AI for ecommerce, done right.Know why the numbers moved.
Revenue dipped. Patricia reads Shopify, Meta Ads, Google Ads, Klaviyo and Stripe together, posts the answer in Slack or Microsoft Teams before standup, and drafts the fix.
An illustrated Larkspur Slack workspace plays example conversations between the team and Patricia, one channel at a time. Pick a channel to read it at your own pace.
Tuesday's numbers: $18.4K, up 12% on last Tuesday. 412 orders, AOV $44.66, MER 3.1 (target 3.0).
- Best seller: the waffle towel bundle, 96 orders
- Watch: Meta prospecting spent $640 at 1.4 ROAS, below your 1.85 break-even. More in #marketing
- Stock: Sage pillowcases run out in 5 days and lead Friday's email. More in #inventory
Works with your store tools, and 1,000+ more
Set it up once. Then it runs every morning.
We set it up with you on a call. Then it runs in your channel.
Connect your tools
Shopify, your ad accounts, Klaviyo, Stripe and your cost sheet. Two stores? Connect both.
Ask once
Brief her in the channel like a new hire. She sets the schedule and keeps it running. Thibaut, CEO of BetterPic, our own online brand, set up the weekly ads report like this:
Every Monday, the last 28 days of Google Ads on real purchases only, by landing page and campaign, with what I should do next.
Approve, then it runs
Anything that spends money or reaches a customer comes back as a draft first. By default, nothing changes without a yes.
I can turn the three page-fix specs into Linear issues for approval.
Give her the job. The answer lands in the channel.
Six jobs Patricia runs for BetterPic, our own online brand. Real threads from our Slack, shortened.
Four delivery flags are open.
| Account | 12-18 Oct | 5-11 Oct |
|---|---|---|
| Spend | $429.55 | $614.12 |
| Purchases | 22 | 31 |
| CPA | $19.52 | $19.81 |
| ROAS | 2.17× | 2.46× |
Spend: $35.02 yesterday, 44% of the normal $78.86 a day. Two stalled campaigns lost 90% of their search views to ad rank, not budget.
Next: check auction eligibility, not bigger budgets. Real purchases only. No account was changed.
The real cause of a spend drop.
- Counts only purchases that are real revenue
- Says when there is too little data to judge
- By default, a budget change stays a proposal until you approve it
Revenue grew for the first time in five months. Satisfaction fell. One cause explains most of it.
Look at this first: delivery slowed on 13 July and never recovered. One order in six now takes over two hours, up from one in 25. About 500 paying customers never downloaded their photos.
Why the month moved, in one line.
- Revenue from Stripe, traffic from GA4, delivery from your data
- One root cause, with the number that proves it
- For a store: orders, refunds and shipping
| USD | Jun | Jul | Aug |
|---|---|---|---|
| Revenue | 37,308 | 40,199 | 38,834 |
| Gross margin | 100.0% | 88.6% | 89.7% |
| Marketing | 8,621 | 7,788 | 8,403 |
| EBITDA | 21,637 | 20,279 | 15,534 |
| Margin | 58.0% | 50.4% | 40.0% |
Margin fell 18 points in two months, for three reasons: a new compute supplier in July, a one-off $3,060 developer payment, and revenue $1,365 below July.
Where the model disagrees: it says August EBITDA is $27,444, on revenue frozen since April. Bank cash says $15,534.
Margin, on cash you can check.
- Revenue after fees, product costs and the costs you name
- The drop explained, the one-off sized
- Turned into a six-slide deck on request
Thibaut, three page fixes are ready for a developer. Search Console data runs to 17 October.
| Page | Impressions | CTR | Position |
|---|---|---|---|
| Funding announcement | 4,389 | 0.09% | 6.48 |
| ChatGPT headshot guide | 1,482 | 0.54% | 7.25 |
| Free LinkedIn photo guide | 1,427 | 0.21% | 9.51 |
Each fix has the exact title, description, intro, headings and schema. Nothing on the site was changed.
