Automations
Repeated work, carried by a system you own
Custom pipelines for the jobs your business does every week by hand. Built on your AI operating system, or on their own. A human approves anything that goes out.
What gets automated
Four kinds of work that come up in almost every business
Content
A week of posts researched, written, designed and queued for approval in one run. Pillars, captions, carousel images, platform adaptations.
Admin and follow-ups
Replies and follow-ups drafted from your own notes and history. Quiet leads chased. Onboarding questions answered from your own documents.
Reporting
The Friday report, the month-end pack, the pipeline summary, assembled from the tools you already use and delivered where you read it.
Research
Competitors, prospects, a market. Any website, any video, the platforms other tools can't reach, pulled in clean and summarised.
Worked example
The Structure & Self content engine, step by step
A physiotherapy practice in Naxxar needed six Instagram posts a week from one practitioner with a full diary. This is what runs now.
01
Topics
A Supabase table holds the topic bank, tagged by pillar. When a pillar runs short, a web search tops it up. Each topic is marked used once it ships.
02
Writing
For each day, a full post in the practice's structure: caption, creative direction, carousel slides, supporting Stories, hashtags, and a Facebook adaptation.
03
Guardrails
Clinical safety is applied to every post automatically: hedged language, no numeric thresholds, no diagnosis claims, an individual-assessment note.
04
Images
Carousel images generated in the brand palette and photo style, then graded before they're accepted.
05
Approval
Each post goes to a Telegram bot. The owner approves or rejects from her phone.
06
Publishing
Approved posts schedule through Buffer to Instagram and Facebook.
How a build is scoped
It starts with a short call about the work that keeps coming back. One workflow is a perfectly good project; most businesses have three or four candidates and we pick the one that's most repeated, most text-heavy and most visible to you. Then you get a fixed written quote before anything is built.
The build happens with working demos along the way, so you steer while it's in progress rather than reviewing a finished thing. Handover is a recording, written notes inside the workspace, and the approval step wired to wherever you actually look, usually Telegram or WhatsApp.
What you own
The code, the accounts, the data, the prompts. It runs on your subscriptions and your machine or your own hosting. If you want Plaios to keep an eye on it as the business changes, that's a light monthly arrangement agreed separately, and it's optional. Nothing stops working if you say no.
Where an operating system helps
An automation built on top of an AI operating system starts with the business already known: the voice, the offer, the clients, the tools. It needs less setup, drifts less, and every automation after it gets cheaper. It isn't required. It is usually the better order.
Questions
Before you book
What's a good first automation?
Something repeated weekly, text-heavy and visible to you: content, follow-up emails, a report you assemble by hand. One workflow, done properly, is a good project on its own.
Do I need an AI operating system first?
Not always. An automation can stand alone. But automations built on top of an OS know your business already, so they need less setup and drift less.
How does approval work?
Anything public gets a human check before it goes out. For Structure & Self, that's a Telegram bot where the practice owner approves each post from her phone.
What does it cost?
Scoped per workflow after a short call, with a fixed written quote before the build starts.
Who maintains it?
You own the code and the accounts. Most clients want a light monthly check-in for fixes and improvements as the business changes; that's agreed separately, and it's optional.
What about GDPR?
Automations run under your accounts, on your data, with the AI provider you choose. Nothing is stored by Plaios. Where personal data is processed, we design the workflow around that from the start.
Tell us the job you keep doing by hand.
Twenty minutes to find the right first workflow. Fixed quote after, before any build.