Find your process among nine systems
Each system does one job. It starts with a Process Snapshot, gets built in two to six weeks and stays under monitoring. Each one says whether it has already run for a client.
Lead Intake
No inquiry goes without a reply and a record.
Execution trace
Lead Intake · ZippyShell Louisiana (US)
25 Mar 2026
-
Runs on: Gmail An inquiry arrives by email name · phone · email · type · date Outcome: received
An inquiry arrives by email Field Type name text phone phone number email email address type moving or storage date date Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
-
Runs on: n8n Reading the text into fields Outcome: fields filled
Reading the text into fields Field Type name text phone phone number email email address type moving or storage date date Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
-
Runs on: n8n Duplicate check by phone and email already exists: skipped and logged Outcome: 2 of 41 inquiries, 25 Jul to 2 Sep 2026 Outcome: new, continues
Duplicate check by phone and email Field Type phone match key email match key Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
-
Runs on: Google Sheets Row written to this month's sheet Outcome: row written
Row written to this month's sheet Field Type sheet month received row one inquiry Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
-
Runs on: Dialpad Text to the customer, by inquiry type Outcome: sent
Text to the customer, by inquiry type Field Type phone phone number message by type: moving or storage Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
-
Runs on: Gmail Email to the customer, for moving or storage Outcome: sent
Email to the customer, for moving or storage Field Type email email address template moving or storage Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
-
Runs on: n8n If sending fails Outcome: error alert sent
If sending fails Field Type step [POTREBNO POJAŠNJENJE: koji korak je pao, sadržaj obaveštenja] Field names come from the flow description. Exact schema: [POTREBNO POJAŠNJENJE: tačna imena polja u ZippyShell toku]
- Who needs it
- a company whose inquiries arrive by email, web form or text, and whose reply depends on who happens to be free.
- Inputs
- email, website form, text message.
- Outputs
- a record without duplicates; a personalized reply by email and text; an error alert when sending fails.
- Timeline
- 2 to 3 weeks.
- Price
- roughly 2,500 to 3,500 EUR.
- Tools
- n8n, Gmail, Google Sheets or a database, Dialpad or Twilio.
- Proof
- ZippyShell Louisiana (US), engaged through Izinga. Live since 25 March 2026. More than 200 emails sent by 2 September 2026. Of 41 inquiries between 25 July and 2 September 2026, the system caught two duplicates. From 27 August to 1 September 2026, 5 of 5 runs succeeded.
Work Orders & Records
A work order is entered once and never retyped.
Execution trace
Work Orders & Records · Suzuki Servis Musa Racing, Belgrade
[POTREBNO POJAŠNJENJE: datum prenosa]
- Runs on: Excel (XLSM) Reading the old file with macros Outcome: 1,865 rows
- Runs on: Supabase Vehicles, duplicate check same vehicle entered more than once: merged Outcome: 69 duplicates Outcome: 1,079 to 1,010
- Runs on: Supabase Work orders, linked to vehicle and customer Outcome: 1,848 work orders
- Runs on: AutoServis OS The shop runs on the new software Outcome: 947 customers
- Who needs it
- a service or technical company that keeps work orders in Excel, on paper or in one person's head.
- Inputs
- an old Excel or XLSM file, new work orders.
- Outputs
- a database, roles (who sees and edits what), reminders, a view of the work.
- Timeline
- 4 to 6 weeks.
- Price
- roughly 4,000 to 6,000 EUR.
- Tools
- Supabase, Next.js, migration of existing data.
- Proof
- Suzuki Servis Musa Racing, Belgrade. 1,865 rows moved. Of 1,079 vehicles, 69 were duplicates, leaving 1,010. The database holds 1,848 work orders. The shop now runs on AutoServis OS, with 947 customers.
Quote & Configurator
The customer builds the request. The quote starts without retyping.
Execution trace
Quote & Configurator · Euro-Karton
live on eurokarton.rs
- Runs on: website The customer picks the box type and dimensions Outcome: customer's choice
- Runs on: website Every choice the customer made, in one place Outcome: inquiry assembled
- Runs on: email The inquiry goes straight to two people at the company Outcome: no retyping
- Who needs it
- a manufacturer or seller whose customer has to specify size, quantity or finish before a quote.
- Inputs
- the customer's choices on the website.
- Outputs
- a structured inquiry by email, a basis for the quote, a PDF if needed.
- Timeline
- 3 to 4 weeks.
- Price
- roughly 2,500 to 4,500 EUR.
- Tools
- web, email, PDF.