I can turn the three fixes into Linear issues for approval.
Pages that earn the click.
- Pages people see in Google but rarely click
- Fixes a developer can ship as written
- For a store: product and category pages
BetterPic is missing from 17 best-of articles in US Google results.
The report lists 41 links, including low placements and name-only mentions.
5 Gmail drafts are ready for the authors. No author email was visible, so the To field is blank. Nothing was sent.
Get on the lists shoppers read.
- Best-of articles for your category, checked weekly
- Missing spots found, emails drafted in your inbox
- Each email stays a draft until you send it
Five headshot brands tracked in ChatGPT and Google AI answers, US English. BetterPic is top 3 on both.
What AI cites most: best-of articles, then Reddit, then brand pages.
Where shoppers ask AI first.
- ChatGPT and Google AI answers for your category, against your rivals
- The sources AI cites, so you know where to show up
- Every month on the 1st
Your store's week, run from Slack or Teams.
A brief every morning. Checks every week. An alert when something breaks. A report every month. You set each one up by asking once.
Yesterday's sales, orders and MER, and why they moved.08:00 · every dayLast week against plan, with three next steps.09:00 · MondaysProducts with under 14 days of stock. Reorder drafted.10:00 · Tuesdays · draftA campaign at 80% of its daily budget by noon.Checked hourly · posts only on a problem
This week's tickets by reason: shipping, sizing, codes.09:00 · WednesdaysPages people see in search but rarely click, with the fix.08:00 · WednesdaysFlow revenue week on week, and any flow that stopped.09:00 · ThursdaysA live email offers a code that has expired in Shopify.Checked hourly · posts only on a problemRefunds by product and reason, with the fix.09:00 · FridaysPromo products against stock, ads against their pages.15:00 · FridaysMonthly report and P&L · 1stMargin by product and channel · 1st
Where AI answers send shoppers · 1stAd account check · 2ndBlack Friday, Cyber Monday and every launch.
Your busiest days leave no time to pull numbers. Set the checks up the week before. She runs them on the day.
The night-before check
Every promo product against stock, every code against its end date, every ad against its page.
Pulse checks through the day
Sales, orders and spend against plan, posted at the hours you pick. The whole team sees the same numbers.
Spend above your floor
Budget moves proposed against your break-even. By default each one waits for a yes, even at 2 a.m.
The recap on Tuesday
What sold, what ran out, and which ads and emails carried the weekend. Saved for next November.
Each app's AI sees its own app. Patricia sees the whole store.
Shopify, Klaviyo and your helpdesk each have their own AI. Each one only knows its own tool. Patricia connects them.
| Compared on | Patricia | A dashboard | One app's AI | An automation tool |
|---|---|---|---|---|
| Sees | Store, ads, email, payments and support | What it imports, as charts | Its own app | The steps you built |
| Gives you | The why, and a drafted fix | A chart | An answer in that app | What you designed |
| Runs on its own | Yes, on your schedule | Scheduled exports | Mostly when you ask | Yes, once you build it |
| Knows your definitions | Saved once, used everywhere | Per dashboard | Usually per question | Hard-coded per flow |
| Before a change | Asks for a yes, by default | Read only | Depends on the app | Only if you build it |
| Where it works | Slack, or Teams in beta | Its own tab | That app | In the background |
Your stores, your rules, your yes.
Every store by name
Connect the US and EU store, two brands, two ad accounts. Each gets a label. She asks which one when it is unclear.
A yes before money moves
Budget changes and messages to customers wait as drafts, with the reason. By default a named person approves each one. The log keeps the record.
Access per person
The team shares one Patricia. A tool is shared with the team or kept personal, and each person reaches only what they may use.
Your definitions, remembered
Say once how you count MER or which orders are wholesale. She uses it in every report, and tells you when two tools disagree.
