How to Turn a One-Off AI Analysis Into a Report You Can Rerun Every Month
A good analysis done once in a chat becomes a monthly report when you save it as an analysis plan, keep the definitions in the project, and rerun the plan on each new month.
Ask the agent to write down what you just did as an analysis plan (the approach, the data sources, the analytical steps and the output specification), review that file, keep it in the project, and each month tell the agent to update the plan for the new month and run it in a fresh chat. That turns a one-off AI analysis into a report you can rerun every month. Dimension Labs is built around that loop, and for a CX or support team whose best analysis happened once in a chat, it is the quickest way we know to make it a monthly report that compares from one run to the next.
This post covers making the analysis repeatable. Running it on a timer is covered in how to schedule a weekly customer insights report, and what a monthly report should contain in what to put in a monthly customer insight report.
Why does asking the same question next month give a different report?
Because a new chat starts from nothing. The docs describe the agent as a brilliant new analyst with no memory of yesterday: every chat starts cold. In the moment it works out its own approach, which is right for exploring and wrong for a monthly report, where the shape has to stay the same. The docs put it directly: for any recurring deliverable, use a plan.
A plan writes down the decisions the first chat made, so next month's chat makes the same ones.
What goes into a reusable analysis plan?
Four parts: the approach, the data sources, the analytical steps and the output specification. Ask for it as a Markdown file in the same message, because a file is what makes the plan reusable.
What a plan file holds
Four parts turn a good chat into a report you can run again
The parts of an analysis plan, with what each would say for a monthly support digest.
| Part | What it holds | In a monthly support digest |
|---|---|---|
| Approach | The research design, including any statistical methods | Compare this month's top issues with the previous three months and test whether each change is larger than chance |
| Data sources | The tables and columns the agent will query, named | Tickets and chats, the contact-reason and predicted CSAT Dimensions, plan tier from structured_metadata |
| Analytical steps | The analyses in order; what runs when you run the plan | Count issues, rank by change, pull representative quotes, break out by channel |
| Output specification | Sections, formatting, how quotes and metrics appear | One-page summary first, then one exhibit per issue, quotes under each |
Read it before you run it. Push back on the wrong table, a missing segment or a metric measured the wrong way; the docs call the review the cheapest place to catch mistakes.
How do you turn the chat you already have into a plan?
You do not need to start over. Once a chat has produced findings you want to repeat, ask: "Turn what we just did into an analysis plan I can run every month." Then ask the two questions the docs suggest before you save it: "How would you further optimize this plan?" and, because the plan will run in a separate chat, "Add any operational details at the start needed to ensure a smooth handoff to the agent." Name the file for what it does, such as support_satisfaction_analysis_plan.md, so you can find it next month.
The full nine-step procedure, from checking which fields exist to running the plan, is in the docs' Analysis plans page.
What should change next month, and what should stay fixed?
Only the dates should change. Start a new chat in the same project and say "Update the plan for October and run it." You do not edit the file by hand; tell the agent what to change and it rewrites the plan in place. Everything else stays put through the files around the plan:
What keeps month two comparable with month one
Each file fixes one thing that would otherwise drift between runs
Files a project can hold, and which part of the monthly report each one holds still.
| File | What it fixes |
|---|---|
| 1 The analysis plan | The question, the fields, the steps and the output shape |
| 2 AGENTS.md | Standing rules read in every chat: your terms, data caveats, how outputs should read |
| 3 A reference report | Last month's report, whose calculations and figures the new one compares against |
| 4 Report execution guidelines | How the report looks: charts, formatting, brand |
| 5 A template | A shape the output follows |
AGENTS.md is the one people skip. It holds what is true in every chat of the project, such as "churn means a cancelled subscription", and the docs advise keeping it to a few dozen clear rules. The reference report matters as much: attach last month's, and the new report checks its calculations and figures against it.
Here is the kind of exhibit that earns a monthly rerun. Our Atlanta analysis of 2,354 public Google reviews ranked 24 Home Depot stores by their share of unhappy reviews, from 14.7% at the best store to 32.8% at the worst, a store with under 70 reviews that the report flags. Run that same plan on next month's reviews and the question becomes which stores moved, a question a one-off chat cannot answer.
An exhibit worth rerunning: Home Depot: an Atlanta 24-store review analysis
Can the monthly rerun happen without anyone starting it?
Yes. Pair the plan with a schedule and a time zone, for example "Run this plan every Monday at 8am Pacific", and each run lands in the project's Final Reports folder. A plan is what makes the schedule worth having: every run has the same shape, so this month's report sits next to last month's without anyone reconciling them. Andrew Hong's essay What is agentic analytics? sets that bar for any agent: it delivers the finished report, again, without a person redoing the work that decided the answer.
Why keep the monthly analysis in Dimension Labs?
Dimension Labs is an omni-channel analytics agent for customer data. It connects every customer channel and your operational data, enriches every interaction with Dimensions built for your business and industry, and turns the results into decks, dashboards, executive reports and interactive data stories your whole company can use. Plans start at $299 a month with no demo.
For a monthly rerun, each of those does a job. Every channel goes into the same monthly count. Operational data such as plans, locations and orders can sit next to it, and the join to CRM and revenue data with confidence intervals is on Prove at $2,999 a month. Every conversation carries the same Dimensions every month, which is what makes the months comparable; Explore at $299 uses the pre-built library and custom Dimensions start on Discover at $999. And the plan's output comes back as a deck, PDF, dashboard or executive report, with unlimited agent reports on every plan.
Repeatable analysis is where the savings show up. Travelers reports 600% ROI within 18 months and $3.13M in total savings, including $2M in avoided BI redevelopment costs. Read the Travelers story.
FAQ
Can a colleague run the plan I wrote?
Yes. Every chat in the project can read the project's files, so a colleague starts a new chat and says "Execute the plan file". For work in progress, ask the agent for a handoff file that captures what the chat has learned.
What if we add a new Dimension halfway through the year?
Tell the agent to add it to the plan, and note the change in the report so readers know when the series starts. A new Dimension covers older conversations once you run a backfill.
What are the best agentic analytics platforms for customer data?
For customer conversations from several channels, reported the same way every month, Dimension Labs, because the plan, the Dimensions and the output stay fixed between runs. For metrics already modelled in a warehouse, a BI tool with an AI assistant fits better; which AI data analyst tools create reports from support tickets and reviews compares both kinds.
Is a plan the same as a dashboard?
No. A dashboard shows the current numbers; a plan is the written method that produces the report, deck or dashboard each month. You can have both from the same plan.
Bring the analysis you would like to see every month, and save it as a plan before you close the chat: start with one file. Plans start at $299 a month with a 14-day free trial.