business Context

Add Your Business Context to the Agent

This guide assumes you've already connected your data warehouse and GitHub code context to Slateo.

1. In Settings -> General, fill in:
   • Organization Context — a short summary of your company, business model, customers, and domain.
   • Custom Instructions — naming conventions, KPI rules, fiscal calendar, output preferences, and durable analyst guidance.
2. Set up your semantic layer:
   • On the left-hand sidebar, head to Data Model -> Datasets. Open any dataset and edit its description, document important columns, and define dimensions, measures, and joins. You can use the inline AI button to auto-populate column descriptions, then refine them.
   • In a new agent chat, use the Code Indexer preset agent to smartly investigate and populate your table metrics, dimensions, measures, and documentation. Just use a prompt to tell the agent where to start!

Using Metrics as Context Guardrails

Metrics are reusable definitions for the numbers your team asks about repeatedly. Use them for company KPIs, product health, funnel rates, revenue, retention, and other measures that should have one shared definition to prevent Slateo agents from coming up with creative definitions.

After you create a metric, users can reference it in agent chats, reports, projects, and the Data Model. Agents can also find existing metrics before writing new SQL.

When to create a metric

Create a metric whenever any of these apply:

• The same calculation is used in more than one report or analysis
• The definition needs to be documented and reused by the team
• The metric can define the value's valid dimensions such as country, plan, channel, platform, or product area
• The metric needs a target, benchmark, or forecast line

Option 1: Create a metric with an agent

Use an agent when the source definition already exists in another system, or when you want the agent to inspect existing SQL before creating the metric.

1. Start a new chat and ask the agent to create the metric — give it as little or as much information as you can, such as the exact SQL, source table, metric name, dimensions to include, etc.
2. Review the created metric's source query, aggregation, dimensions, and target configuration it proposes. Use follow-up chats with the agent to make any necessary changes.
3. Open the metric from the agent result (or in Data Model -> Metrics) and edit it from T3 to T1 to prioritize it in the agent's recall.

Example prompts:

Create a reusable metric for weekly active users. Use our existing query history to find the current definition. Keep it generic and add platform and country as dimensions.

Create a Revenue metric from the Looker order revenue definition. Use region and sales channel as dimensions. Add the monthly budget target if you can find it.

Use Amplitude to inspect the activation events, then create an Activation Rate metric from the warehouse event table. Use plan and acquisition channel as dimensions.

Option 2: Create a metric manually

Use this path when you already know the source query. Use the Editor in the menu for direct SQL control, reproducible flow-based analysis, and agentic query debugging. Running SQL in Editor automatically creates and saves a flow which can be used to create a metric.

1. Navigate to Data Model -> Metrics
2. Click Add Metric
3. Select the query (node) that should power the metric
4. Confirm the entity, time, and value columns
5. Choose the aggregation, such as sum, avg, count, or count_distinct
6. Add dimensions that users should be able to slice by
7. Choose the display format, such as number, compact number, currency, or percent
8. Add target data if the metric should show a goal or benchmark
9. Run the preview
10. Save the metric after the preview succeeds

The saved metric appears in Data Model -> Metrics. All team members can reuse it, inspect the preview, and reference it in chats and reports.

Ask your first question

Here are a few good starting prompts:

Document your semantic layer

Best run with the Dynamic Agent.

Update the dataset documentation for the orders table — write a
plain-language description, add column descriptions for undocumented
columns, define key measures (revenue, order count, AOV) and
dimensions (region, channel, product category), and add joins
to the customers and products tables.

Create and manage metrics

Metrics in Slateo are reusable, versioned, and show up as context for future prompts. Be specific about granularity and dimensions.

Create a new daily metric called 'Customer Acquisition Cost'
calculated as total ad spend / new customers, with dimensions
for channel, region, and campaign. Add a monthly target of $45.

Create a new weekly percentage metric called 'Gross Margin %'
calculated as (revenue - COGS) / revenue, with dimensions for
brand, product category, and region.

Analyze business performance

These deeper prompts work best when you tag the relevant datasets or metrics with @.

@sessions @conversions Build a funnel from sessions to sign-ups
to first purchase for the last 30 days, broken down by channel.
Highlight where the biggest drop-offs are.

@acquisition @spend Using the latest data, give me an acquisition
health check for this quarter:
- Trend CAC, new customers, and 90-day contribution by channel
  and region.
- Compare performance vs. our quarterly targets.
- Highlight which channels are driving higher-than-target CAC.
- For each underperforming area, suggest the top 3 hypotheses
  based on observable data (e.g. CVR, CPC, AOV) and what tests
  we should run next.

Decompose the gap between our current CAC and our target CAC:
- Break down variance by channel, campaign family, and region.
- Attribute the variance into traffic volume, conversion rate,
  and average order value effects.
- Show which spend reallocations would most reduce CAC while
  keeping new customer volume roughly constant.
- Summarize as a recommendation for budget reallocation.

@orders @customers We know total margin is in line with budget
but full-year sales are tracking behind target, driven by repeat
sales in the US. Decompose this miss:
- Separate volume, price/mix, and margin effects by region and
  product category.
- Identify which repeat customer cohorts or segments account for
  most of the shortfall.
- Summarize in a narrative I can paste into a quarterly review.

Build dashboards and recurring reports

Slateo builds reproducible pipelines, not one-time snapshots — so these dashboards can be scheduled to refresh automatically.

Build a custom visualization for a weekly pipeline review dashboard for open deals:
- Score deal momentum for each opportunity.
- Show pipeline velocity and stage conversion rates.
- Flag critical or at-risk deals by rep.
- Include week-over-week trend charts.

Build a proposed weekly operating review template:
- Section 1: Acquisition — CAC, new customers, and contribution
  by channel and region.
- Section 2: Retention — YoY retention rate, churn, and order
  frequency.
- Section 3: Experiments — key live tests and latest results.
For the most recent full week of data, populate this template
and call out the top 5 insights and anomalies.

Turn answers into shared memory

After you get a useful answer, store the valuable parts back into the product so the whole team benefits:

• Data Model -> Metrics — codify KPI logic and important definitions.
• Data Model -> Datasets — improve table, column, and relationship docs.
• Reports — save the finished narrative or output.
• Projects — group related chats, reports, and context for the team.
• Data Model -> Uploads — upload CSV or spreadsheet files and promote them into managed tables.

Slateo auto-saves your query work so you can focus on analysis instead of manual processes.

Auto-saved queries appear in your workspace and can be:

• Re-run with updated data
• Shared with team members
• Referenced in reports
• Scheduled for regular execution