Task manager for assignment follow-up
Say it the way you would say it out loud. The system reads the sentence, sends the assignment to Telegram and keeps asking how the work is going until it is finished. Nobody has to learn anything: they get four buttons in the messenger they already have open all day.
Who it is for
Retail and service chains, restaurants and cafés, clinics, franchises, factories, construction, logistics. Typically 50 to 300 people, five sites or more, each with its own manager.
Twenty assignments come out of it. By the next one, nobody remembers half.
The only way to find out what is happening there is to call. Every day, round and round.
They will not fill in cards in a portal. They all have Telegram, and that is enough.
And you will not need one. Setup takes an evening; nothing here is harder than a notebook.
Who it is not for, plainly: engineering teams running sprints, agencies billing by the hour, and anyone who needs document routing and approvals. Those are different products, and we will not pretend to be them.
Why assignments go missing
As long as creating a task means opening a form and filling six fields, work will be handed out by voice. And nobody can check a voice a week later.
A spoken assignment does not formally exist. No owner, no deadline, nothing to open and look at. A week later two people hold two different versions.
Asking “so where are we?” is awkward for you and insulting to them. Not asking is not an option either: there is no other way to know.
That is what happens when people who work with their hands are pulled into a web portal. We do not pull anyone in: the bot is enough.
Then no tool helps at all. Which leaves us one requirement: assigning work here has to be faster than typing a message.
How it works
Voice, text, a forwarded message or the notes from a whole meeting. It turns them into assignments and asks about anything it did not understand instead of guessing.
They accept, propose a different deadline or ask a question and hand the turn back to you. All by buttons, all recorded in the assignment history.
As the deadline approaches, the system asks how it is going. One tap updates the card. You only get shown what went wrong.
Features
No “integrates with 500 services”. Just what actually gets used in a company where work is handed out by voice.
Your screen
No kanban, no Gantt, no endless list. Three zones, and the first two fit into a single glance over morning coffee.
Colour here means risk and nothing else. Priority, type and project are never colour-coded, otherwise red stops meaning anything at all.
Industries
One product, but every business hands out different work. Here are the typical stories.
Daily walk-rounds against a checklist, assignments to site managers after the stand-up, a shelf photo attached to the card. You can see which sites are slipping without calling all of them.
Morning checks, equipment repair requests, hygiene tasks. The shift changes, but the assignment stays with the role rather than with one person.
Work for front desk staff and head nurses, recurring checks. Anything touching patients is classified, and every card opening is on record.
Assignments to shift supervisors, fixing findings, daily rounds. People can reply straight from the floor without a long break.
Work for site foremen, defects captured with photos, “blocks” links between jobs. Deadlines move by agreement, not after the fact.
Requests by route and area, recurring equipment checks, follow-up on fixes. Overload shows up before someone actually stalls.
Pricing
Other systems charge for every employee. But in a chain of a hundred and fifty people, most of them sign in for a minute a day to press “done”. Paying a full seat for that makes no sense. Our subscription is a step based on headcount: one sum for the organisation, and how many managers are inside it does not change that sum.
Two weeks, no payment and no limit on people, so the decision comes from the work rather than from a website.
A monthly subscription per organisation: the step is chosen by how many people you have. Appoint as many managers as you need — it does not affect the bill.
Runs on your own server: local language model, confidentiality levels, access log, works with no internet. Annual licence.
We are revising the plans right now, so we quote the numbers on the call — together with the maths for your headcount and number of sites. The principle will not change: one price per organisation by step, with no charge per seat. The only separate line is the usage of an external language model above what the step includes; with the model inside your own perimeter there is nothing to pay for at all.
When data cannot leave the building
Deployed with a single command on your server. The language model and speech recognition run locally, and no assignment is sent to an outside service.
An assignment can be classified: the circle of people who see it narrows and delivery to external channels is blocked. Neither a project nor a job title overrides that.
Who opened a card and when. Entries are only ever added; neither the author nor an administrator can erase them.
In the cloud, company data is separated at every single query rather than by a visibility setting. There is a dedicated set of automated checks for it.
Web, mobile, email, Telegram. In a closed network the external ones are simply switched off, and the interface says so plainly.
Really. The assignment arrives in Telegram with four buttons: accept, propose a date, ask a question, mark it done. There is a web version for them too, but the rollout does not count on it and usually nobody opens it.
The card is shown to you before it goes out, and any field is one tap away from being fixed. If the model is unavailable, rule-based parsing takes over and the assignment still gets created.
It does not ask people to behave in a way they do not. Exactly one person creates cards and moves statuses: the one handing out the work. And it takes them half a minute.
One evening. Staff are uploaded as a file, everyone gets an email with a link, and Telegram is connected with one tap. We run the first weeks of the pilot with you and go through the results together.
It depends on where the model runs. In the self-hosted edition it runs on your hardware and nothing leaves. In the cloud edition ordinary assignments are parsed by an external model, and anything carrying a confidentiality mark is never shown to it — that is a check in the code, not a promise in a contract. And the assistant sees exactly what the person asking is allowed to see: it works with their rights, not its own.
The cloud version runs on servers in Russia. The self-hosted version never leaves your own network.
Russian and English, including the parsing of spoken assignments. Each person gets the interface and notifications in their own language, so a mixed team is not a problem.
Twenty minutes: you name three real assignments from your last meeting, we create them while you watch and show you exactly what your people will see. Then a pilot with your team.
A one-minute form: company, name, email. We reply to it and send an invitation. Prefer email — hello@worksync.ru.