- Proof
- Euro-Karton. The box configurator is live on eurokarton.rs. The customer picks the box and dimensions, and the inquiry goes straight to the inboxes of two people at the company, with no retyping. You can try it yourself.
Owner Report
One screen instead of a report built at night.
Execution trace
Owner Report
partial proof: analytics inside AutoServis OS
- Runs on: Supabase Data from systems you already have Outcome: collected
- Runs on: Next.js A weekly overview on one screen Outcome: overview
- Runs on: email An alert when a number jumps Outcome: alert
- Runs on: n8n If the data doesn't arrive Outcome: error alert sent
- Who needs it
- an owner who pulls numbers from three spreadsheets before every meeting and trusts none of them.
- Inputs
- data from systems you already have.
- Outputs
- a weekly overview and alerts when a number jumps, on one screen.
- Timeline
- 2 to 4 weeks.
- Price
- roughly 2,500 to 4,000 EUR.
- Tools
- Supabase, Next.js, email.
- Proof
- partial. Analytics run inside AutoServis OS. As a standalone system, there is no proof yet.
Systems we build that haven't run for a client yet
We say so openly. For these, the Process Snapshot also checks whether they're feasible for you, before any larger commitment.
Company Knowledge Base
Answers from your company's own documents, with a source on every answer.
Execution trace
Company Knowledge Base
planned flow, no client proof yet
- Runs on: Supabase Documents, emails and notes go into the base Outcome: stored
- Runs on: pgvector Search across the company's documents Outcome: passages found
- Runs on: Claude An answer with a source on every answer Outcome: answer and source
- Runs on: MCP Your company's AI assistant sees the base Outcome: connected
- Runs on: n8n If storing or search fails Outcome: error alert sent
- Who needs it
- a company where only one person knows the answer, or it sits in an email from two years ago.
- Inputs
- documents, emails, notes.
- Outputs
- search, answers with sources, a connection that lets your company's AI assistant see the base.
- Timeline
- 3 to 5 weeks.
- Price
- roughly 3,000 to 5,000 EUR.
- Tools
- Supabase, pgvector, Claude, MCP.
- Proof
- no client proof yet. The same approach runs in Marrow, a personal notes system Amar built for himself, and in our internal CRM with its own MCP server.
Document Reader
Data from an invoice, contract or bank statement enters the system without typing.
Execution trace
Document Reader
planned flow, no client proof yet
- Runs on: scan or photo Invoice, contract or bank statement Outcome: received
- Runs on: Claude or GPT Reading fields from the image Outcome: fields proposed
- Runs on: Azure Document Intelligence Tables from the document Outcome: rows proposed
- Runs on: a person Check before saving Outcome: approved
- Runs on: Supabase Saved to the database Outcome: saved
- Runs on: n8n If reading fails Outcome: error alert sent
- Who needs it
- a company where someone retypes documents from paper or PDF.
- Inputs
- a scan or photo of an invoice, contract or bank statement.
- Outputs
- fields in the database, checked by a person before saving.
- Timeline
- 3 to 5 weeks.
- Price
- roughly 3,000 to 5,000 EUR.
- Tools
- Claude or GPT for image reading, Azure Document Intelligence for tables, Supabase.
- Proof
- none yet. Before production, accuracy is measured on 50 of your real documents.
Voice Receptionist
A call gets answered even when nobody can pick up.
Execution trace
Voice Receptionist
planned flow, no client proof yet
- Runs on: Twilio Incoming call Outcome: received
- Runs on: Retell or Vapi A reply in Serbian Outcome: conversation
- Runs on: n8n Appointment booked Outcome: logged
- Runs on: Twilio Handoff to a person when needed Outcome: transferred
- Runs on: n8n Call record Outcome: logged
- Runs on: n8n If a call goes unanswered Outcome: error alert sent
- Who needs it
- a company that misses calls during work, after hours or when everyone is out on site.
- Inputs
- an incoming call.
- Outputs
- a reply in Serbian, a booked appointment, a handoff to a person when needed, a call record.
- Timeline
- 4 to 6 weeks.
- Price
- roughly 4,000 to 6,000 EUR, plus a monthly cost for voice and telephony platforms (estimated 150 to 400 EUR), passed on at actual cost.
- Tools
- Retell or Vapi, Deepgram, ElevenLabs or Azure voice, Twilio, n8n.
- Proof
- none yet. Before any proposal we run a 20-call test in Serbian, so you can hear how it sounds.
WhatsApp Intake
A customer's WhatsApp message becomes an order or inquiry in the system.
