Margin by product line, with orders in Shopify and costs sitting in a spreadsheet. It joins the two and shows every step.
Operator
Hand it the whole job.
Operator is the AI built into Switchboard. It runs the analysis, builds the dashboard, and changes how your numbers are defined when the business changes. Teach it a process once and it runs it again, on demand or on a schedule.
Operator
Churn jumped in July. Find out what's driving it, and set us up to catch it early next time.
Ran a query · 3 sources
Compared cohorts
It's the June cohort. The first-order promo pulled in buyers who don't come back: their second-order rate is 31% against 58% for full-price customers.
Second Orders by Cohort Saved to Retention
April 58%
May 55%
June 31%
Good. Watch it weekly.
Scheduled · Mondays 8am
Done. Every Monday I'll check each cohort's second-order rate and flag any that fall below 45%.
It knows what your fields mean.
Every field it reads carries a plain-English note on what it means and how it's calculated, so a column is never just a column.
It knows how your business runs.
What you sell, how you make money, which promo ran in March, and how your industry reads its numbers.
It knows what you've taught it.
Corrections stick and routines carry forward, so the hundredth answer is sharper than the first. What it learns stays yours, never shared across customers.
The work you can hand it.
Ad spend that went quiet on Tuesday. It checks the sync, finds where the data stopped, and gets it flowing again.
Add contribution margin, remap a channel, pull in a new field. Operator proposes the exact change and waits for your yes.
A weekly scorecard with revenue, ad spend, and fulfillment status. It lands in the sidebar, shared with the team, reading live numbers.
Which discount codes to cut, how much to produce this week, where the budget should move. You get a recommendation and the numbers behind it.
A Monday fulfillment plan: what's in stock, what to make, which orders ship first. This week's numbers, same format.
And when a job should run without you, Operator hands it to an Agent
Every answer comes off the same model.
Operator reads one model of your business, where each number is defined once. Its answer matches the dashboard, and the dashboard matches the books.
01Ask
Ask for the change.
In the same conversation as everything else.
Treat exchanges as refunds in net revenue.
02Propose
See the exact change.
Operator drafts it against the model and shows what it touches.
Model change · orders_output
~ net_revenue: exchanges count as refunds
+ exchange_rate, defined once
2 dashboards read these fields
03Approve
Nothing applies without your yes.
Approve it and every dashboard reads the new definition.
Approve Revise
Applied · 2 dashboards updated
Bring us the question your team keeps re-answering.
Operator's context doesn't write itself. Our team connects your tools and defines your numbers the way you actually use them, and that foundation is what the answers stand on. Bring the report you rebuild by hand every month, and we'll show you what Operator would do with it.
Book a demoFAQ
How do I know an answer is right?
Ask it to show its work. It walks you back through the steps: which fields it read, what they mean, and how fresh the data was. If something looks off, you trace it to the source instead of taking it on faith.
What happens when Operator isn't sure?
It asks. When a question could mean two things, like which margin or which date range, Operator asks before it answers. And bigger moves, like changing how your data is organized, always wait for your approval.
How does Operator know what a field means?
Behind every chart is a definition: what counts as a repeat order, how refunds are treated, where a "day" starts. Switchboard writes those down in plain English, and Operator reads them before it answers. That's why two people asking the same question get the same number.
What happens when it gets something wrong?
Tell it. Operator walks the number back through the query and the fields it read, and you see exactly where it went sideways. The correction sticks, so the same mistake doesn't come back next month.
Is our data used to train AI models?
No. Operator runs on enterprise AI APIs that don't train on your data, and what it learns about your business stays in your workspace, never shared across customers.
Get started
Hand it a hard question.
Book a demo. We'll put Operator through a full analysis end to end, and you can pull on every thread it shows.