AI agents: Management and analytics

What it does: Losses in tenge before purchase and AI's contribution after.

What a person decides: The final decision on the outcome.

For whom: Owner.

How to measure: Revenue brought in by AI.

What it does: A morning summary and three decisions for today.

What a person decides: Decisions.

For whom: Manager.

How to measure: Time to get the day's picture.

What it does: Conversion by stage and bottlenecks.

What a person decides: Actions.

For whom: Manager.

How to measure: Funnel conversion.

Example result: “Situation: in September, 212 leads → 98 bookings → 69 visits → 41 payments. Key point: the biggest drop is between booking and visit — 30% of those who booked didn't show, up from 18% a month ago. Options: 1) confirm every booking a day ahead — cheap, catches people who forget; 2) prepayment for bookings — cuts no-shows more, but some customers drop off before booking. Recommendation: 1, then compare in a month. Today: assign the front desk to confirm tomorrow's bookings.”

What it does: Metrics and best practices by branch.

What a person decides: Standards.

For whom: Manager.

How to measure: Gap between branches.

What it does: A forecast based on history and seasonality.

What a person decides: The plan.

For whom: Manager.

How to measure: Forecast accuracy.

What it does: Goals, hypotheses and a quarterly plan.

What a person decides: Approval.

For whom: Owner.

How to measure: Plan completion.

What it does: Outcomes and tasks from a meeting recording.

What a person decides: Tasks.

For whom: Staff member, manager.

How to measure: Meeting decisions carried out.

Example result: “Minutes of the October 7 team meeting. Attendees: Yerlan, Dinara, Marat. Decisions: 1) ‘Autumn cleaning’ promotion from October 15 — owner Dinara, copy by October 12. 2) New price list — Yerlan, for approval by October 20. 3) Confirming bookings a day ahead — Marat, starting Monday. Open question: ad budget for the promotion — decide by Friday. Next meeting — October 14. To decide: create a task for each item.”

What it does: An agenda from open tasks and outcomes with owners.

What a person decides: Decisions.

For whom: Manager.

How to measure: Meeting tasks completed on time.

Example result: “Situation: 23 open tasks; 4 overdue, 3 waiting for a manager decision, 2 with no owner. Key point: only these 9 go on the agenda; the rest are on track. Options: 1) a 20-minute meeting on just the 9 tasks — fast; 2) review all tasks — more complete, but takes an hour. Recommendation: 1. Agenda: AC repair — no owner; Arman invoice — overdue; new location opening — waiting on a lease decision, plus 6 more items. Outcomes after the meeting: decision, owner, deadline. Today: send the agenda to the team before 9:00.”

What it does: A risk summary based on company data.

What a person decides: Measures.

For whom: Owner.

How to measure: Risks spotted in advance.

What it does: Learns from edits as style examples.

What a person decides: Removing examples.

For whom: Everyone in the company.

How to measure: Share of replies with no edits.

Example result: “Style memo from the last 10 reply edits. 1) Polite ‘you’ and first names — corrected 6 times: ‘Dear customer’ became ‘Hi Aigerim.’ 2) Price in the first line, not at the end: 4 edits. 3) No ‘unfortunately’ at the start of a reply: 3 edits. 4) In messengers — no more than 3 sentences. Conflict: one edit added an emoji, another removed it. To decide: emojis — yes or no.”

What it does: Sends an announcement to the right groups, collects “read” confirmations and reminds those who haven't read it.

What a person decides: The text and recipients.

For whom: Manager, staff member.

How to measure: Share who have read it.

What it does: Turns a manager’s note into a clear task: outcome, deadline, owner, and how to check it.

What a person decides: Assigning the task.

For whom: Manager.

How to measure: Tasks done on time.

Example result: “Task: update the price list on the website. Owner: Aigerim. Outcome: the Prices page shows the new prices for 8 services from the accounting file, with no old ones left. Deadline: Friday, October 17, 18:00. How to check: the manager opens the page and compares it with the file. Missing: who grants access to the site — please confirm.”

What it does: Turns a voice note transcript into a task list with owners and deadlines.

What a person decides: Who gets what and by when.

For whom: Manager.

How to measure: Manager ideas that became tasks.

Example result: “Situation: the note has 5 tasks. Key point: two are urgent — sign the Bereke contract by Thursday (Arman) and reply to customer Assel’s complaint today (Dinara). The other three have no deadline — needs clarifying. Options: assign everything now, or the urgent ones today and the rest at the team meeting. Recommendation: urgent ones now. Today: tasks with deadlines for Arman and Dinara.”

What it does: Turns a long work chat into decisions, tasks with owners, and open questions.

What a person decides: Assigning tasks.

For whom: Manager.

How to measure: Time spent reading chats.

Example result: “Situation: the sales team chat had 214 messages this week. Key point: the promo moves to November 1 (Marat), and Aigerim updates the price list by Friday. Open question: who takes over the Bereke account now that Timur has left. Recommendation: assign an owner at the team meeting. Today: confirm the price list deadline with Aigerim.”

What it does: A project plan by task: milestones, owners, what’s blocked, and the risks.

What a person decides: Project deadlines and resources.

For whom: Manager.

How to measure: Projects delivered on time.

Example result: “Situation: the Online Booking Launch project has 6 tasks, 2 done. Key point: payment setup is blocked waiting on the bank contract, and the November 1 launch depends on it. Options: 1) launch booking without online payment — we hit the date; 2) wait for the contract — the date slips. Recommendation: 1. Today: call the bank and get a signing date.”

What it does: A launch plan from the project tasks: price, copy, staff training, announcement — what’s ready and what’s in the way.

What a person decides: Price and launch date.

For whom: Manager.

How to measure: Launches without delays.

Example result: “Action: move the teeth whitening launch from November 1 to 8 — front desk training hasn’t started, and without it bookings will get mixed up. Ready: price approved, website copy written. Impact: a launch with no booking errors. Risk: the announcement is scheduled for November 1 — update the date. Fallback: if training is done by October 30, go back to November 1.”

What it does: An opening checklist from the project tasks: documents, staff, cash register, advertising, deadlines — what’s urgent.

What a person decides: Opening date and budget.

For whom: Manager.

How to measure: Locations opened on time.

Example result: “Situation: 3 weeks until the Satpayev St. location opens; 7 of 12 tasks done. Key point: the cash register isn’t registered, and you can’t open without it. Options: 1) file the registration today; 2) open later. Recommendation: 1. Today: have the accountant file to register the cash register.”

What it does: Rules for staff: what can and can’t go into AI and who checks the output — based on company policies.

What a person decides: Approving the rules.

For whom: Manager.

How to measure: Data handling violations.

Example result: “Action: approve three AI rules. 1) Never send customers’ personal data — national ID numbers, phone numbers, diagnoses — to AI (personal data policy). 2) Staff review any AI-written customer message before sending. 3) Prices and terms come only from the company price list. Impact: lower risk of data leaks. Risk: people will work around the rules — assign an owner. Fallback: revisit the rules in a month.”

What it does: What to hand off to whom during vacation: open requests and tasks, what can be decided without the boss, and what waits.

What a person decides: Who gets decision authority.

For whom: Manager.

How to measure: Issues that waited for the boss to return.