About Loop Loot
Written the way we'd explain it to a business owner, not a marketing team.
The problem
Most small businesses have a handful of tasks that eat an evening every week: typing up the same quote from scratch, chasing an invoice someone forgot about, retyping a job application into a spreadsheet. None of it is hard. It's just repetitive, and it's always the thing that gets pushed to after hours.
What we've seen work
Before this, we spent years advising some of the largest private equity funds, consulting firms, and manufacturers in the world on how to actually use AI: not the slide-deck version, the version that changes how a team works day to day. Across companies worth billions and companies with twelve employees, the pattern held: the real value was never one enormous integration project. It was two or three specific, unglamorous tasks a team already did every week, automated properly. We looked at AI opportunities across more than 30 companies in the past three years and never found an exception.
Most small businesses have already dipped a toe in: someone on the team uses ChatGPT for something. Few have gotten past that, because the jump from "I asked a chatbot a question" to "this actually runs as part of how we work" looks like it needs a data team and a budget they don't have. It doesn't. It needs someone to find the two or three specific tasks worth automating first. That's exactly what the first conversation with Loop Loot does.
How it works
Start from a template, or describe your business
Pick a working example close to what you need and adapt it. Or, if nothing fits, just describe your business and the task in plain language through a chat, the same way you'd explain it to someone you just hired.
It proposes a concrete direction
What would start it (you sending an email, a schedule, you typing something in), what it would actually do, and what you'd need to have ready: a pricing rule, an example email, whatever the task calls for.
The software builds and tests it. No one else looks at your data
It runs against real inputs before you ever see it, not a hypothetical: nothing gets presented as "done" without having actually run first. This step is automated; no other person is involved in building or reviewing it.
You choose whether to approve the final thing
It stays off until you turn it on. And even once it's running, you're still the one who decides: most external actions, like sending an email or changing a stored record, wait for your go-ahead by default, not the other way around.
The rules it doesn't bend on
- It asks instead of guessing. If an input is ambiguous or a detail is missing, it sends you a question, not a best guess dressed up as an answer.
- Nothing external happens without review, by default. An email to a customer or a change to stored data waits for your approval unless you explicitly turn that off for a specific step.
- It never invents a fact. An email address, a price, a name: if it isn't in what you gave it, it won't make one up to fill the gap.
What it can't do yet, said plainly
It can't send a separate message to each item in a list in one run yet: today a daily digest of, say, five overdue invoices comes back as one email with five clearly separated sections, not five emails. It can't read a bank account or any other authenticated external system on its own. And it doesn't yet remember, on its own, that it already reminded you about something yesterday. Some workflows repeat a reminder each run until you tell them otherwise. None of this is hidden behind vague language; you'll hear about a real limit like this directly, before you approve something built around it.