The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
A London ice cream maker was taking wholesale orders by email, in every format you can imagine.
Now an AI agent reads the wholesale inbox, matches each customer's shorthand to real products and creates the order. Production, traceability and kitchen checks are moving into the same Unstack workspace, so a trace or recall takes seconds.
290 orders created by the agent so far.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Section, a matcha shop that runs itself
The problem A lot of matcha brands skip the story. You buy a tin and never learn where the leaf grew, how it was shaded or how it was ground. And the instructions come as a little booklet in the box that ends up in a drawer by the second bowl.
The idea Tell the story on the website, and move the instructions there too. Section is a concept matcha brand: the site follows one leaf from the field to your bowl, then sells one box with the matcha and the five tools a barista uses. There's no booklet in the box. The guide lives on the site: every part lettered A to F, four drinks, each step read aloud, and pages that turn by themselves while your hands are busy. Live: pure-section-space.base44.app
Automation depth Behind the counter, the shop runs itself:
Saves time: an order lands on the packing board the moment it's placed, with where it ships. Nobody copies orders out of an inbox.
Improves operations: stock counts itself. Every order takes its items off the shelf (a box takes one of each part). At 3 left the shop says "Only 3 left" and I get an email. At zero, the item sells out on its own.
Increases leads: a sold-out item offers "Tell me when it's in". That's a list of people who want to buy. When I type a new count, each of them gets one email.
Enhances the customer experience: the buyer gets an email at every step: got it, packed, on its way. The last one links to the guide, so the steps arrive before the box does.
The packing board is mine, behind a sign-in, and addresses are kept where only I can read them.
** It all runs on Base44: its database, Google sign-in, one backend function for the emails, and its hosting. **
Design quality
One idea runs through the whole brand: everything is drawn in one black line, and the only colour is the matcha's green.
The first screen is a small matcha bar. Scroll, and it opens into its own drawing.
The phone layouts are designed, not shrunk. The box, its price and its key stay at the foot of the first screen, and the guide's Back and Next stay under your thumb.
How I built it I tried style directions in Figma, built the site and automation in Base44, and I published from there. The video shows both sides: buying the box, and running the shop.
Real-world use Section is a concept, and no payment is taken. Everything else is built to run as a real shop: real orders, real emails, a locked staff board and private addresses. Add a payment provider, and a matcha brand could start selling with it.
Ohhhhh this is very nice
Section, a matcha shop that runs itself
The problem
A lot of matcha brands skip the story. You buy a tin and never learn where the leaf grew, how it was shaded or how it was ground. And the instructions come as a little booklet in the box that ends up in a drawer by the second bowl.
The idea
Tell the story on the website, and move the instructions there too. Section is a concept matcha brand: the site follows one leaf from the field to your bowl, then sells one box with the matcha and the five tools a barista uses. There's no booklet in the box. The guide lives on the site: every part lettered A to F, four drinks, each step read aloud, and pages that turn by themselves while your hands are busy.
Live: pure-section-space.base44.app
Automation depth
Behind the counter, the shop runs itself:
Saves time: an order lands on the packing board the moment it's placed, with where it ships. Nobody copies orders out of an inbox.
Improves operations: stock counts itself. Every order takes its items off the shelf (a box takes one of each part). At 3 left the shop says "Only 3 left" and I get an email. At zero, the item sells out on its own.
Increases leads: a sold-out item offers "Tell me when it's in". That's a list of people who want to buy. When I type a new count, each of them gets one email.
Enhances the customer experience: the buyer gets an email at every step: got it, packed, on its way. The last one links to the guide, so the steps arrive before the box does.
The packing board is mine, behind a sign-in, and addresses are kept where only I can read them.
** It all runs on Base44: its database, Google sign-in, one backend function for the emails, and its hosting. **
Design quality
One idea runs through the whole brand: everything is drawn in one black line, and the only colour is the matcha's green.
The first screen is a small matcha bar. Scroll, and it opens into its own drawing.
The phone layouts are designed, not shrunk. The box, its price and its key stay at the foot of the first screen, and the guide's Back and Next stay under your thumb.
How I built it
I tried style directions in Figma, built the site and automation in Base44, and I published from there. The video shows both sides: buying the box, and running the shop.
Real-world use
Section is a concept, and no payment is taken. Everything else is built to run as a real shop: real orders, real emails, a locked staff board and private addresses. Add a payment provider, and a matcha brand could start selling with it.
Walkthrough: first comment
4.4%. That is the bottom 90% of firms' share of OpenAI + Anthropic spend, in the Ramp data Apollo's chief economist published Oct 24.
The top 10% of firms pay the other 95.6%.
I run a one-person company on AI agents. I am one of the tiny spenders.
What Torsten Slok wrote (Apollo, Oct 24; TechCrunch picked it up today):
→ "Revenue per worker is rising for the Mag 7 but is flat for the S&P 493"
→ "Profit margins are rising for the Mag 7 but are flat for the S&P 493"
→ Of the top 1% of AI spenders, 80.4% are VC-backed. Of the median spender, 13.1%.
Hedge: these are firms on Ramp, not all of corporate America.
The line that landed: mainstream firms now view token spending "not as a source of growth but as a cost center that weighs on margins."
He is right about my books. Tokens are a cost line, not a growth line.
The payoff does not show up at the model level. It shows up per task, or not at all.
What I watch for:
Rule 1: Price the task, not the model. If the tokens cost more than the hour they replace, the task goes back to me.
Rule 2: Judgment stays with me. Slok: "AI may be a great calculator, but in business, intelligence is mostly judgment." The agents draft, I sign.
Rule 3: The latest model is not a default. Corporate America, per Slok, "does not need the latest model". Same here, task by task.
Slok says the AI story "will be decided in the S&P 493, not the Mag 7."
I think it is decided one task at a time, inside the 4.4%.
Your token bill: cost line or growth line?
