For executives
Personal agentTime for management, not for searching chats
Assignments, commitments, and meeting prep are gathered in one place.
Potential with stable agent operation. Not a measurement of savings.
For owners and teams: incoming requests, agreements, finance, and marketing. We pick the role and set the agent up for your process.
In Telegram we'll discuss your task and the next step. Diagnostics and implementation are agreed separately.
30 office chairs for a new location
Over the last 4 months
30+ companies
200+ agents
Pick a task — see an example of the finished result and the agent that does it.
Hello! We need 30 office chairs for a new location. Can you deliver to Tashkent by Friday?
Hello! Which model are you interested in, and which district should we deliver to? I'll clarify the details and pass it to the manager to check availability and lead time.
Replies to incoming messages on its own. A repeat inquiry — only after your approval. Prices and the decision on a deal stay with a person.
All that's left for the manager is to check availability, price, and delivery.
Another task? Every role is in the catalog, with search.
Full catalog →Our team has implemented 200+ AI agents in 30+ companies, from personal assistants to department-level teams.
Here are 11 stories from our practice. Not every implementation is written up as a case, and some projects cannot be disclosed because of non-disclosure agreements (NDAs). We do not publish client names.
For executives
Personal agentAssignments, commitments, and meeting prep are gathered in one place.
Potential with stable agent operation. Not a measurement of savings.
Clothing retail chain
31 agentsWarehouse, purchasing, finance, and sales work from one picture of inventory.
Result of a retail chain in roughly two months.
IT company · ERP
8 agentsInstead of an incomplete report, missed calls and promised callbacks are visible.
July → October 2026, snapshot as of October 29. People and agents worked together.
Restaurant chain
5 agentsOne overview of the register, delivery, call center, and guest database, instead of scattered reports.
A segment of guests who stopped coming was found. This is not the number of customers brought back.
Services company
24 agentsThe agents take over part of the work on content, data, and process control.
Result of a company about three months after launch.
Residential developer
11 agentsAnalysis of calls and CRM shows objections, managers' mistakes, and growth points.
Scenario: one additional deal per 100 meetings. Growth not yet achieved.
Education
12 agentsReconciling CRM, calls, and documents shows where the team needs to step in.
A problem that was identified. Not the number of customers or payments recovered.
Agricultural production
1 agentA daily reconciliation of work chats, warehouse, and ERP helps catch a problem sooner.
A specific production-control episode from the case.
Steel and metallurgical plant
Personal agentThe executive gets a summary of production, assignments, and documents before signing.
The volume of assignments recorded, not the number of tasks completed.
Construction company
3 agentsConstruction, procurement, and finance link requests, confirmations, and the payment calendar.
18 + 22 + 16 checks. A launch-stage result, not a measured financial outcome.
Product IT company
24 agentsProduct, development, sales, and finance are united under shared oversight.
Result about a month and a half after launch.
Entrepreneurs from Uzbekistan on how they use agents in their work.
Entrepreneur, retail (Turkish Baby), Tashkent
Otabek handed the agent the checking of reports from the work group. The agent finds errors in calculations, suggests corrections, and flags possible problems at the till. The owner gets an extra check on the financial information he relies on to manage.
From the client's testimonial
«I double-checked it myself — all the calculations add up»«I implemented AI in my business as an agent that oversees my ERP system, and made it my personal assistant. It helps me manage employees — especially to oversee the work of the finance department and managers. I added it to the work group: it takes my routine tasks and reports and checks them. I double-checked it myself — all the calculations add up; we've reached the point where it can be trusted calmly. If the finance people make a mistake, the agent finds it right away and gives a solution and advice. With its help we're building our own ERP system — it sees shortcomings in the work and helps fix them, and immediately flags possible problems at the till. As a personal assistant it leads me toward my goals: I uploaded my workout plan and big business goals — and, like a good conversation partner and mentor, it suggests what to focus on. Every day it prepares summaries of books and reminds me of prayer times. I use it for networking too — a database of acquaintances and partners is being built: who can help with what, who works where. I ask about a person — it finds them in the database.»Read in fullCollapse
FMCG distribution, Tashkent
In work chats, people discuss money, make arrangements, and wait for replies. The agent assembles a daily summary for Kobil: who he hasn't replied to, which promises are still unfulfilled, and which questions he needs to come back to. It helps him restore context and decide what to do next.
From the client's testimonial
«It reminds me of all this and pulls it together into one complete report»«What I like most about the agent is that it tracks my chats and Telegram conversations in real time, keeps context well, and tells me who I need to reply to and what about. And every day it produces a real report straight from that context: who wrote to me and I didn't reply; who talked about money; it reminds me of agreements — who asked me to do something that I haven't done — it reminds me of all this and pulls it together into one complete report. These things are great! This is how I set up my own agent.»Read in fullCollapse
Serial entrepreneur (restaurants, cafes, retail), Tashkent
Twice a day, Sherzod and his managers get a sales summary — with figures and suggestions on how to increase them. The agents also help with payroll sheets, reports, and tracking of assignments. Managers have a regular basis for decisions about sales and the work of the team.
From the client's testimonial
«The data really helps»«The agents help me and my managers: we set up a sales summary twice a day — with figures and ways to increase them. The data really helps. They digitize processes, from payroll sheets to reports, help manage staff, delegate tasks, and monitor their completion. Next month we'll save almost UZS 100 million on the payroll fund.»Read in fullCollapse
The savings amount in the testimonial is the client's forecast for next month, not an achieved result.
