Board and tickets
Photos, videos, voice notes, documents and transcripts stay attached to the original request.
Management platform · integrations · AI
Operations and AI content
A management platform built for a studio that designs and produces exhibition fit-outs. It connects site requests, operational management, quotes, documents and content for the website and social channels.
At a glance
- Problem
- Operational requests arrived as photos, voice notes, messages and documents, and had to be carried by hand into the office workflow. Online communication also meant preparing similar material several times for different destinations.
- What I built
- A system that connects Telegram, an operational board, administrative tools and a multichannel editorial area.
- My role
- I handled analysis, design, development, AI integration, release and maintenance.
- Distinctive element
- People working on site keep using a familiar channel. The system turns what arrives into structured work objects, and the editorial area uses AI to prepare content that stays editable before publication.
The problem
- Operational requests spread across messages, voice notes and photographs.
- Quotes and bills of quantities with items, measurements, quantities and printing.
- Documents, suppliers, purchases and jobs to connect to the work.
- Content to prepare several times over for website and social.
How it works
I started by watching how requests were created and which tools were already in use. From there I split the system into three connected surfaces. Telegram collects input from the field, the internal area organises the work and the editorial area prepares content.
Tasks and calendar
Tasks carry owners, deadlines, priorities and work areas.
Quotes and bills of quantities
Offers are organised into items and chapters with measurements, quantities, numbering and PDF output.
CRM and administration
Documents, purchases, sales, suppliers and invoices are linked to projects.
Document signing
Documents can be sent to a dedicated page through a personal link.
Multichannel editorial area
A piece of content can be prepared for journal, portfolio and social channels with dedicated versions and previews.
From the field to a task
The most important path is also the shortest. People working on site keep using the same channel they would use for any other message.
- Message
A person on site sends a photo, video, voice note, document or dictated text to the studio bot.
- Transcription
Voice notes are converted into text and the original material remains attached to the request.
- Ticket
The office sees the request appear on the board with the received material and can review it before turning it into operational work.
- Task
The ticket can become an assigned task with a deadline and priority.
The bot sends a readable confirmation back to the sender so they know the request has reached the office.
From editorial work to the channels
Content preparation follows four steps, with AI preparing the channel versions and the person keeping the final decision.
- Writing
A person prepares the source content, chooses the intended tone and attaches the images.
- Destinations
The channels are selected from the configured website and social destinations.
- Preparation and previews
AI can adapt the text and format for each channel. Every version remains editable and is shown through a preview that reflects the publication context.
- Publishing
Publication only happens after an explicit human confirmation.
UX decisions
Telegram stays the entry point from the field
People working on site can send material through a channel they already use instead of filling in a new interface.
The AI does not publish on its own
Prepared or adapted content stays editable and needs a final confirmation from a person.
Results and product status
- The system is in service and keeps evolving module by module.
- Field requests flow into the board.
- Quotes and bills of quantities are produced inside the platform.
- The editorial area handles content for website and social with previews and human approval.
The images come from the real system and are published only after names, references and identifying data have been obscured.