Salesforce at the club – Episode 2
A prompt instead of typing up delivery notes
The weekly ordering ritual
After a successful weekend, our cold room usually looks exactly how you would want it to: pretty empty. So we order on Monday, the order is picked on Tuesday and delivered on Wednesday. A perfectly normal process.
The unpleasant part only starts afterwards.
Because as soon as the stock arrives, the paperwork begins.
Check the delivery note. Review the line items. Compare quantities. Check prices. Then enter everything into the system manually. None of it is particularly complicated.
But these are exactly the tasks that take time week after week, introduce errors and, honestly, are not much fun.
At some point, I asked myself:
Why should a person still have to type data out of a PDF?
The answer was surprisingly simple: They should not.
Our supplier already emails us the delivery note as a PDF before the delivery arrives.
It contains everything we need: product numbers, quantities, pack sizes, purchase prices and order numbers. In other words, everything needed to create a goods receipt automatically.
Why the platform suddenly becomes so useful
The emails already arrive in Salesforce. Email-to-Case automatically assigns them to the right supplier and stores them with their attachments. The data is already where we need it. It would be a shame to keep asking someone to type up PDFs at that point.
So AI enters the picture.
A prompt reads the delivery note, understands its content and converts the information into a structured format. Salesforce then takes care of the rest. For each line, it finds the matching product using the product number and automatically creates a goods receipt.
Quantities, prices, the delivery date and order number are transferred directly, and stock levels increase automatically. Technically, the solution is surprisingly unspectacular: a prompt template and a small Apex action connected through a flow.
Done.
And that is exactly what I have come to love about the platform.
The individual building blocks already exist. You just need to connect them sensibly.
The data was already there. AI just needed to understand it.
AI does not replace the human straight away
For now, we deliberately still trigger the process manually. When a new delivery note arrives, the result can be checked and confirmed. That was an important step for us.
We are not trying to automate everything with AI straight away. We prefer to introduce it into existing processes step by step.
Observe first. Then improve. Build trust. Only then automate. After several months of use, that trust has grown enough for us to take the next step. In future, the whole process should start automatically as soon as a new delivery note arrives.
Observe first. Then improve. Build trust. Only then automate.
The real gain is not the time saved
Incidentally, the real value is not in the time we save.
It is in the quality of the data. Price changes are detected automatically, and relevant changes are sent straight to Slack.
New products can be created faster, and when checking incoming deliveries, stock quantities often only need to be confirmed.
Of course, we can still measure the benefits:
Three hours a month is hardly revolutionary, even for a club.
But it is these small improvements that make the difference in the end.
AI does not always have to be SPECTACULAR. It has to CREATE VALUE.
Sometimes it is enough that nobody has to type up a delivery note any more.
And that is what I like so much about our project now. We are not looking for grand showcases. We are looking for the boring tasks nobody enjoys and letting AI take them over.
And perhaps that is the nicest realisation: a club is ultimately a perfectly normal business, just with better music and later working hours. That is exactly why we find the same small opportunities for improvement here as you probably would in almost any other company.