Entrepreneur, education, Tashkent
«After implementing the agent, the first thing I started doing was automating my personal life and work. For example, I send the agent a meeting — and it immediately puts it in my Google Calendar, writing in the description the topic and… the questions I need to ask this person. Then I fully automated finances: we set up a sheet in Google Sheets, I enter income and expenses, and at the end of the month the agent suggests what can be cut and what to pay attention to. I gave it access to my Notion and CRM — and it started finding bottlenecks in the business. That's how we realized we needed a head of sales. The agent drafted the job description, we posted it on HeadHunter, and started interviewing. After each interview I sent it the audio recording, and it analyzed the candidates and gave feedback. We hired one in the end — the agent made him an offer and helped with onboarding: a checklist, a monthly and a weekly plan. I'm also running customer development interviews — my goal is a hundred of them. I retell the conversations to the agent by voice, and it gathers insights: what to implement, what product to build, where to market. And all this in less than a month. I recommend it to everyone. When I heard how much this agent costs, I thought: well, okay. But after using it, I understood right away — this isn't a product for that price. It should cost a lot more!»Read in fullCollapse
Entrepreneur, manufacturing, consulting, Tashkent
«After I installed this assistant, work sped up — if not 10 times, then definitely 5. Every morning it gives me a short digest: what can be done, what's left, what deserves attention. I've handed the calendar over to it completely, and… projects too. Even where I don't attach importance to something myself, it finds it and says: 'pay attention here.' I run my calendar, projects, and finances through it — right now I'm recording personal finances, and it calculates everything. It gives me a report whenever I want, not just at the end of the month. Little by little I'm connecting the company too, so it keeps its financial reporting as well. I added health as well. The most interesting thing: whatever idea comes up, whatever I want to load onto it — first I think 'will it handle this?', but it handles it. It does everything well, even better than I expected. I'm very pleased and happy that I installed this AI assistant.»Read in fullCollapse
The speed improvement is the client’s personal estimate.
We start with one agent, where we expect the greatest business return relative to implementation cost.
First, a free conversation. We'll discuss your task and define the next step. Then the work follows a clear sequence:
Before diagnostics, we sign a contract and a non-disclosure agreement. Then we move on to the company's internal processes and data.
We review the team's processes, tools, and routine work. We choose the first task by its potential return and define what to teach the agent and what to connect it to.
We agree on the role, tasks, connections, action boundaries, and success criteria. You know in advance what work you are handing over and how to accept the result.
You make the upfront payment after approving the specification. By then, the agent's tasks, scope of work, and implementation terms are clear.
We connect the tools, configure the agent, and test it on real work tasks. We show your team how to assign tasks, assess results, and adjust the rules.
A clear timeline before launchAfter the specification is approved, the upfront payment is made, and the necessary access is provided: approximately 3 days to 2 weeks. We confirm the exact timeline for your task.
Care — agent updates and supportCare includes agent updates, bug fixes, priority support, and adjustments to settings and connections within the agreed role. The first month is free from acceptance. After that, we agree on the cost for your company based on the number of agents.
We start with one task. You know in advance what's included in implementation and which expenses remain after launch.
Process review, agent setup, tool connections, and testing in real work. We agree on scope and cost before starting.
Updates, bug fixes, priority support, and adjustments within the agreed role. The first month from acceptance is free; we agree on the ongoing cost in advance.
Required subscriptions, usage-based fees, and the infrastructure the agent runs on. The expense depends on the selected services and workload.
The first conversation is free. We agree on in-depth diagnostic terms separately; new roles and tasks require a separate scope agreement.
How to plan an agent budgetThe one where the agent has the greatest potential return relative to the cost of implementation — ROI. During diagnostics, we compare benefit, complexity, and data readiness. We start with one agent, assess its work, and then decide on the next step.
The first conversation is free: we discuss your task and define the next step. In-depth diagnostics follows the contract and NDA: we study processes and internal data, select the first task, and prepare the agent specification. We agree on diagnostic terms before starting.
No. You talk to the agent in Telegram by text or voice. We handle installation and setup; you provide the task context and working rules.
A description of the process, examples of input data and the desired result, rules, and a responsible person. We agree on the services and permissions needed before connecting anything.
It depends on the role and agreed rules. For example, Support replies to incoming messages on its own, but sends a follow-up only after your approval. Prices and decisions on deals stay with a person.
Yes. An agent can misunderstand a task, rely on incomplete data, or give an incorrect answer. At launch, we test it on work examples and agree on how to review results. Important decisions and actions requiring approval stay with a person.
Before connecting, we agree on the data involved, access permissions and processing terms, taking Uzbekistan’s legal requirements into account. We are also introducing a protective layer that masks personal data in text requests sent to external AI services. We document the available safeguards separately for each project. External AI services and infrastructure may process data to carry out tasks.
Data handling details
The task, data readiness, connections, and amount of setup. Before starting, we agree on implementation cost, Care terms, and the necessary AI service and infrastructure expenses. The first conversation is free. Diagnostic terms are agreed before it begins; implementation is paid separately.
On azimai.uz, you choose an individual agent for a task. Our separate NOIB project builds agent teams for a department or company. Describe your task in the conversation and we'll suggest the right route.
An agent team for your business — NOIB
Tell us how the work is organized today and what you'd like to hand over. In the first conversation:
In Telegram we'll discuss your task and the next step. Diagnostics and implementation are agreed separately.
Implementation and Care: the azimai team, NOIB LLC. Founder — Azim Orifjonov ↗
Before connection: data and control ↗