AI agents
AI agents watch your business. You keep the decisions and the control.
Davirix AI employees work on the same database as your staff. They see sales, stock, debts, and cash, find problems on their own, propose a fix with evidence, and check whether it worked. You decide how much freedom each one gets.
The loop
Observe, propose, act, verify
- 01
Observe
Six signals every 15 minutes: low stock, stock-out, transfers stuck over 48 hours, overdue debt, overdue tasks, and cash discrepancies. Plus a daily snapshot of revenue, margin, cash, debts, and stock value.
- 02
Analyse
The AI executive splits the picture across sales, warehouse, finance, control, and procurement specialists and pulls their findings together.
- 03
Propose
Every proposal carries evidence, a severity, a short summary, and a concrete action. Duplicates are suppressed and stale proposals close after 72 hours.
- 04
Act
You accept, and the action runs as the AI employee through the normal command path: the same rights, policies, and audit as a person.
- 05
Verify
After the deadline it compares before and after: achieved, partial, or failed. If not resolved, it opens a follow-up task for whoever accepted.
AI employees
A ready AI employee for each part of the business
Created from a template in a nine-step wizard: objective, instructions, scope, autonomy, budget, approval, and schedule.
Watches the whole business with five specialists: sales, warehouse, finance, control, and procurement.
- Daily business review
- Proposals with evidence
- Result checks
Answers questions on stock, prices, and customer debt. Replies to customers on Telegram in the mode you choose.
- Stock and price answers
- Customer debt lookups
- Telegram replies
Tracks low stock, stock age, and sales velocity, and flags what needs moving or reordering.
- Low stock
- Slow and dead stock
- Sales velocity
Answers on ticket status, the queue, warranty terms, and serial-number history.
- Ticket status
- Warranty checks
- Serial history
A business advisor for the owner: reads the data, never changes it, and answers with a detailed analysis.
- Read-only access
- Detailed answers
- Owner only
Teaches staff how to use the system, in Uzbek and Russian, from more than a hundred guide pages.
- On every screen
- Uzbek and Russian
- 100+ guide pages
Autonomy
You choose how much freedom each AI employee has
Analysis only
Gives an opinion from the situation it's handed. No access to tools.
Draft
Reads data and prepares proposals. Changes nothing.
With approval
A person approves every action first. This is the default.
Automatic within policy
Works on its own inside the risk level and policy rules you set; anything beyond goes for approval.
Safety
AI can't get more rights than a person
- An AI employee is a separate actor; its role can never be wider than its creator's.
- Write rights need a dedicated policy rule for AI, limited by branch and amount.
- Every AI run and tool call is logged with its reason, confidence, and model.
- If no one is available to approve, nothing executes. There is no silent pass.
- Daily limits on runs, tool calls, and conversations.
- One setting switches AI off entirely, and the ERP keeps working.
Where the AI works
We configure the AI for your company
- Objectives, instructions, and scope for each AI employee, from your process
- Who approves which proposals, and which actions run automatically
- A knowledge base built from your company's documents
- A daily AI summary for managers on Telegram
- A voice agent and further AI actions
AI agents
What the AI agents do in this industry
Questions
Can the AI move money or delete documents on its own?
No. The AI only performs actions you allow through policy. Money and stock documents stay with people; the AI prepares tasks and proposals around them.
Which model is used, and where does our data go?
Today it runs on OpenAI models through the AI gateway. Each AI employee sees only the data within its scope, and AI can be switched off entirely.
What if the AI proposes something wrong?
Every proposal comes with evidence, and you can reject it with a reason. Accepted proposals are checked against the outcome, so you can see which ones actually helped.
What does it take to turn AI on?
During rollout we connect the AI provider, create AI employees from templates, and set up the observation schedule.
Next step
See the AI agents on your own data
In the demo we show the AI employees, the proposals screen, and the approval flow on an example from your industry.
- Live demo, not slides
- Rough estimate of scope and timeline
- Answers on migration from your current system
Or contact us directly