Service and support tickets with AI intake and customer portal

The problem: service requests arrive everywhere and in no order

When a customer calls about a broken laptop or a software fault, the request lands wherever the phone was answered — a notebook, a spreadsheet, a memory. By the time the technician arrives, half the context is missing. The customer does not know where the case stands, the manager does not know what is open, and the same issue gets reported twice because no one checked.

Service work needs structure from the moment the request arrives. Not after someone has time to type it up, and not only for the cases that feel important.

What the platform does instead

The ticket module captures every request in one place, with a timeline that starts the moment it is created. A customer can open a ticket by voice note, by typing into the portal, or by sending photos from a phone via QR code. The platform transcribes the voice note, analyses it with AI, and extracts the company, the employee and the issue. When recognition is confident, the ticket is created automatically. When it is not, the note waits for review.

The customer portal runs at support.ioncomputers.bg and gives every client company its own login. An employee there can see their own tickets, add follow-up notes or photos, and track status without a phone call. The portal adapts its colour theme per company.

Inside the platform, each ticket carries a timeline of events, attachments, checklists and documents. A technician can dictate findings on site, and the system will parse them. Monthly reports group tickets by customer, with numbering that resets each month and protocols that can be invoiced manually.

How intake works

When a voice note arrives, the platform transcribes it and sends the text to the AI with a structured prompt. The AI is asked to extract the customer name, the employee name, the asset being reported, and a short description of the issue. It replies in JSON.

If all four fields come back with high confidence, the ticket is created and the customer is notified. If any field is uncertain, the note goes into a review queue. A person reads the transcription, fills in what is missing, and approves. The ticket then appears in the system as if it had been confident from the start.

The same intake process applies to text typed into the portal. Photos go straight into the ticket as attachments, without analysis.

The customer portal and how it connects

The portal is a separate interface, not the main platform. A customer company administrator receives a login link by email. That administrator can then invite employees from the same company. Each employee sees only their own tickets.

When an employee creates a ticket, they can type a description, record a voice note, or scan a QR code to turn their phone into a camera. The QR code generates a one-time upload token. The phone opens a page where the camera works, the photo is sent, and the token expires. The ticket appears in the main platform immediately.

The portal does not allow editing of closed tickets, but follow-ups can be added to open ones. Status changes are visible in real time.

Monthly reports and numbering

At the end of each month, the platform generates a report per customer. The report lists every ticket opened that month, grouped by status. Each ticket carries a number that resets on the first of the month — so August tickets run from 001 upward, and September starts again at 001.

The report can be printed or exported. Invoicing is manual: the manager reviews the report, selects which tickets to bill, and creates the invoice outside the platform. The ticket system does not generate invoices on its own.

What the AI sees and what it does not

The AI receives the transcription and a prompt that defines the four fields. It does not see previous tickets, does not access the customer database, and does not decide priority. It extracts, it does not infer.

When the AI is uncertain, it says so in the JSON response. The review queue shows the transcription, the attempted extraction, and a form to correct it. The person doing the review is the one who decides whether the guess was close or wrong.

Integration with the rest of the platform

Tickets link to the customer record, so the "who is calling" popup can show open cases. The call log can attach a call recording to a ticket. The task system can create follow-up tasks from a ticket event.

The ticket module does not yet connect to supplier ordering or quoting. It is a separate stream, built for service and support, not for sales.

What comes next

The current release handles intake, timeline, portal access and monthly reports. The next step is automatic task creation from ticket events — so when a ticket is marked "awaiting parts", a task to order the part is created without manual entry. After that, integration with the phone system will allow a ticket to be opened directly from a call popup.

Frequently asked questions

Can a customer create a ticket without the portal?
Yes, by phone. The person taking the call can dictate a voice note or type the ticket directly into the platform. The portal is optional.

What happens if the AI gets the customer name wrong?
The ticket goes into the review queue. A person reads the transcription, corrects the name, and approves. The ticket is then created with the correct customer.

Can two customers have the same ticket number?
Yes. Ticket numbers reset each month and are scoped per customer. Customer A's ticket 015 in September is separate from customer B's ticket 015 in the same month.

Does the platform store call recordings inside tickets?
No. The recording stays in the phone system. The ticket stores a link to the recording and the transcription, but not the audio file itself.