There is a problem in business that we don’t talk about enough.
Businesses have an enormous number of problems they want to solve. Business owners know what needs to be done, but they often don’t have the knowledge, skills, time, people, or systems required to do all of it properly.
Their employees face the same problem.
And technology, until now, has largely been a collection of tools that helps them do the work.
So the business owner ends up carrying an enormous operational burden.
Accounting. Procurement. Inventory. Payroll. Compliance. Reporting. Customer management. Purchasing. Reconciliation. Forecasting. Administration.
Every function requires knowledge.
Every function has processes.
Every function has exceptions.
And every function produces an enormous amount of information that someone has to interpret and act upon.
For years, we’ve tried to solve this by hiring more people and giving them better software.
Then AI arrived.
And suddenly, something changed.
For the first time, we have machines that can understand language, documents, context, patterns and complex instructions at a level that makes many previously unimaginable forms of automation possible.
We are getting remarkably close to the thing businesses have always wanted:
Something that can just do the work.
But there is a catch.
AI by itself cannot run a business.
At least not reliably.
A frontier model can look at an invoice and understand it.
It can categorize expenses.
It can analyze financial statements.
It can reason about inventory.
It can identify anomalies.
It can write reports.
It can even make remarkably sophisticated recommendations.
But understanding something is not the same as operating it with precision every single day.
Consider a restaurant.
An AI can understand that chicken is an inventory item.
It can understand a recipe.
It can calculate theoretical food consumption.
But it doesn’t automatically know that the kitchen actually consumed 3.7 kilograms of chicken today.
It doesn’t inherently know how much was wasted.
It doesn’t know whether the variance came from over-portioning, spoilage, incorrect counting, theft, a recipe change, or an error in the stock records.
To determine that reliably, you need a properly designed operational system.
You need recipes.
You need units of measurement.
You need opening and closing stock.
You need purchasing records.
You need production records.
You need sales.
You need wastage.
You need transfers.
You need adjustments.
You need rules.
You need history.
You need controls.
And then you need AI sitting on top of all of that, interpreting the information and helping determine what should happen next.
The same is true for finance.
A frontier model may be capable of understanding a company’s finances extraordinarily well.
But running the finances accurately every day requires transaction structures, source documents, account mappings, reconciliation rules, audit trails, permissions, validation, exception handling and historical state.
The intelligence is not enough.
The intelligence needs a system to operate inside.
This is where I believe the next generation of business software is going.
The old model was simple:
Software gives humans tools. Humans do the work.
The emerging model is fundamentally different:
Software structures the domain. AI performs the work. Humans manage the exceptions and decisions.
That distinction is enormous.
Because AI doesn’t eliminate the need for software.
AI changes what software needs to become.
We don’t necessarily need one giant AI that knows everything about every business.
We need deeply structured systems that understand how particular domains actually work.
A procurement system needs to understand procurement.
An inventory system needs to understand inventory.
A restaurant operating system needs to understand recipes, ingredients, production, consumption, wastage and sales.
A financial system needs to understand transactions, accounts, reconciliation and controls.
The AI then becomes an intelligent worker inside those systems.
It can read.
It can classify.
It can reconcile.
It can investigate.
It can predict.
It can recommend.
It can execute.
And increasingly, it can act.
But the system gives it the structure, data and boundaries required to do those things reliably.
That is the part that I think is being underestimated in the current AI conversation.
The future isn’t simply about having smarter models.
It is about building better environments for those models to operate in.
And that creates an extraordinary opportunity.
There are thousands of things businesses currently pay people to do because they are too complicated, fragmented or context-dependent to automate reliably.
Many of those problems were previously considered fundamentally human problems.
I don’t think they necessarily are anymore.
But solving them won’t come from putting a chatbot on top of an existing piece of software.
It requires rebuilding the underlying operational system around what AI can now do.
That is one of the reasons we built Qvian.
We are not interested in simply adding an AI button to business software.
We are interested in taking the actual mechanics of running a business and structuring them so that AI can participate meaningfully in the operation.
Because we are now solving problems that, only a few years ago, would have seemed almost impossible to solve through software.
And perhaps the most exciting part is that we are still at the beginning.
The question is no longer:
“Can AI do this?”
The more important question is:
“What would we have to build around AI for it to do this reliably?”
That is where I believe the next generation of business technology will be built.
Not around AI alone.
But around AI + structured systems + domain knowledge + operational data + human oversight.
That combination could fundamentally change how businesses are run.
And eventually, the amount of work required from a business owner may become dramatically smaller.
Not because AI replaced the business owner.
But because we finally built systems capable of carrying much more of the operational burden.
