Requests and approvals
A bot or portal takes the request, checks the fields, moves it through the approval route and keeps the status. Managers approve in a messenger.
example: expense reports — a photo of the receipt becomes a finished table
business process automation
We remove manual work from your processes
Bots in messengers, client portals, integrations with 1C and ERP, reports that build themselves. We start with one task and take it to a working service.
1C, ERP — the accounting and resource-planning systems your finance and production teams already work in.
before
Requests in email
Excel per plant
Messenger threads
Paper forms
1C exports
after
One intake point
Rule-based approval
Written to 1C and ERP
Report on schedule
process diagram: five manual inputs → one flow
02 / services
What we automate
Six tasks we are asked for most often. Any of them can be taken on its own.
Reporting
We pull data from dozens of files and systems, normalise it and build the summary. The report arrives on schedule, not on request.
example: plant summary from Excel files sent by email
Documents
Contracts, quotes, print forms and labels generated from one set of data. Numbering, templates and the archive stay in the service.
example: contract and invoice built from the request
Client portals
Roles, requests, documents, mailings. Clients and contractors work in a portal instead of a mailbox, and each of them sees what to do next.
example: contractor portal with requests, acts and payment status
Integrations
1C, ERP, CRM, payment services, payouts to contractors, corporate mail and calendars, marketplace and supplier APIs. Data moves without manual exports.
example: receipts and payouts for closed requests
Data collection and checks
Supplier price and assortment monitoring, quote comparison, checks for mismatches between systems. Differences show up the same day.
example: daily digest of supplier price changes
03 / cases
What already runs
Industry and scale instead of names: the companies have not agreed to be named.
Tickets from Yandex Messenger into SimpleOne
- task
- “People write in a messenger while tickets live in SimpleOne. Requests get lost between the chat and the system, and agents answer in two windows.”
- done
- Yandex Messenger integration with SimpleOne: a chat thread creates a ticket, agent replies return to the same chat, statuses and attachments stay in sync. Packaged as an installable platform solution.
- result
- one chat instead of two windows; package 1.0 installs into SimpleOne without a developer; reusable across the platform's customers.
Portal and site instead of email and Excel
- task
- “Requests arrive by email, assignments go out in threads, documents and points are tracked in Excel. Who is at which stage — we find out by phone.”
- done
- Website and client portal: requests, contractor assignments, documents, award letters, point balances, mailings, a text editor for non-technical staff, roles and permissions, plus a separate staging site for acceptance.
- result
- 44 review items worked through and closed; one portal instead of email, Excel and phone calls; own server data and access stay with the client.
Expense reports in a messenger, not a spreadsheet
- task
- “Expense reports are assembled by hand, receipts get lost, and the report reaches finance and the ERP late and with errors.”
- done
- A bot in Yandex Messenger: amounts, receipt uploads, meetings, completeness checks. Export to the ERP and submission for e-signature, with statuses returned to the employee in the chat.
- result
- in a chat a report takes minutes; root cause found: expense and VAT fields vanished in the ERP; paperless signing goes through the HR e-signature system.
Three Excel requests into three plants' plans
- task
- “Three request files, three plants, each with its own plan form. We copy by hand, get lost in rail cars and mixed loads, and formulas break.”
- done
- Request parsing, allocation by plant, splitting rail cars into trucks, mixed loads across days, filling each plant's own form exactly. A one-click desktop app.
- result
- 43 tests check the allocation on every run; Mac and Windows one-click run, no Python setup; formulas intact the paste skips service columns.
Who actually works in mail and cloud storage
- task
- “We need to see who works in corporate mail and storage, by department and by month. Asking managers gives impressions, not facts.”
- done
- Audit-log collection from Yandex 360 via API, event breakdown by person and department, monthly reports showing spikes and silence.
- result
- 234 people in one report, no manual exports; 7 reports monthly, repeatable with one command; facts instead of impressions.
Forum venue: models, plans and dimensions from one dataset
- task
- “A contractor abroad needs plans, elevations and models from one agreed geometry. Every client change means recalculating the whole set.”
- done
- Parametric models of rooms and structures, seating layouts, dimensioned plans and elevations, a handover package for the contractor, with built-in geometry checks.
- result
- 41 files of models in the handover package; 13 rooms built from the hotel's own spec; checks overlaps, aisles, clearances.
Where the basket is cheaper: five chains compared
- task
- “We need to compare the whole basket, not single items: where it is cheaper to buy regularly. By hand that is a week of work and still inaccurate.”
- done
- Price collection from five chains for a 31-item basket, conversion to unit prices, matching of comparable products, weekly and monthly basket totals, plus a no-loyalty-card figure.
- result
- 31 items checked across five chains in one run; unit price per kilo, litre and piece; with and without card the cost of the discount is visible.
we also do
Other things we have shipped
- local data anonymisation before AI chats: a browser extension plus an on-device service
- office and parking booking bot — bookings mirrored to a spreadsheet on the company drive
- digest bot for a salon chain: subscriptions, payments via YooKassa, notifications
- inbound email triage: message → analysis → a note in the team chat with a category
- payouts to self-employed contractors and tax receipts: integration with the operator service
- supply contracts: client details from cards → ready documents with numbering
- labels and stickers: 76 catalogue items → print-ready PDF
- quotes and cost estimates: a quote generator, side-by-side comparison of contractor bids
- websites and landing pages: a 30-project portfolio, product and service pages
04 / process
How we work
Four steps. Each one leaves you something you can look at.
012–5 days
Discovery
We look at the process as it is, count volumes and mark the manual steps. You get a process diagram, an integration list and an estimate.
022–3 weeks
Prototype
A working slice on your own data: the bot takes a request, the report is calculated. You see the result before the main spend.
03from 1 month
Rollout
We finish the service, connect 1C and the other systems, deploy on your servers and train the staff.
04subscription
Support
We watch the service, make changes and send a monthly report: what ran, what we fixed.
You can start with one task — one report or one bot. No end-to-end programme is needed to get a result.
05 / stack
Stack and approach
Python and FastAPI for services, Node where it fits, PostgreSQL, Docker. Bots on aiogram. Integrations over APIs. Deployed on our servers or yours — whichever suits you.
data and access
Code and data stay with you. We take the minimum access and only while the work runs, then hand everything over with documentation.
06 / pricing
Price guidance
The exact figure comes after discovery and is fixed in the estimate.
Small automation
One report, one export, one calculation. Usually one to two weeks.
A bot or a client portal
Request intake and approvals, roles, documents, notifications, one or two integrations.
Support
Monitoring, changes as needed, a monthly report on how the service ran.
We name the price after discovery and fix it in the estimate: it depends on the scope of work, not on the type of solution.
07 / questions
Questions and answers
Discovery takes a few days, the prototype 2–3 weeks, rollout depends on scope. The exact term is named after discovery, before the work starts.
Someone who knows the process, access to test data and one approver. A written specification is not required.
That is the usual case. We go through the process together, write it up ourselves and get it approved before any code.
You do. Source and data stay with you, deployed on your servers, with minimum access granted.
A monthly subscription: changes, monitoring, a service report. One-off work is also possible.
From the scope after discovery. The figure and the scope are fixed in the estimate; later changes are agreed separately.
08 / request
Describe your task
Two lines are enough: what is done by hand and how long it takes. We will reply and suggest where to start.
Request sent
We reply during working hours, usually the same day. If it is faster, write to us on Telegram.
direct
We reply during working hours, usually the same day. Discovery starts with a single 30-minute call.