What an integration actually involves
Most businesses we speak to are not short of software. They are running a stack that grew one decision at a time, where nothing talks to anything else, and the owner cannot see their own business without asking three people and waiting.
That is the problem the word integration actually names. Not adding another tool. It is connecting what is already there so the work moves between systems without a person carrying it, and so the numbers can be seen in one place without somebody rebuilding a spreadsheet every month.
The AI part is often the smallest part of it. The work is in the systems on either side: what shape the data is in, what each system will let you read, and what the business has actually agreed the process should be.
What the systems on the other side are actually like
The word integration suggests two systems that would like to talk to each other. What is usually there is two systems that were never designed to be read from, and a decade of data typed by people who were busy.
On one engagement, eight years of accounting records and a customer list assembled from email domains had almost nothing in common. Matching them on exact name linked six companies out of roughly two thousand. The accounting customer code existed on sixteen. That single gap was why every customer portal showed an empty order history: years of trading was sitting in the database and was invisible, and no amount of AI would have found it, because nothing was wrong with the AI.
It took three passes to fix, and the failures in between are the useful part. Matching on words in company names linked two unrelated businesses because both contained the word group. A rule had to be written that branches of one parent organisation are separate customers and never the head office. Companies had to be created from the clean accounting names rather than the messy ones. The end state attributed 99.8 per cent of the history, and the remainder is blank in the source system itself.
The shortcut that was measured, then cut
A decade of design files lived in one mail archive, saturated with every customer's work at once. The bulk import had brought in the attachment records without the files behind them. The fast recovery was to match by filename across messages. It was cut, because it would eventually surface one customer's artwork inside another customer's folder, and because measuring it first showed it recovered almost nothing anyway.
And the things that were simply wrong underneath
- A database change recorded as applied whose tables had never actually been created. Three screens were dead and a board was empty, and nothing reported it.
- A permission rule that worked for staff by accident and locked out every customer.
- Duplicate version numbers in the change history, which meant a test copy of the database could not be rebuilt for months, so the tests that mattered most were the ones being skipped.
Why this is on a services page
None of it is an AI problem and all of it decides whether the AI is fed anything true. If a proposal you are reading does not mention the state of your data, either somebody has looked and is not telling you, or nobody has looked.
What implementation involves, in the order it usually happens
Almost every engagement starts smaller than the problem that prompted it, and that is deliberate. The pain is usually described at the level of the whole business; the sensible first piece of work is one process that matters.
Integration is the part people search for and implementation is the part they actually buy. The difference is not pedantry. Integration is getting two systems to exchange something, and it is often a week of work. Implementation is getting the business to run differently afterwards, and that is the part carrying the risk, the cost and the arguments.
One process, working properly
Quoting, intake, the follow-up that keeps getting missed, the report somebody rebuilds by hand. Contained, quick to judge, and it either runs or it does not, which makes it the cheapest way to find out whether the rest is worth doing.
The same process, measured
Once something runs, the question becomes whether it is better. That needs a number that existed before we arrived, which is why the first engagement usually establishes one.
The connected system
Everything joined, your own knowledge in the core of it, and visibility across the business rather than in six places. This is the one with a real obligation on your side: your people's knowledge has to go into it.
Where we are, and what we are not claiming yet
We work with Sydney businesses and we are Australian-run, through an Australian company. What we are not doing on this page is listing a street address or claiming a service radius, because we would rather publish that when it is precisely true than publish it now and correct it later.
A small claim is the tell
This market has watched a lot of people make confident claims about AI that did not survive contact. A firm that will not overstate its own front door is making a small, checkable promise about how it handles the larger ones. If a Sydney address matters to you, ask us and you will get a straight answer.
Who you actually deal with
Two founders, both on every engagement, both reachable. There is no bench, no account manager, and no junior doing the work while a senior name appears on the proposal.
That is a limit, not a positioning line. It caps how many engagements can run at once, and it is the reason we would rather turn down timing than take work we cannot be present for.
What we will tell you that you may not want to hear
- If the process is not agreed, software will not settle it. It will encode one person's version and make the disagreement more expensive to resolve.
- If the return does not clear the cost, we will say so. A project that cannot pay for itself is one we would rather not have delivered.
- The prototype is the easy part. If somebody has shown you something impressive in a week, they showed you the part that takes a week.
- Deep back-end work is a larger undertaking than the front-end work we are strongest at. We would rather say that at the start than discover it in month three.
Nobody who contacts us arrives comparing suppliers
Nobody who has contacted us has arrived comparing suppliers. They arrive asking where to start, which is a different question, and it is answerable before anyone quotes anything.
The assessment takes about ten minutes and returns a written view of where AI would pay in your business and where it would not, including the cases where the answer is not yet. It is free and the report is yours whether or not we ever speak again.