Foremen check stock, budgets and payments in 1C right from the site
Foremen call the office many times a day: "when is the tile arriving?", "has the subcontractor been paid?", "what's left in the budget?" A Telegram bot answers these questions itself: it pulls the data from 1C UNF, the company's accounting system, and shows it in two taps. Foremen confirm deliveries with a photo of the delivery note.
14 sites at once and one office phone
For the estimate, take a company that fits out commercial premises: restaurants, showrooms, small offices. It runs 14 sites at once, each with a foreman and 2–4 crews of different trades: electrical, finishing, plumbing. Accounting is kept in 1C UNF on the company's own server. Materials go from the warehouse to the sites, and subcontractors are paid on a schedule.
A foreman spends the day on site with a phone in one pocket and a hammer drill in hand. Opening 1C on a laptop there isn't realistic, so he calls the office. The estimate assumes 6–8 calls a day from each foreman about deliveries, subcontractor payments and what's left in the budget.
Fourteen foremen make about a hundred calls a day. If each call takes 2.5–3 minutes, including looking up the answer in 1C, the office manager spends 4–5 hours a day on the phone and fits the rest of the job in between calls. Deliveries get entered into the warehouse records when someone gets round to it, sometimes 1–3 days later. Until then other foremen can't see the real stock and call again.
A Telegram bot talks to 1C UNF through its HTTP services and WireGuard
The foreman opens the bot and in two taps gets what used to take a phone call. The bot is built on aiogram 3, a Python framework for Telegram bots, and lets in only phone numbers from an approved list. A service in the middle, written in Node.js and Express, talks to 1C UNF through 1C's own HTTP services. All traffic between the rented server and the company's 1C server runs through an encrypted WireGuard tunnel.
The foreman sends a command or taps a button in a message. Each section has its own dialogue flow in aiogram 3.
An approved list of phone numbers, linked to the employee's record in 1C. No password.
Node.js and Express on a rented server. It keeps frequent answers for 5 minutes, queues requests and retries them on failure.
An encrypted tunnel between the rented server and the company's 1C server. 1C can't be reached from outside.
1C's HTTP services return data on request: stock, movements, postings. Deliveries are recorded with a photo of the delivery note attached.
Buttons that are easy to tap with one hand
The main user is a foreman on site. So the bot works through buttons inside the message, with at most two taps to any answer. Main menu: "My sites" → choose a site → "Orders / Payments / Budget / Confirm receipt". There's almost nothing to type.
Confirming a delivery with a photo
When the truck arrives at the site, the foreman photographs the delivery note and taps "Confirm receipt" on the order card. The bot takes the photo, stores it in MinIO file storage linked to the 1C document, and records the material as delivered to the site in UNF. Stock in 1C updates at that moment. Other foremen see the new figures in the bot within 5 minutes, which is how long the middle service keeps frequent answers.
The bot reports important events itself
The foreman doesn't need to open the bot to find out what's new. "The tile truck has left the warehouse, arriving at 14:30." "Subcontractor Ivanov has been paid, the work certificate can be signed." "Site Z is 3% over budget." These messages are triggered by events in 1C UNF and by scheduled checks.
How access to 1C is protected
1C UNF stays on the company's own server and isn't exposed to the internet. The WireGuard tunnel between the rented server and the 1C server encrypts all traffic. Only numbers on the approved list can sign in, and a site is visible only to the employees added to its record in 1C. The middle service checks access rights again and won't show a foreman someone else's site.
What the solution is built from
The single source of data: sites, budgets, materials, subcontractors, payments
A REST interface published inside 1C itself, with no third-party connectors
A Python framework for Telegram bots: step-by-step dialogues, buttons in messages, handling taps
The middle service: sign-in, cache, queue, retries, access rights, requests to 1C
An encrypted tunnel between the rented server and the company's 1C; 1C isn't exposed to the internet
A rented server for the bot and the middle service, for example at Selectel in Russia. The company can pick the host
S3-compatible storage for photos of delivery notes and signed work certificates
Delivers messages and notifications to foremen
Parts of this setup already run in my own projects. The knowledge-base assistant is a Telegram bot with three access levels. My selective traffic routing runs over a WireGuard tunnel. And I worked with 1C's exchange formats when I built the Treasury statement converter, which produces files for import into 1C.
What to expect after launch
The figures are for the same company with 14 sites. The estimate assumes the bot answers 19 out of 20 questions foremen ask: stock, payments, budget, delivery times. For your company I'll redo the numbers from your actual sites and calls.
only unusual questions still need a call
4–4.8 hours a day; on a ₽80k monthly salary that's ₽40–48k a month
the entry appears as soon as the foreman confirms receipt with a photo of the delivery note
Where these numbers come from. Calls now take 4–5 hours a day. If each foreman is left with 0.4 calls a day, fourteen foremen make 5–6 a day, which is under 20 minutes. That frees 4–4.8 hours of an eight-hour day, or 0.5–0.6 of a full-time job.
There's one more gain that's hard to put a number on: the foreman always has the budget balance in front of him. Today that number comes from the office manager. With the bot, the foreman sees it straight from 1C and can catch an overspend while the decision is still open.
How long it takes and when it pays back
from the start of work to launch. The exact time depends on whether your 1C already publishes HTTP services
one-off, plus ₽12k a month for support. The exact figure depends on the number of sites and bot sections
By this estimate the office manager's freed-up time is worth ₽40–48k a month, or ₽28–36k after support. Leaving support aside, the investment pays back in 10–12 months; including support, in roughly 13–17.
The same approach can be used with other 1C configurations: Trade Management, 1C:ERP and Integrated Automation (KA). Any configuration will do if it has HTTP services or can publish them. Prices for other work are on the Pricing page.
Where the same approach fits
The typical situation: an employee in the field has no access to the accounting system and has to call for every number. With the bot they handle simple 1C operations right on the phone, without a full 1C client. This comes up in many industries:
- → Logistics and delivery: a driver marks "cargo received" and "shipped" in the bot and shares where the truck is
- → Manufacturing: a shop floor supervisor checks the shift plan, marks work as done and records defects with a photo
- → Service and cleaning companies: a technician on site records the work done and the parts used
- → Sales reps: an order placed in the bot goes straight into 1C, and nobody has to forward it to a manager
- → Retail: a shop assistant sees the stock and orders goods for the store without calling the merchandiser
- → Field staff in any company where 1C runs on its own server and nothing is exposed to the internet
- aiogram 3, Express and 1C HTTP services together fit any task where people away from the office need numbers from 1C
- A WireGuard tunnel between the company's server and a rented server: 1C isn't exposed, and all traffic is encrypted
- Sign-in by an approved list of phone numbers, no passwords needed
- Confirming a delivery with a button and a photo of the document: records are kept right on site
Blog article on this topic: A Telegram bot as a 1C client: how it works, where it fails and what to check first
Related cases
Converter for Federal Treasury statements into 1C
Reads XML statements from the Federal Treasury's GIS system in both format versions, V3 and V4, and builds a…
Forecasting events with several language models: calibration and Brier scoring
Five AI agents search open sources, a sixth reconciles their answers, and a calibration step adjusts the…
Language models analyze trading signals right on a laptop: Ollama and RAG on an M3
qwen2.5:14b is the main model, llama3.1:8b the backup, and LLaVA:7b reads chart screenshots; ChromaDB finds…
Start with the 5,000 ₽ audit: a written report and a payback estimate
I'll tell you what to automate first, when it will pay for itself and whether your task needs AI at all. Sometimes it doesn't.
Or just ask a question. During business hours I reply within 2 hours