How Kriptomat achieves 62% AI resolution, saving 172 hrs each month

Last year Kriptomat had a problem: their support team was stretched, response times weren't as fast as they wanted and with a global customer base they really needed 24/7 support.

So they turned to My AskAI.

Since integrating their My AskAI customer service agent into their Intercom helpdesk it has resolved over 5,547 support tickets and is currently deflecting 62% of their tickets and saving their team 172 hours every month.

Now you’re probably asking yourself: “how does Kriptomat do this?!” and “how can I do this for my business?”.

Let’s take you through their setup and show you how.

Kriptomat AI customer service results on Intercom at a glance

What does Kriptomat do?

Let’s kick off by explaining a little bit about Kriptomat:

Kriptomat are an EU-licensed cryptocurrency platform that's been around since 2018 and is based out of Estonia. If you run a crypto business on Intercom, our roundup of the best AI customer service tools for crypto companies covers the wider field.

They currently have over 400,000 users across Europe and their platform was built to make buying, selling, holding, and staking digital assets simple for everyday users.

With a focus on accessibility and compliance, Kriptomat aims to bridge the gap between traditional finance and the crypto world by offering features suited for beginners and more experienced users alike.

Which helpdesk does Kriptomat use?

Kriptomat use Intercom as their helpdesk platform, and receive the majority of their support requests through the Intercom chat widget, with some additional volume coming in via email.

They had looked at Intercom’s Fin but the $0.99 per resolution pricing just didn’t make economical sense for them (we compare the two in detail).

So they chose to sign up for My AskAI’s approved Intercom app instead, and run our AI agent for Intercom in the inbox their team already had.

How did they train their AI customer service agent?

Kriptomat had an Intercom help center already, so they connected this to their AI agent using our Intercom connector to get things going.

But they also had a lot of useful training data (tutorials, blog posts, terms and conditions) on their site too (over 7,000 pages of it in fact!) that they didn’t want to go to waste.

So they added this to their AI agent’s knowledge as well using our unlimited website sync to ensure that the content always remains fresh and their AI agent was as knowledgeable as could be.

“Personally, I'm a big fan of the direct integration with our pre-existing help articles, and how easy it is to re-train the agent when it's providing outdated information.

It has helped our team out immeasurably especially during heavy inquiry surges over the past few months!!”

- Hannah DiBella, Support, Kriptomat

When did they decide to turn on ‘direct replies’ to customers?

They decided pretty much from day 1 that the fastest way for them to learn how their new agent would perform would be to put it in front of their users and start gathering feedback.

In the worst case scenario, they figured there were always lots of ways the users could request to speak to someone and hopefully in some cases it would give a number of users faster answers.

They were pleasantly surprised to see that from go-live, due to their extensive collection of articles and tutorials, that they were able to see a ~50% resolution rate.

What was the biggest thing they did to improve their AI agent’s resolution?

But Kriptomat didn’t want to stop at 50%, they wanted to see how much further they could take their AI customer service agent.

So they regularly checked in on their Insights to see what people were asking, reviewed and fixed their knowledge gaps and created custom answers to make sure their agent had all the knowledge it could to answer questions.

Through this continual improvement process, they have managed to increase their average AI resolution to where it sits today, at around 62%. Our self-learning now does much of that gap-closing on its own, drafting a new help article each time the AI meets a question its knowledge cannot yet cover.

How do they customize their AI agent setup to work for their business?

Each business manages support in their own way and has nuances to their setup, that’s why our AI agent has multiple customization options to fit around you.

For Kriptomat, this came in the form of:

• Using our different types of guidance to ensure their AI customer service agent responded on-brand and handed over if legal or fraud issues arose
• A feature allowing them to auto-close AI responded tickets after a certain time period and, more impactfully
• The use of our AI Copilot in Intercom, to enable their agents to use our AI, even after a conversation has been handed over.

“Amazing tool, keep it up guys, My AskAI has revolutionised the way we offer support to our customers!”

Martina Mitic, Head of Customer Success, Kriptomat

What impact is their AI customer service agent having now?

Well, let’s see:

As a result of Kriptomat’s continual efforts to review and add custom answers to their AI customer service agent knowledge, their AI agent is currently:

• Resolving 62% of tickets automatically each month across almost 1,700 tickets (that’s around 1,000 tickets per month their team doesn’t have to deal with directly).
• Saving their customer service team 172 hours per month (assuming a 10 minute per ticket handle time).
• Achieving a 61% AI CSAT score across those 1,700 tickets.

Where do they go from here?

There is always going to be a limit to how far static knowledge alone can take an AI agent and Kriptomat know this.

That’s why they are looking to take the next step to provide their AI agent with access to customer data using our User data API.

This could help their agent respond to questions about transactions or account balances.

Once they’ve managed to get this in place, their next ambition will be to look at features like tools and tasks to automate more time-consuming tasks like ID verification.

Watch this space!

Read more case studies or explore our plans and pricing for My AskAI.

Learn more about building an AI-first support operation in our complete guide.