01 About

Started in 2016 joining systems together. Still the same job, better tools.

Mergical is run by Rezz, which is short for Reza Davaee. It began in 2016 doing systems integration for Australian businesses: making the quoting tool talk to the accounting system, getting the e-commerce orders into the warehouse, stopping the same customer details being typed into three places.

That work never went away. What changed is that a large part of it used to be impossible. If a task needed someone to read a document and understand it, no integration could help, and it stayed manual. A person opened the PDF, worked out what it was, and typed it in.

Around the early 2020s that stopped being true. The tools got good enough to read a supplier invoice, understand a customer enquiry, and know when they were unsure. Mergical moved into AI automation because it finally made the original job possible to finish, not because AI became something to sell.

That order matters. A lot of businesses are being sold AI by people who have never watched how the business actually works. The unglamorous part, sitting with the person doing the job and following it end to end, is the part that decides whether any of it is worth doing. It is also the part most vendors skip, because it is slow and does not demo well.

Who Mergical works with

Australian small and medium businesses, typically 3 to 50 staff. Trades and services, professional services, wholesale and distribution, e-commerce, and local service businesses.

The common thread is an owner who still knows every part of the business, and a handful of jobs that eat hours every week without needing anybody's judgement.

Who Mergical is not for

Enterprises with an in-house platform team. They can build this themselves, usually cheaper and better suited to their environment, and we would just be in the way.

Anyone who wants AI because it is AI. There has to be a job it does and a number it moves, or there is no point.

02 Principles

Six things we actually work by.

These are not values on a wall. Each one changes what gets built and what gets refused.

Start small, prove it, then expand

One workflow first. It runs alongside the manual process until it has earned trust. Businesses that automate five things at once end up trusting none of them, and quietly go back to doing it by hand.

Automate the repetition, not the judgement

The parts of a job that are the same every time are worth automating. The parts that need somebody who knows the customer, the site or the trade are not. Most of the skill is telling the two apart, and being disciplined about the line once you have found it.

The client owns everything

Accounts in your name on your billing, the configuration, and the documentation. If you stop working with us tomorrow, everything keeps running and somebody else can pick it up by reading what we wrote down.

Say when AI is not the answer

Sometimes the problem is a process nobody has fixed, or one person doing two jobs, or a form that asks for the wrong things. Automating a broken process just produces broken results faster. We would rather lose the work than build that.

Plain English, always

Every document, every explanation, every conversation. If you cannot tell your bookkeeper what a workflow does and why, we have not finished explaining it. Jargon in this industry is usually there to stop questions being asked.

Careful with data

Business-tier AI services with no-training terms, so your documents are not used to train anybody's model. Access scoped per workflow, so an invoice process sees invoices and not the whole business. If a workflow needs something sensitive it does not strictly need, we design it differently or we do not build it.