Main content

Relate 2024 - Zendesk adds AI capabilities in support of 'human-first' CX

Phil Wainewright Profile picture for user pwainewright April 16, 2024
Summary:
As its Relate 2024 conference opens, Zendesk argues for AI that puts humans first, announcing new AI capabilities from a combination of recent acquisitions and its own in-house development.

Tom Eggermeier presents the keynote at Zendesk Relate in front of a huge image with the slogan 'AI that puts humans first'
Tom Eggemeier, Zendesk, on stage (Relate 2024 screengrab)

In the year since launching Zendesk AI at its Relate annual conference last May, Zendesk has grown its AI capabilities both organically and through three separate acquisitions. As this year's Relate opens today, it is presenting the combined fruits of those efforts as "AI that puts humans first," in the words of its CEO Tom Eggemeier.

Today's announcements include a co-pilot that will help human agents to interpret intent so they can better anticipate customer needs, as well as suggest ways they can streamline workflows, Zendesk says it will also learn from experience to improve future interactions. There are new capabilities for customer service admins to maintain and improve knowledge base accuracy and relevance, while new AI-powered reporting is slated to help service leaders monitor and optimize performance, across both human and AI resources. In a briefing for today's event, Matthias Goehler, Zendesk’s EMEA CTO, told us that AI analysis of customer feedback can help identify shortcomings in knowledgebase content, as well as suggest improvements:

We have AI capabilities where we are analyzing your knowledge base constantly against all the incoming inquiries and identifying whitespaces. We are identifying things like, this is not really helping the customer based on the feedback that we get from the customer — because the customer can rate every single interaction.

[There is] a constant feedback loop that you can take into account and say, this might be outdated, or you have gaps, you have stuff perhaps in your knowledgebase that you don't need any longer. We can identify this and make sure the knowledgebase is at the right level and the right quality — and then now use generative AI to also automatically suggest new articles or changes in articles ... to make the knowledgebase better.

Added from acquisitions

In addition to these organic enhancements to Zendesk AI, the acquisitions have added important new capabilities. Last month's acquisition of Ultimate has expanded the range of customer support automation, adding the ability to customize responses and workflows in complex use cases, while allowing for more seamless hand-offs between intelligent automated agents and their human counterparts. This soon after the acquisition, Ultimate is still a separate add-on to Zendesk's existing agent automations, but will be integrated over time. Goehler comments:

We are, after the acquisition, significantly investing in the engineering teams and in the products to make sure we can continue the innovation. But we can also work on integrating it to make it one seamless offer.

We have always been very strong on TCO, we have always been very strong on ease-of-use, speed of implementation, that's our DNA. This is what we will continue to do with the acquisitions as well.

More progress has been made with the earlier acquisitions of predictive workforce tools from Tymeshift last year and automated Quality Assurance (QA) from Estonian startup Klaus, acquired in January this year. Both are being introduced at Relate as Zendesk products, in the form of Zendesk WFM (WorkForce Management) and Zendesk QA, respectively. The acquisitions have brought additional AI resource into the company, as Goehler explains:

Both Tymeshift and Klaus have very strong AI capabilities. That was one of the reasons why we acquired them. Tymeshift in terms of using AI to create forecasts and schedules, so that you're always staffed at the right level across languages, skills, channels, peak times, et cetera. Klaus in terms of making sure, with auto-QA, [using] AI to monitor 100% of all interactions, and then filter out the ones where I really need to focus on.

These capabilities help Zendesk customers manage customer service at scale, with Workforce Management helping to rapidly assign resources where they are most needed across a large service organization, while the Klaus technology provides an invaluable extra check on service quality across both automated and human agents. Manual call quality checking typically only looks at a small sample of customer interactions, whereas the automated Zendesk QA tools make it possible to analyze every interaction. As well as monitoring voice calls by human agents along with emails and messaging, this also extends to automated interactions, ensuring that chatbot interactions also measure up to quality standards. Goehler points out the value of this additional QA check in the era of generative AI:

More and more [customers] use generative AI to create answers... Now hopefully — that would always be my advice to customers — you define the base on which the bot is creating the answers off, so that's your knowledge base, that is your help centre articles, that is your processes that you have defined that you go through, et cetera. So that the bot is not making up answers on its own.

But still, if it's using generative AI, even if you give the content to the bot, 'This is the right answer,' it still creates a text out of it. So it's a very, I think, sophisticated approach now to make sure you can also monitor and QA the answers of the bot and see, 'How was tonality there? How was sentiment there? Was the greeting okay?' and all of that stuff.

The automated analysis of each interaction looks at the phrases and tone used in the conversation, and then provides a QA rating that can trigger corrective action, whether it's additional mentoring and training for human agent, or fine-tuning of the AI model.

My take

Automation has long been a feature of customer service, dating back to the earliest interactive voicemail systems, but the advent of generative AI has brought new urgency to this trend. Zendesk is talking about customers achieving automation of as much as 80% of customer service interactions. From a productivity point of view, that sounds like great news, but from a risk management point of view it creates an obligation to analyze those interactions and make sure that they're delivering what customers need. Auditing the output of AI-fueled automation is often neglected but it's a crucial element in the enterprise AI stack. It's good to see Zendesk giving it the attention it deserves.

Loading
A grey colored placeholder image