Jamz Flow

Jamz Flow · AI automation team

We build systems that take repetitive work off your team's desk.

AI workflows, internal tools and agents for teams that want their time, energy and focus back. We take each one from the first process map to a system that runs on its own.

See our workBook a call

How we build

Automation should make people's work lighter, not just cheaper. Every system we deliver follows these rules.

Facts come from code

Dates, reply times and totals are calculated, not guessed. AI handles the parts that need judgment, and code checks its answers wherever the right answer can be known.

No failure stays silent

Limits, timeouts and odd outputs are caught, retried or handed to a person with the details. When something breaks, someone hears about it.

People keep the decisions

We automate the repetitive steps and leave judgment where it belongs. Manual corrections are never overwritten by the system.

Your data stays yours

Read-only access where possible, personal details masked before any text reaches a model, and results stored in your own accounts.

Selected work

Systems that are built and running, or in testing where marked. Client names stay private; the problems and the solutions are real.

AI content pipeline with a quality gate

For a US structural engineering firm

  • Client pilot
  • Google Apps Script
  • Gemini API
  • Google Sheets
The problem

The firm needed steady, expert blog content. Every technical claim would appear under a licensed engineer's name, so getting it wrong was not an option.

What we built

A pipeline that plans twelve weeks of SEO topics against the site's existing posts and competitors, drafts each article, and runs it through two review layers: code checks for broken or invented links and formatting, then an independent AI editor that must score 80 out of 100. Claims an engineer has to confirm are flagged, and the flags are capped, so the model drops claims it cannot support.

In testing, the gate stopped a draft that confused a collar tie with a rafter tie, and another that made up inspection prices.

Status

Delivered in September 2026. The client reviews and approves each article; publishing stays a five-minute manual step because the site's CMS has no content API.

Inquiry tracking and lead dashboard

For the same engineering firm

  • Client system
  • Gmail API
  • Apps Script
  • Gemini API
The problem

Every inquiry landed in one shared inbox. Nothing was recorded, so no one knew where leads came from, how fast they got a reply, or why most never became projects.

What we built

A read-only system that separates real inquiries from vendor pitches and spam, records each one, and keeps its stage current as the conversation moves. A dashboard shows sources, win rate, reply speed, lost reasons, and who is still waiting for a follow-up. It checks the inbox every hour.

Status

Deployed in September 2026 for inquiries from October onward.

Meeting scheduling assistant

  • Automation tool
  • Google Calendar
  • Google Meet
What we built

Shows open times in the other person's own time zone, books the agreed slot in Google Calendar, creates a Google Meet link and sends the details to both people.

Content publishing automations

  • Automation tool
  • Telegram
  • YouTube
What we built

A news radar that watches trusted sources and publishes matching posts to a channel on its own, and a bot that republishes new YouTube videos the moment they go live.

Private video calls on our own website

Meeting rooms built into this site, no app or account needed

  • Live on this site
  • WebRTC
  • On-device AI
The problem

Calls with clients depended on outside apps: accounts to create, links that expire, paid plans, and calls that struggled on slow connections. We wanted meetings to happen right next to our work and our proposals.

What we built

A meeting room inside our site. We create a room, send one link, and the other person joins from any browser. Video goes directly between the two browsers, and when a network does not allow that, a relay on our own server carries it.

Up to four people, screen sharing, chat, and background blur in three levels that runs on the device itself, so the picture never leaves it. Everyone appears in their own orientation and is never cropped, and each person can move and resize their own small picture.

No per-minute fees: when the two browsers connect directly, the call uses no server traffic at all.

Status

Live on this site and ready for client calls. From idea to launch took one day.

AI front desk for clinics

Talks to patients on the clinic's website or by phone, by voice or chat, and books visits.

  • In testing
  • Voice and chat
  • Phone line
  • Management panel
The problem

Patients call and nobody answers. The line is busy, it's after hours, or the receptionist is busy answering the same questions again and again. Some bookings also get lost on paper.

What we built

A receptionist that talks to patients on the clinic's website, by voice or chat.

  • It offers the first free time.
  • It takes the name and number and repeats them back to be sure.
  • It books the visit so two people never get the same time.
  • If the voice drops, the conversation continues in chat.

It can also answer the clinic's phone line: callers dial a normal number, and the same receptionist picks up.

The clinic sees everything in one panel: today's bookings, the list of people who got in touch, and the text of every conversation.

It never makes up answers. If it doesn't know something, it sends the question to the clinic. The clinic answers once, and from then on the receptionist answers by itself.

Status

In testing with a sample clinic; the phone line connection is built and in testing too. Open to pilot partners.

AI sales follow-up agent

Finds which customers need a call, talks to them, takes the order and updates the CRM.

  • In testing
  • Voice and text
  • Sales dashboard
  • CRM
The problem

Distributors lose sales quietly. A sales rep visits but no order comes in, or a good customer starts buying less. The call team can't reach everyone, and the reasons customers give rarely reach the CRM.

What we built

An agent that runs the follow-up calls end to end.

  • Every day it ranks customers by value and urgency, with a plain reason for each.
  • It talks naturally, asks why they haven't ordered and answers with a real offer.
  • It takes the order, repeats it with the total and confirms it.
  • It books the next call and passes problems to the right person.
  • It checks a sample of sales visits with the customer, so the manager knows which visits really happened.

The sales manager gets a dashboard: revenue recovered, reach rate, conversion, revenue at risk and the reasons customers give.

It never makes up prices or promises. Every offer comes from the company's settings, and anything it isn't sure about goes to a person.

Status

In testing with sample data; the full flow works end to end. Open to collaboration with distributors for a pilot.

AI video editor that runs on your own computer

  • In progress
What we built

A local video editing app that uses AI to cut editing time, with no footage leaving the machine. It grew out of years of producing video content for brands.

Tools we work withGoogle Apps ScriptClaude APIGemini APIOpenAI APIGoogle Workspace APIsWebRTCSupabasen8nMake

A small automation advisor

What could run on its own in your work?

Answer five short questions. In under a minute you get a few concrete ideas for systems that could free up your team's time.

Which task eats your team's week?

Tell us about it by email or book a short call. We'll reply with how we would approach it, and whether automation is the right answer at all.

Book a callSend an email