Automation
Control and automate your tasks
AirMind is a natural-language interface to the whole of Airdec. It reads every number in your account and runs the task you describe. It is in development.A natural-language interface to the whole of Airdec. It reaches every number in your account and does not stop at answering. Describe the task and it carries it out in the product, then repeats it every Monday if you ask. It is in development; what follows is what it will do.
One text box instead of the whole app
Ask, and it answers. Tell it, and it does the work. Say it once, and it keeps doing it.Three verbs, one interface. Ask, and it answers from your data. Tell it, and it does the work inside Airdec. Say it once, and it keeps doing it every week without being asked again.
It reaches all of your data
“Why did leads drop in Plovdiv last week?” is a question, not a filter combination. AirMind works out which figures answer it and pulls them from the same aggregators the dashboard uses: campaigns, leads, funnel, geography, reaction time, points of sale.
It carries out the task
Reassign a batch of leads, move a target, add a point of sale, pause the form that is filling up with spam. The work you would click through five screens for becomes one sentence.
It operates Airdec, not a copy of it
AirMind drives the product through the same interfaces your team does, under the same account and the same permissions. It cannot reach a screen you could not open yourself, and every change it makes shows up in the audit trail with its name on it.
Say it once, it repeats
“Every Monday at nine, send me the branches that missed their target.” Described once in the chat, it becomes a standing routine. No builder, no trigger dropdowns, no automation tab to learn.
Real components, not a paragraph
The answer arrives with the KPI cards, charts and lead lists rendered inline. They are the same components as the rest of Airdec, drawn from server payloads rather than retyped by a model into a sentence.
It will not invent a number
Every figure AirMind shows comes from a tool result it can point at. The model writes the sentence; the database writes the number. That separation is the design, not a setting.
What you would type
No syntax to learn. The left column ends in an answer, the right one ends in something having changed.
Where it stands
Built so far, and honestly
AirMind is in development. This is the order it arrives in.
Reading is built
Nine internal tools are live, each reading the aggregators the dashboard already displays.Nine internal tools are live: leads, campaigns, funnel, geography, reaction time, points of sale and the rest. Each one reads the aggregators the dashboard already displays, so an answer and a chart cannot disagree.
The conversation is next
Streaming replies, the tool loop and the inline components that let an answer render a real KPI card. Being built now.Streaming replies, the tool loop, and the inline component protocol that lets an answer render a real KPI card instead of describing one. This is the part being built now.
Then it acts, and learns your business
Tools that change things rather than only read them, standing routines, and a knowledge base of your own vocabulary.Tools that change things rather than only read them, each one permission-checked and logged. Then standing routines, then a per-organisation knowledge base of your products, regions and naming, so AirMind answers in your vocabulary rather than in generic marketing language.
AirMind, in detail
Can I use AirMind today?
Not yet. It is in development. The data layer is built and tested; the chat surface and the acting tools are not. We would rather say that than ship a demo that answers three rehearsed questions. Book a demo and we will show you what the current assistant page does in the meantime.
What will it be allowed to change?
Whatever your own account can change, and nothing else. AirMind acts as you: the same role, the same permissions, the same audit trail. Anything destructive is confirmed with you in the conversation before it runs, and every action it takes is listed under its own name afterwards.
How is this different from asking ChatGPT about my numbers?
A general model has to be given the numbers, and will happily do arithmetic on them, which is where the wrong figures come from. AirMind never computes: it calls a tool, the database answers, and the model narrates what came back. And a general model cannot press a button in your CRM; this one is wired into it.
Will it replace the dashboard?
No. The dashboard is the answer to the question you ask every day. AirMind is for the ones you ask occasionally, where knowing which screen to open is half the work, and for the jobs where the screen was never the point, only the clicking.
What happens to my data?
The assistant service holds no database credentials at all. It receives the figures it needs over an authenticated internal call and returns text. There is no copy of your leads sitting inside an AI service.