We run BetterPic, our own online brand, on Patricia.
BetterPic sells AI headshots online. Its ads, reports, search work and outreach run on Patricia. The six threads above come from there.
Pricing
One price for the whole store. Every seat included.
Pay for the work she does. A spend cap is on by default, and credits never expire.
22 morning store briefs, 20 product descriptions, and 40 refund checks in one example month
- 20,000 credits included every month
- Every seat included, never per-seat
- Metered to real work, spend cap on by default
- Credits never expire and roll over
- Smart Meeting Notetaker, save $20 per employee per month
What a task costs, on average
A refund checked against your policy, reply drafted
~$0.13
~56 credits
A product description in your brand voice
~$0.34
~150 credits
The morning store brief
~$0.68
~300 credits
A Black Friday plan: emails, ads and a schedule
~$2.66
~1,183 credits
Security and trust
Built to be trusted with your store's data.
She joins only the channels you invite her to and the tools you approve. Every action is logged, and by default a store change waits for a yes.
Your data stays yours
Never used to train anyone else's model, and deleted if you cancel. Each person sees only what they should.
Every action logged
A full audit trail of what she did, when, and on whose say-so. Nothing happens off the record.
She checks her own work
She checks her work against your brand voice and your saved definitions before you see it.
Sharper every week
The more feedback you give, the better she knows your store, your products and your customers.
Secure to connect
The compliance most tools cannot show.
Connect Shopify, your ads and Stripe knowing the paperwork is done. Check it all in the Trust Center.
How your data is protected, in plain languageFAQ
Questions from store owners.
How do ecommerce brands use AI?
In two ways. Facing the shopper: search, recommendations, a chat widget. Behind the scenes: reading the numbers, checking ad spend, catching stock problems, sorting support. Patricia does the second, across all your tools, in Slack or Microsoft Teams.
How is Patricia different from Shopify Sidekick?
Sidekick lives inside Shopify and sees Shopify. Patricia lives in Slack or Teams and sees Shopify together with your ads, Klaviyo, Stripe and helpdesk. You can use both: Sidekick for admin tasks, Patricia for questions that cross tools.
Is this ecommerce automation, like Shopify Flow or Zapier?
Not quite. An automation repeats the steps you built. With Patricia you ask for the outcome in plain words. She works out the steps, explains what she found, and asks before she changes anything, by default. Keep your automations.
Does Patricia work with Shopify?
Yes. Connect Shopify from the dashboard and she reads orders, products, stock and refunds. More than one store? Connect each one and give it a label.
Will she change ad budgets or prices on her own?
Not by default. Budget and campaign changes wait in the channel as a proposal, with the reason. A person approves, edits or skips it, and every change is logged.
Does she reply to customers?
She drafts. She groups repeat questions and drafts replies and macros. A draft goes to a customer only after someone on your team approves it.
How do I know the numbers are right?
Every report names its source and dates, and uses the definitions you saved: what counts as revenue, how you count MER and margin, which orders to leave out. When two tools disagree, she shows both numbers.
We already have a dashboard tool. Why add Patricia?
A dashboard shows a chart. Patricia explains what changed and drafts the fix: the budget move, the email swap, the reorder. Keep the dashboard. She does the work around it.
Slack or Microsoft Teams?
Both. Slack, and Microsoft Teams in beta. The whole team shares one Patricia, and each person reaches only the tools they may use.
Where is our store data kept?
Patricia is SOC 2 Type II audited. Your data is stored in the US (US East), kept apart from other customers, and never used to train anyone else's model.
What does it cost?
$45 a month for the whole team, with 20,000 credits. That covers, for example, 22 morning store briefs, 20 product descriptions, and 40 refund checks. Free credits come with the onboarding call, no card needed. Credits never expire.
Go deeper on each job.
Bring your store. Leave with tomorrow's brief scheduled.
Connect Shopify and one ad account on the call. You leave with the morning brief running.
No credit card.