top of page

13th of June 2024

How Ami helps to save over 10,000 animals a month

The client this case study is based on is one of the UK’s leading animal welfare charities, which receives approximately 24,000 to 35,0000 conversations per month. With this volume of conversations, Ami’s aim is to help the members of the public (MOPs) with their queries as quickly as possible, to maximise the number animals the client can help. A factor which slows this process down is vague queries.

Vague queries refer to queries that are ambiguous in nature, and require Ami to request more information in order to provide an effective answer. An example of this is if a customer were to only say ‘Delivery’. This query can mean the customer intends to ask about a range of things relating to delivery, from the delivery timeframe to making a complaint about a delivery they have received. When encountering these vague queries, Ami usually asks the customer to provide more information on the topic or asks questions to extract relevant information, which lengthens the conversation and thus slows down the overall resolution time. Additionally, data from other clients suggest that the more questions Ami asks, the more likely the customer is to exit the conversation prematurely (before their query has been resolved).

A way Ami combats this issue is the URL recognition process, which aims to make conversations more efficient by reducing the number of questions Ami needs to ask when encountering vague queries. This case study details the background leading to the implementation of the URL recognition set up, the current set up, and the result of this set up.

Vague Journeys: Previous Set Up

Prior to the implementation of Ami’s URL recognition set up, Ami’s vague journey set up in this client is as below:

This set up includes Ami asking the customer what best describes their query, then listing popular conversation topics for that type of animal for the MOP to choose from. This set up is used for every vague conversation, regardless of the URL the customer is on.

The decision to implement Ami’s URL recognition came about because the likelihood of MOP’s queries being about the topic of that URL is very high. For instance, if the MOP has a vague enquiry as above, and is conversing with Ami on the page about reporting cruelty or neglect, we predict that they likely would want to report cruelty or neglect towards an animal.

By implementing the URL recognition, Ami would be able to predict the MOP’s intent based on the URL, thus increasing Ami’s accuracy by reducing the number of vague journeys occurring and improving MOP conversation experience.

URL Recognition Implementation

The implementation of this URL recognition set up involves Ami considering the URL on which the conversation is taking place when encountering a vague query, and then skipping the initial vague question that asks the MOP what best describes their query. To confirm the effectiveness of this strategy, we have initially asked whether the MOP is enquiring about the topic on the URL as a confirmation question.

For instance, if the MOP asks a vague query such as ‘There’s a cat’ on the URL about reporting cruelty/neglect, Ami would ask ‘Please confirm, are you enquiring to report cruelty/neglect?’.

Similarly, if the MOP asks the same query on the URL about giving up a pet, Ami would ask ‘Are you enquiring about giving up your cat?’.

If the MOP says yes or gives any other affirmative answers, Ami would provide the appropriate advice. For instance, offering to escalate the MOP to an agent or phone number to report the cruelty/neglect, or providing a link about pet care support and rehoming a pet.

If the customer says no or gives any other negative answers, Ami would respond similarly to the old vague journey set up by asking the MOP what best describes their query, then list popular conversation topics for that type of animal, but leave out the topic Ami had already asked about (the recognised URL topic). For instance, if the conversation occurs on the reporting cruelty/neglect URL and Ami has already asked if the MOP is enquiring to report cruelty/neglect, Ami would ask if the MOP is enquiring about reporting an injured animal, a stray animal or other:


As of March 2023, 10% of total chats involved the URL recognition function; these would have otherwise been considered vague journeys. The reduction of vague journeys then in turn leads to an improved customer experience.

Out of these chats, in 94% of conversations the MOP’s intent matches the topic of the URL page. Considering the accuracy of the URL recognition in predicting the MOP’s intent, the next step is to remove the confirmation question entirely, to ensure an even more concised conversation, further decreasing the resolution time and improving the MOP’s experience with Ami.

As predicted, Ami is able to accurately recognise the MOP’s intent based on the URL on which the conversation occurs, and ask for confirmation in a more precise way. This leads to a decreased number of vague journeys, therefore a decreased resolution time and an increased number of conversations about animal welfare resolved in a time period.



Sarin Chinavicharana



Couldn’t Load Comments
It looks like there was a technical problem. Try reconnecting or refreshing the page.
bottom of page