01AI operations

Systems that dothe work daily..

Marketing MIX is an international marketing studio founded in Kyiv in 2014 and headquartered in Ottawa since 2024. Below are five AI systems we built for our own operations and keep running in production. Each one replaces a job somebody used to do by hand.

02What runs
  • 01

    A digital colleague inside the CRM

    Staff ask her in plain language, she answers from live CRM data.

    Runs on: Festival CRM · our own worker on a local machine

    How it works

    1. 01A question goes into a queue; a worker picks it up and answers from the same records the person is allowed to see.
    2. 02Two passes: one drafts the answer from the data, the second checks it against the request before it appears.
    3. 03Her reach is the employee's own permissions, recalculated on every call.
  • 02

    A watcher over work chats

    Reads the chats nobody has time to read and reports what matters.

    Runs on: Telegram chats · CRM finance queue

    How it works

    1. 01Every morning the owner gets one short digest: how each chat moved, which ones went quiet, which questions are still open.
    2. 02A photo of a receipt in a chat lands in the finance queue of the CRM with the sum attached.
    3. 03Requests for access turn into a card for the manager instead of a lost message.
  • 03

    Volunteer check-in by geolocation

    People mark themselves on site; the server decides whether to trust it.

    Runs on: Festival CRM · volunteer cabinet in the browser

    How it works

    1. 01Two taps in a personal cabinet: I'm here, I'm done. Hours are counted from that.
    2. 02The server compares coordinates with the venue and refuses readings that are too far or too vague.
    3. 03Anything doubtful is flagged for a coordinator instead of being silently accepted.
  • 04

    A photo quiz in place of a contact form

    The visitor taps pictures, the studio gets a qualified enquiry.

    Runs on: mmix.biz · edge worker · advertising API

    How it works

    1. 01Pictures instead of fields: what to move, from where, how far. Nobody writes an essay.
    2. 02The answer lands as a lead with its source and campaign attached.
    3. 03The same event id goes to the ad platform, so one enquiry stays one enquiry.
  • 05

    An agent that hunts and negotiates

    Give it a thing to buy; it finds, bargains and brings back a deal.

    Runs on: Local marketplaces · Telegram · a separate browser profile

    How it works

    1. 01Searches the local marketplaces on a schedule and reads the listings the way a person would.
    2. 02Writes to sellers, negotiates within the rules it was given, and never commits on its own.
    3. 03When a price is agreed, the owner gets one message with the deal and the pickup.
03Why this page exists

We run all of it ourselves first.

Everything above carries our own operations: our festival, our chats, our leads, our purchases. We sell the same work after it has survived our own week.

04How we work with you

Three ways in.

  • 01

    Find the work worth automating

    We sit with your team, watch a week of real routine, and come back with the places where a machine pays for itself — and the places where it does not.

  • 02

    Build the system

    Assistants over your own data, pipelines between the tools you already pay for, agents that carry a task end to end. Yours to keep, with the decisions written down.

  • 03

    Train the team by the hour

    Hands-on sessions for owners, managers or the whole team, on your own tasks. Your people leave with prompts and habits that hold after we go.

Tell us what eats your week.

Write what your team does by hand. We answer with what we would automate first, and what we would leave alone.