Execution trace
WhatsApp Intake
planned flow, no client proof yet
- Runs on: WhatsApp Business API The customer's message Outcome: received
- Runs on: Claude Message into a structured order or inquiry Outcome: fields filled
- Runs on: WhatsApp Business API Confirmation to the customer Outcome: sent
- Runs on: Supabase Order saved Outcome: saved
- Runs on: n8n Handoff to a person Outcome: forwarded
- Runs on: n8n If confirmation or saving fails Outcome: error alert sent
- Who needs it
- a company whose customers write on WhatsApp, where orders get retyped by hand or lost in the chat.
- Inputs
- the customer's message.
- Outputs
- a structured order or inquiry, a confirmation to the customer, a record, a handoff to a person.
- Timeline
- 3 to 5 weeks.
- Price
- roughly 2,500 to 4,500 EUR, plus Meta's per-message charge.
- Tools
- WhatsApp Business API (through Infobip or Twilio), n8n, Claude, Supabase.
- Proof
- none yet.
SEF Sync
An invoice from your software or spreadsheet reaches SEF, Serbia's e-invoicing system, without manual entry.
Execution trace
SEF Sync
planned flow, no client proof yet
- Runs on: software or spreadsheet Invoice Outcome: ready
- Runs on: n8n Submission to SEF through its API Outcome: sent
- Runs on: SEF API Status tracking Outcome: status logged
- Runs on: Supabase Archive Outcome: stored
- Runs on: n8n If SEF rejects the invoice Outcome: error alert sent
- Who needs it
- a company that enters invoices into SEF by hand or retypes them from a spreadsheet.
- Inputs
- an invoice from your software or a spreadsheet.
- Outputs
- submission to SEF through its API, status tracking, an archive.
- Timeline
- 3 to 5 weeks.
- Price
- roughly 3,000 to 6,000 EUR (Compliance line).
- Tools
- n8n, the SEF API, Supabase. Your company's SEF administrator generates the API key.
- Proof
- none yet.
Don't see your process?
Describe it anyway. If it doesn't fit any of the nine systems, we'll say so in our reply, and explain why.
Every system is bought in three steps
- Process Snapshot.
400 to 700 EUR, 5 to 7 working days. A process map, the risks, a phased plan with prices and a baseline measurement. The full amount is credited against the system price.
- One Process, One System.
2,500 to 6,000 EUR, 3 to 6 weeks. Including data migration, training, documentation and 30 days of monitoring.
- Monitoring.
150 to 400 EUR per month, 6 months minimum. Failure tracking, small changes, a monthly report.
A company that's losing inquiries can start smaller: Single Automation, 600 to 900 EUR, 5 to 10 working days.
All packages and what moves the priceTechnologies we work with
We choose tools to fit your process and what your company already uses. We list them so you know what we work with.
Running for clients or in our own production
- Automation
- n8n.
- AI models
- OpenAI GPT, Anthropic Claude, Google Gemini, with structured output and function calling.
- Connecting AI assistants to data
- MCP (Model Context Protocol).
- Speech to text
- Groq Whisper, OpenAI transcription, in Serbian.
- Chip card reader
- vehicle registration documents, in AutoServis OS.
- Data
- Supabase and Postgres, pgvector, Google Sheets, migration from Excel, XLSM and VBA.
- Applications
- Next.js, React, TypeScript.
- Email and messaging
- Gmail API, IMAP, Resend, Dialpad SMS.
- Workplace tools
- Google Workspace.
- CRM
- our own CRM on Supabase (internal).
- Hosting
- our own server with Caddy, Vercel.
- Monitoring and testing
- error alerts in every flow, nightly backups, Playwright.
We also work with these, not yet delivered for a client
- Automation
- Make, Zapier, Microsoft Power Automate.
- Open AI models on your own server
- Llama, Mistral, for data that can't leave the company.
- Voice and telephony
- Retell, Vapi, ElevenLabs, Deepgram, Azure Speech, Twilio.
- Documents
- Claude and GPT for image reading, Azure Document Intelligence, Google Document AI.
- Data
- Airtable, Access.
- Messaging
- WhatsApp Business API, Infobip, Viber, Twilio SMS.
- Workplace tools
- Microsoft 365, Copilot.
- CRM
- HubSpot, Pipedrive.
- Serbia
- SEF API, eUprava, e-Otpremnica.
- Monitoring
- Sentry.
One process is enough to start
Tell us which one and where it lives today. The description commits you to nothing. A written reply with a price range arrives within 48 hours.