?? Large Language Models for API Calls with Kyriba ??

?? Large Language Models for API Calls with Kyriba ??

Read time: 4 minutes

Hello,,

?? Welcome (back) to The AI Treasury Insight Newsletter! ???

In this week's?edition, we discuss the usage of LLM (Large Language Models) in calling APIs and the research conducted by Kyriba on that matter.?

Let's break down their solution and see how the research around AI in Treasury will transform Treasury Departments.? Please note that this article is not sponsored, we just found their solution interesting and worth talking about!???

???What does "Calling an API" mean????

Before we get into the nitty gritty, let's make sure we properly understand all the terms at hand here. Starting with an "API call" and breaking down what an API is

???Demystifying APIs???

API in Simple Terms?????

  • An API, or Application Programming Interface, is like a menu in a restaurant
  • Imagine you're sitting at a table with a menu of choices to order from
  • The kitchen (system or application) will prepare your order, but you need a way to tell them what you want
  • The waiter (API) is the messenger that takes your order (request) and tells the kitchen (system) what to prepare, then returns the dish (data or functionality) back to you.

API in Corporate Treasury???

In a Corporate Treasury context, APIs can be used to:

  1. Automate Transactions: Send payment orders directly from internal systems to banks without manual input.
  2. Retrieve Account Information: Automatically pull bank balances or transaction details into internal treasury management systems.
  3. Connect to Third-party Services: For example, getting real-time currency exchange rates from external providers.
  4. Integrate Different Systems: Make sure treasury software, ERP systems, and other corporate applications can "talk" to each other, sharing data seamlessly.

Making the connection with something you might have heard of in Treasury: the type of connectivity you set-up with your bank.?

If the connection is a bit streamlined (meaning you don't go on your bank's portal), there are high chances that you are connected to your bank(s) either via the SWIFT Network, or via H2H (Host-To-Host).?

API is another (serious) option to consider when seting up your bank connectivity as it can real time data.??

Essentially, APIs allow Corporate Treasury departments to automate tasks, ensure data accuracy, and integrate various systems to operate more efficiently.

???What is an "API Call"?????

API Call in Simple Terms???

Using the restaurant analogy from before, an "API call" is like placing an order from the menu.

When you tell the waiter (API) you want a specific dish (data or function), you're making a "call" to the kitchen (system) to prepare that dish for you.?

The waiter then brings back the dish (the result) to you.

API Call in a Treasury Application???

Imagine a treasury management system (TMS) that a company uses to manage its bank accounts. The TMS needs to know the current balance of a particular bank account at the start of each day.

  1. Placing the API Call: The TMS sends a request (makes an API call) to the bank's system asking for the current balance of a specific account.
  2. Bank's System Processing: The bank's system receives the API call (request), fetches the balance for that account, and prepares to send it back.
  3. Receiving the Result: The bank's system sends the account balance back to the TMS, completing the API call.

In this scenario, the "API call" made by the TMS allows it to automatically fetch the account balance without someone manually logging into the bank's portal and checking. This automation ensures that the data in the TMS is up-to-date and accurate, making treasury operations more efficient.

?? Keep reading to understand how LLM can enable smooth, AI drive API calls??


???Kyriba's Deep Dive into LLM for Treasury Management???

Now the "issue" with API is that, it still requires a bit of knowledge around what parameter to enter in order to get the right data from my "Calls". Also, how to make an API call in the first place??

  • With Large Language Model, a Treasury could simply talk to its machine for it to interpret correctly what data it needs to retrieve and what "Call" to make

Retrieve your balances for the last quarter???

  • Let's say you want to retrieve your balances for the last quarter, but don't know any account number or any other details
  • You could communicate with your LLM with the simple request "Retrieve my Balances for the last quarter"

Kyriba's LLM solution

  1. Indicating which parameters are missing?The first thing Kyriba's solution will do is to pin-point what parameters are missing in the request, for the Treasurer to be?guided through the process from the startTo the specific request "Retrieve my statements for the last quarter", Kyriba's LLM would typically ask you for:The ActualsA Statement IdentifierA Date TypeNow remember, as the Treasurer who didn't have much information to begin with, you actually might not know what those parameters are, let alone indicating the right ones to the LLM!
  2. Displaying all the accounts to retrieve the correct parametersSince you don't actually know what to input as actuals, statement identifier nor date type, why not asking the LLM?Kyriba's solution proposes to request the following "Show me all my accounts created last month"And guess what, the app does just that, and displays all the accounts created last month...
  3. Being more specific and getting the parameters neededNow that you have all the information, and you know what you want to focus on, why not being more specific?Let's ask the app to?"Get my accounts in France in Euro created last month"From there, you can find the accounts created last month in France, in Euro, along with the parameters required to complete the initial request
  4. Using the account code and date type to know your balanceNow that you have the information to your parameters (i.e. Bank Account Code), let's ask the app the following:"What are my actual balances for the past quarter (value date) for account with code LLM_FR1?"And now the model extracts the data and gives it to you under the form of a comprehensive plotYou can even ask the LLM to give you the average actual balances for any of those account!

?What to see what it looks like? Click here to read watch the video!

Thank you Basile Etienne for the very insightful video ;)

If you want to dive deeper into Treasury-related topics or have specific areas you'd like us to cover in future editions, we encourage you to visit our website.

There, you'll find a wealth of resources and can also reach out to us directly to share your feedback and suggestions.

Have a question? Answer to this email and I will make sure to answer you!

要查看或添加评论,请登录

Corporate Treasury 101的更多文章

社区洞察

其他会员也浏览了