Will AI Assistants Know How to Forget, Too?

clock Oct 09,2026
gestionale conosce azienda

“The data’s there, but I need to tidy it up before sending it over.” If this sentence comes up at every meeting, it’s worth pausing for a moment. Your business software records orders, issues invoices and keeps daily operations running. Yet whenever someone needs to understand what’s happening in the company, a colleague has to export a file, cross-reference it with another and call the person who knows how to interpret it.

Meanwhile, the business has changed. It has launched an e-commerce channel, added services to its products and delegated responsibilities that once rested entirely with the owner. But the software still uses categories and procedures defined years ago. People have found ways to keep things moving, building little bridges over each limitation.

That’s how a system can continue working while describing a company that no longer exists. Recognising this doesn’t mean you need to replace it immediately. It means understanding how much work it now takes to reconcile what happens in the business with what the system can represent.

The Real Work Has Moved Elsewhere

An Excel spreadsheet alongside your business software isn’t proof of inefficiency. It might support a simulation, an occasional analysis or a routine check. It becomes a different kind of signal when it holds essential information the main system cannot handle.

Think of delivery dates updated only in the sales manager’s spreadsheet. Or job costs calculated manually because labour, materials and subcontracted work remain separate. In these cases, the spreadsheet stops being a supporting tool: it becomes the place where decisions are made.

The useful question is how much the company depends on that extra step. If the file isn’t updated, who loses visibility? If the person maintaining it is away, who knows how to reconstruct the figures? The answers reveal where the process has drifted away from the software.

When Old Categories Distort the Picture

Imagine a company that sells machinery and has gradually built a subscription-based maintenance business. If its software still treats everything as product sales, total revenue might be correct while the value of the two activities becomes difficult to distinguish.

New questions arise. How much does it cost to support each customer? Which contracts consume more hours than expected? How much revenue depends on renewals? If the data collection process wasn’t designed to capture this information, the final report will require painstaking reconstruction—or remain incomplete.

The limitation may therefore sit well before the reporting screen: in master records, transaction categories or the way costs and revenue are linked. Even an attractive dashboard inherits the definitions on which it is built.

“We’ve Always Done It This Way” Gets Built into the Software

A customisation can solve a real problem. Years later, however, it may still be active even though that problem has disappeared. An approval still goes to a role whose responsibilities have changed, or people keep completing a field that nobody reads anymore.

Each step carries a small cost in isolation. Repeated across hundreds of transactions, it becomes time taken away from other work. And when a procedure is too cumbersome, people may start working around it: entering placeholder values, sharing information by email or filling in missing data at the end of the month.

Cloud and AI Won’t Fix a Flawed Definition

Moving business software to the cloud can change how it is deployed and managed. Connecting it to other tools can reduce some manual data entry. Neither step, on its own, clarifies what the company means by an active customer, a completed order or a profitable job.

The same applies to artificial intelligence. Asking an assistant which customers generate the highest margins is useful, provided costs are allocated consistently. If some support hours are recorded in messages and others in the business system, answering the question requires information that may not be available.

Automation therefore needs some less glamorous groundwork: deciding which data to collect, who is responsible for it and when it must be updated. Without that agreement, even a quick answer may offer little value.

Adapt or Replace? Start with a Real Case

To assess your system, choose an important process and follow it from beginning to end. Take an order, for example: from quotation to delivery, through to payment and margin calculation. Include the steps that happen outside the software, because that is often where the work needed to make everything add up is hidden.

This review may reveal different kinds of problems. An available feature that was never configured needs a different response from a missing integration. Poorly entered data may point to a training issue; a structure unable to represent the new business model may require a more fundamental rethink.

Any potential replacement should also be tested against real cases. A convincing demonstration should show how the new system handles a complicated return, a partial delivery or a contract amended midway through. These situations reveal how well the product fits everyday work.

The right business software should let people recognise their company in the data they consult. If explaining every figure requires starting with “Yes, but…”, the system may still be working for an earlier version of the business.

Categories

Create your account