Intelligent Search

The Intelligent search widget provides a list of your customer's transactions, where they can easily query or ask questions about their finances and stay up to date with their most recent transactions.

This guide will walk you through how to use and navigate the Intelligent Search widget effectively.

📘
If you're interested in Intelligent Search but do not wish to use Bud's widget you can integrate directly with our Intelligent Search APIs which can be found here.

Getting started

In order to use Intelligent Search you will first need to create a customer on the Bud platform and ingest transaction data for that customer. This can be achieved through either Open Banking or through Bud's First Party Ingestion API.

Once you have transaction data on the Bud platform you can generate the Intelligent Search widget by calling the Intelligent Search Frontend Widget URL endpoint. This will generate an embeddable link that can be placed within your online or native applications.

How it works

Intelligent Search works by displaying a customers enriched transactions back to them, it then utilises text embedding as well as large language models in order to allow customers to search or ask for anything related to their transactions and receive a list of relevant transactions and an insight in response. This works not only across all of Bud's enrichments such as searching for categories or merchants, but also searching for more obscure terms such as "car" or asking questions such as "when was I last paid?".

The various elements of Intelligent Search are documented below.

Transaction List

Provides a date ordered list of a customers transactions based on transactions stored in the Bud platform
The list of transactions can be filtered using the date filter on the right hand side
Free text query input at the top allows customer's to search
Each transaction can be clicked into to open the transaction details view

Transaction Details View

An overview of details that Bud holds on a given transaction
Both raw information such as the date, time and status of the transaction as well as enriched information such as category l1, l2, merchant_name and location information is displayed
If an exact location has been identified then a map will be present showing the location of the transaction

Queries and questions

Questions and queries are automatically routed to the relevant model based on the input from the customer
Results include a list of relevant transactions for the customer as well as an insight personalised to the query/question
Smart filters are also returned giving customers suggestions of subsequent queries that they could ask

Insights

Insights returned will be generated based on the query or question that the customer has asked
Examples of insights include category totals, merchant totals, spending trends and increases in regular payments to name a few
Customers are able to view the transactions used to generate an insight by clicking on the "View {x} transactions" beneath the insight

If you have any questions, please contact us via the chatbot (bottom-right of screen 👉) or via a support request or check our FAQs.