Atlas: the AI marketing analyst grounded on your warehouse data
Ask any attribution question in plain English. Atlas writes the SQL, builds the chart, and cites every join it used. Answers come from your actual attribution tables, not from generic AI training data.
What Atlas is
An AI analyst that works with your data, not generic training data
Most AI assistants in marketing tools are wrappers around a large language model with access to pre-aggregated dashboard data. They can answer questions about what is already visible in the UI. They cannot join to new datasets, recompute attribution on a custom cohort, or explain the SQL behind an anomaly.
Atlas is different because it operates directly on your warehouse tables. When you ask "which content piece appears most in the middle of won deals on U-shaped attribution?", Atlas writes a query against your attribution fact table, executes it in your warehouse, and returns the results with the query shown. You can copy the SQL, modify it, and run it yourself.
Atlas is built on AttriByte's warehouse-native architecture. Your raw event data stays in your warehouse. Atlas receives only the aggregated results it needs to answer each question. No PII or raw events are transmitted to the AI model.
Atlas is included in every AttriByte plan. It is not a separate AI add-on or a premium tier feature. The Growth plan at $1,200/month includes Atlas with full attribution query access across all six models.
What Atlas can do
Six things Atlas does for your attribution workflow
Attribution Q&A in plain English
Ask "which channel drove the most pipeline last quarter?" or "what is the first-touch contribution of paid LinkedIn?" Atlas writes the SQL, runs it against your warehouse, and returns a chart with the numbers and the query that produced them.
Proactive insight surfacing
Atlas monitors your attribution data for anomalies and trends: a channel whose CPA dropped 30% week-over-week, a campaign that is winning pipeline on last-touch but contributing nothing on first-touch. It surfaces these without waiting for you to ask.
Budget reallocation recommendations
Based on current attribution data, Atlas recommends specific budget moves: which campaigns to increase, which to cut, and what the projected pipeline impact is. Every recommendation cites the underlying attribution numbers.
Segment suggestions grounded in your data
Atlas identifies high-intent account segments from your attribution and pipeline data and proposes them for audience activation: accounts that consumed five or more touchpoints in a 30-day window, or accounts where mid-funnel engagement spiked.
Grounded on your warehouse data
Atlas does not answer from general AI training. It answers from your actual attribution tables and pipeline data. Every response cites the join, filter, and aggregation it used. You can verify any number directly in your warehouse.
Raw data stays private
Your raw event rows and PII are never sent to any AI model. Atlas operates on aggregated query results only. This is enforced at the architecture level, not a policy statement.
What to ask Atlas
Any attribution question your team has asked in a spreadsheet
Atlas handles the full range of attribution analysis that marketing and revenue ops teams currently run in spreadsheets, BI tools, or ad-hoc SQL queries. Type the question in the same language you would use in a Slack message to your data analyst.
Atlas surfaces the answer as a chart, a table, or a plain-language explanation depending on what fits the question. Every response links to the underlying SQL so you can verify, extend, or save the query for a recurring report.
Example questions
- Which channels contributed to the deals we closed in Q3, on a W-shaped model?
- Compare first-touch vs last-touch credit for our Google Ads campaigns this month.
- What is the average number of touchpoints before a demo request from enterprise accounts?
- Which campaigns have a pipeline ROI above 3x on linear attribution?
- Show me accounts with 8+ touchpoints in the past 30 days that are not yet in an active opportunity.
- What percentage of our last-touch conversions came from branded search?
- Which content piece appears most often in the middle of won deals on U-shaped attribution?
Privacy and safety
How Atlas handles your data
AI and data privacy are not in conflict if the architecture is right. Here is exactly what Atlas does and does not do with your data.
What Atlas receives
Aggregated query results from your warehouse: numbers, labels, and chart data. Not raw event rows. Not PII.
What Atlas does NOT receive
Raw event records, personal identifiers, form field values, or any un-aggregated row-level data from your warehouse.
Is my data used to train the AI?
No. Your query results are never used to train or fine-tune any AI model.
What permissions does Atlas have?
Atlas has read-only access to the attribution tables AttriByte manages in your warehouse. It cannot modify data or write to any table.
Can Atlas push audiences or activate segments?
Atlas can propose audience segments. Activation (pushing a segment to Meta, Google, LinkedIn, or any other destination) requires explicit user confirmation. Atlas cannot activate autonomously.
FAQ
Atlas AI analyst: common questions
What is Atlas?
Atlas is AttriByte's built-in AI marketing analyst. It answers attribution questions in plain English by writing SQL against your warehouse attribution tables, building the chart, and showing the query it used. It also surfaces proactive insights and recommends budget moves based on your current data.
Does Atlas use my raw customer data to train the AI model?
No. Your raw event data and PII stay in your warehouse. Atlas receives only the aggregated results of queries against your attribution tables. Your data is never used to train any AI model.
How does Atlas differ from a standard AI chatbot on top of a dashboard?
Most AI chat features on top of dashboards operate on pre-aggregated summary data or a flat dataset. Atlas is grounded on the full attribution tables in your warehouse: it can slice by any dimension, join to any table you grant access to, and recompute attribution dynamically. The answers are not limited to pre-built chart types.
Can Atlas take actions, or is it read-only?
By default, Atlas is read-only. It can propose audience segments for activation, but write actions (pushing a segment to Meta or HubSpot) require explicit approval from a user with activation permissions. Atlas will never modify your data or push to a destination without a confirmed action.
Which attribution models can Atlas query?
Atlas can query all six of AttriByte's attribution models: first-touch, last-touch, linear, time-decay, U-shaped, and W-shaped. You can ask Atlas to compare models side-by-side in a single question.
Ask Atlas your first attribution question.
Atlas is included on every AttriByte plan. Connect your warehouse and start getting answers in minutes.