So we have all figured out that chatGPT is great at understanding text and having a conversation. We also know that we can feed it with conditions. That is all fantastic but in the end we want the conversation to result in something that we can use on our business logic like summarizing an order into an XML. Let’s try it.
So chatGPt is great at understanding but the output from chatGPT must also be easy to understand for other programs so that the resut can be used in more traditional programming or systems. So I decided to try it by creating this prompt
We are receiving pizza orders. start the conversation by welcoming me to pizzeria Venezia and ask if I would like to order and wait for customer input. If you feel uncertain about the placed order ask the questions needed. When the order is understandable summarise it in a table and ask for confirmation. If the order is confirmed then reply with an xml in the format:
<order>
<pizza>
<name></name>
<extras></extras>
<remove></remove>
<count></count>
</pizza>
</order>
I am the customer you are the pizzeria.
Those are our available pizzas, any ingredients mentioned can also be ordered as extra to other pizzas
MARGHERITA
95:-
VESUVIO
ham 110:-
At the end of the prompt I just copied and pasted the menu from a random pizzeria website just to see if chatGPT figured it out and it did.
Welcome to Pizzeria Venezia! How may I assist you today? Would you like to place an order?
Yes, what do you have?
We have a wide variety of pizzas with various toppings and extras. Would you like me to go over the menu with you or do you have a particular pizza in mind?
give me the menu
And it gave me the menu so I decided to put it to the test right away.
I would like one margherta and two bussola but exchange the shrimps for olives on one of them.
And it managed to summarize it into a table which I confirmed and received an XML that I can then use in my business logic. Extremely easy to do and implement. This can be applied to any customer integration like taking orders, handling customer service chats or why not combine it with voice to text and text to voice technologies? Imagine how many customers you could service instantly.
Add to that: it speaks any language…
