Meet the team that delivers the why before the meeting.
The Analyst Agents are the FP&Agents for reporting and analysis: a growing team of autonomous agents that answer the what and the why, flag what looks off, and trace every figure to the source.
The problem
It takes too long to get to the root cause.
The question lands in Slack, the answer lives in a model, and by the time it ties out the moment has passed.
Every answer is a fresh export
Each question restarts the pull-pivot-paste cycle, and the team becomes a report factory.
Numbers nobody can defend
An answer without lineage dies the moment someone asks where it came from.
Problems surface too late
Root cause takes hours of digging, and the line that quietly doubled waits until someone notices.
What the Analyst Agents do
Every question answered, every figure traced.
The Analyst Agents are Cube's FP&Agents for reporting and analysis: a growing team of autonomous agents that fetch any slice across any dimension, explain variance and root cause, and flag anomalies before anyone has to ask. Every figure they return traces back to the source transaction.
Analysis
Fetches any slice across any dimension: department, entity, region, product.
Smart Variance
Variance and root cause, ranked by driver and drillable to the transaction.
Support
Product guidance and troubleshooting in the flow of work.
Anomaly Detection
Watches the numbers and flags what looks off, before anyone has to ask.
Every agent here works under Charlie, the super agent: one question in, one decision-ready answer back.
In the flow of work
The why, answered where you ask.
Ask from Slack, the browser, or your AI assistant; the answer comes back with the sources attached.
Why is gross margin down 2 points?
Delegate the drilldown
From the board to the vendor, answered.
Ask on any board, and the agents drill from the summary figure to the transactions behind it.
Do some analysis on our travel expenses in Q2. Drill down to the transactional details and identify the key drivers of cost by vendor.
Q2 FY26 travel expense analysis
| Period | Travel expense |
|---|---|
| Apr-26 | $142,300 |
| May-26 | $141,850 |
| Jun-26 | $144,750 |
| Q2 total | $428,900 |
Spend was flat across the quarter: each month landed within ~$1,500 of the others. A stable travel run-rate, no one-off spikes; the variance to budget came from headcount growth, and every line below opens to its receipts.
Top vendors driving travel cost
| Vendor | Spend | % of total |
|---|---|---|
| United | $91,440 | 21.3% |
| Delta | $70,770 | 16.5% |
| Marriott | $51,470 | 12.0% |
| Uber | $43,320 | 10.1% |
The payoff
Answers without the assembly.
The answer, not another export
Ask in plain English and get the finished answer, instead of a data pull you still have to work through.
It already knows your business
Answers come from your accounts and your definitions, so revenue means what you mean by revenue.
Receipts attached
Every number opens up to the real transactions behind it, so you can defend it in any room.
Root cause
The margin move, decomposed.
Smart Variance splits the change into drivers, and the drilldown below opens the biggest one to its transactions.
Trace to the source
Every number opens.
The delivery-cost step above, opened to the accounts and transactions behind it.
| Jun 03 | AWS | usage · June | NetSuite | $71,300 |
| Jun 15 | Cloudflare | enterprise plan | NetSuite | $18,450 |
| Jun 28 | Fastly | overage · June | NetSuite | $9,900 |
| + 24 more transactions | ||||
Cube MCP Server
Your AI reads it, and writes it back.
The Cube MCP Server gives Claude, ChatGPT, Copilot, and the agents you build read and write access to the same clean, decision-ready data the Analyst Agents work on.
Read: decision-ready answers
Ask any figure, any slice, any why from Claude, ChatGPT, or Copilot, and get the same traced answers and anomaly flags the Analyst Agents surface in Cube.
Write: work that lands in Cube
Save variance commentary, flag items for follow-up, and push report views back into Cube, with permissions and the audit trail intact.
Use cases
Where the Analyst Agents go to work.
Variance analysis
Every line over threshold flagged, timing separated from structural.
Ad hoc answers
Every department has questions. Finance has answers, with sources.
Metric deep-dives
NRR by cohort, margin by segment, burn by month, sliced on demand.
Anomaly detection
The line that quietly doubled gets flagged the day it happens.
See it on your own workflow.
Bring the variance that took your team a day last quarter and watch it decompose in minutes.
Get the why before the meeting.
Book a demo and run your hardest variance question through the Analyst Agents in a live sandbox.
Part of the FP&Agents: one super agent, four specialist teams.
