Employee training reminders and staff broadcasts for Magnum

Magnum Cash & Carry, the largest retail chain in Kazakhstan, assigns training to its staff in an LMS that sends no notifications. azamat.ai built a separate engine to send them. It follows about 50,000 course assignments, reminds employees through the company’s Telegram bot and sends each manager a daily list of new courses on the team and who is behind.

In the LMS Deadline passed
The engine Tells the manager who didn’t finish
EmployeeManager
Magnum Life bot
  1. Last month

    Hello, Olga Viktorovna!

    Your team has been assigned: Food safety rules.

    • Aliya Zhumabayeva
    • Timur Akhmetov

    11:00
  2. New course

    Hello, Olga Viktorovna!

    Your team has been assigned: Customer service basics, due by the end of the month.

    • Aigerim Nurlanova
    • Daniyar Seitkali
    • Yerlan Abdrakhmanov

    11:00
  3. After the deadline

    Hello, Olga Viktorovna!

    The deadline for Customer service basics has passed.

    Not completed:
    • Daniyar Seitkali

    11:00
Main menu
Translated from Russian.

The second part is a broadcast tool for HR. AI drafts a message from company documents in Russian and Kazakh, a person approves it, and it goes only to the chosen city, branch or department.

  • Employees get weekly reminders until they complete a course, and managers get a daily summary of who is behind.
  • HR broadcasts in Russian and Kazakh are drafted by AI, approved by a person and sent only to the chosen audience.

Why assigned training went unfinished

Magnum assigns training to its staff in an LMS, and every course has a deadline. The LMS recorded the assignment and stopped there. Employees heard about a new course only if someone remembered to tell them, so many put it off or forgot, and store managers had no easy way to see who was behind.

By the time a deadline came, a large share of courses hadn’t even been opened. Big store chains know this pattern well: the information sits in one system and never reaches the shop floor. We look at it more closely in AI for retail in Kazakhstan.

  • Employees forgot about assigned courses or kept putting them off.
  • Managers couldn’t see who on the team was behind.
  • Nobody warned people that a deadline was close.
About 50,000 course assignments for over 12,000 employees
  • not started
  • in progress
  • completed

Over 40% of all course assignments not even started

How training reminders and manager summaries work

The engine follows each course from the day it’s assigned. The employee hears about it straight away, then gets a weekly reminder while the course is open and a warning shortly before the deadline. Once the course is complete, the bot thanks them and the reminders stop. If the deadline passes first, the employee is asked to finish as soon as possible.

The manager sees the same course from the team’s side: who has been assigned it, a daily summary of who is behind, a note when someone finishes and, once the deadline passes, a list of who didn’t. Tests follow the same pattern, from a passed test to a reopened retake. Nothing is sent at night.

Employee
Manager
Who didn’t finish Manager

Customer service basics

The deadline has passed. Not completed:
• Daniyar Seitkali

Translated from Russian.

Staff broadcasts by city, branch and department

For news that isn’t about courses, HR has an admin panel. AI drafts the message in Russian and Kazakh from company documents, and a person checks it. Then HR narrows the audience step by step, from a city to a branch to a department, and the message goes only to those people. Questions about the documents themselves go to Olzhas, Magnum’s AI knowledge assistant.

AI draft

Dear colleagues! A reminder of the rules for the daily morning briefing in Magnum stores.

Approved by HR
City
  • Almaty
  • Astana
  • Shymkent
Branch
  • Stores
  • Warehouses
  • Head office
Department
  • Sales staff
  • Directors
  • Security
Store sales staff in Almaty Employee bot · delivered

How the LMS integration is built

The engine runs as a separate service next to the LMS. Through the day it reads course and test results, notices what has changed since the last check and decides who should hear about it. Messages go out through Magnum’s employee bot in Telegram, and HR broadcasts can also go to WhatsApp.

An admin panel holds the audiences, schedules and approvals, and every release goes out through CI/CD. The same engine can sit next to another LMS or HR system. That’s how we usually approach AI for HR: we connect to the systems a company already runs.

Stack

Python the engine and the admin panel
PostgreSQL employees, assignments, message log
Redis schedule and task queue
Employee bot API delivery to Magnum’s Telegram bot
WhatsApp second channel for broadcasts
LLM broadcast drafts from company documents
Admin panel audiences, schedules, approval
CI/CD tested releases

Safeguards against duplicate and stray messages

The engine works on live data for the whole company, so its safeguards are written into the code, and the team can pause sending at any moment.

  • No duplicates

    Every message has its own key, so the same reminder can’t go out twice.

  • Once per person

    An employee who belongs to several audiences still gets a broadcast only once.

  • Nothing out of date

    Courses that closed long ago are recorded and never brought up again.

  • A bad number doesn’t hold up the rest

    An invalid phone number is filtered out before sending, so everyone else still gets their message.

  • A kill switch

    One switch pauses all sending. Messages then wait in a queue until the team has checked them.

  • A gradual rollout

    Delivery can open to a chosen list of phones first, so new templates are checked on real messages before everyone gets them.

  • Quiet at night

    Messages go out only during the day. Anything that falls due overnight waits until morning.

What changed for HR and managers

The engine follows about 50,000 course assignments for over 12,000 employees and more than a thousand managers, and no one has to track progress by hand.

  • Employees Weekly reminders until completion

    Training reminder

    Please remember to complete Customer service basics before the deadline.

  • Managers A daily summary of who is behind

    Customer service basics

    Not finished yet:
    • Aigerim Nurlanova
    • Daniyar Seitkali

  • HR Broadcasts in minutes, approved by a person

    Morning briefing rules, for all store staff in Almaty.

    Approved and sent

Built in: no duplicates nothing out of date quiet at night

FAQ

Connect a reminder engine to the LMS. It reads who has been assigned which course, messages each employee in a channel they already use, such as a company Telegram bot, and keeps reminding them until the LMS marks the course complete. A warning goes out as the deadline gets close, the reminders stop once the course is done, and nothing is sent at night.

By reading and comparing. The engine regularly pulls course and test results through the LMS API, compares them with the previous read and turns each change into an event: a new course, a completion, a deadline that has passed. The LMS itself barely changes. Magnum only added managers’ contact details so that summaries know whom to address.

Yes. In Magnum’s admin panel AI drafts the message from company documents in Russian and Kazakh, a person reviews and approves it, and HR picks the audience by city, branch, department or position. The message then goes out through the employee bot in Telegram and can also be copied to WhatsApp.

Through the company bot. When people on the team get a course, the manager receives the list with due dates, then a daily summary of who is behind and, after the deadline, a list of who hasn’t finished. Nobody has to open LMS reports.

Yes, as long as the system can share course results through an API or a regular export. The engine only needs assignments, statuses and deadlines. The rules for when to remind people and whom to tell stay the same.

Does your LMS or HR system stay silent too?

We connect to systems that have no events, rebuild the events from their data and deliver messages in the channel your staff already use, with a person approving anything an AI step writes.

Brief (optional)