Zendesk extends its generative AI offering, balanced by new data privacy measures

At a special event in New York today, CX vendor Zendesk announced further enhancements to its generative AI offerings, most notably a voice transcription and summary capability, along with new data privacy measures to allow greater control over customer data management. Introducing the event, Tom Eggemeier, CEO of Zendesk, commented:

AI is reshaping how we all think about customer experience, and quite frankly, employee experience as well. AI is already having an enormous impact on how we all do business. And some of the biggest changes are going to come fortunately, or unfortunately, to the customer service space.

He cited the example of longstanding Zendesk customer, the upmarket London department store Liberty, which adopted Zendesk AI shortly after it was first introduced in May. He outlined the results at Liberty:

They've seen their customer satisfaction up by over nine percentage points already. Their first reply time is down by 73%. The time it takes to resolve the average ticket is down by 11%. They are providing quicker, more personalized service and their customers are noticing this. The customers don't know it's because of AI. They just know they're getting better service, which is connecting them with the Liberty London brand even more.

Extending Zendesk AI

The new generative AI capabilities announced today build on and extend the Zendesk AI package unveiled in May at the company's annual Relate conference. Today's announcements include:

• Expanded intent detection founded on Zendesk's CX-specific large language models to serve additional industries including insurance and travel & tourism, or to match internal IT and HR use cases. These models can be further adapted by taking common intents that an individual company encounters from its customers or employees, and editing them for their specific needs. For example a travel agency might want to detect frequently asked questions by travellers and escalate them to agents who are specially equipped to handle these issues.
• New tools to help agents, administrators and managers optimize and customize their CX processes. For example, Zendesk AI can identify gaps in the existing question and answer flows based on common topics customers are asking about and suggest new flows to plug these gaps. Or admins can adjust the tone of a bot's responses to be more casual or professional to match the brand persona.
• The ability to summarize key information from multiple help centre articles, and automatically provide a more accurate and relevant answer. Bringing this information together in a more digestible format will enable agents and bots to respond to customers in a more consistent and natural conversation flow.

The launch of AI for Voice brings the ability to not only provide an immediate transcript and summary of the conversation, but also identify how the customer felt about the call. This promises to save time spent manually recording conversation details and allow agents to focus on ensuring the customer’s issue has been properly addressed.

Data privacy

For customers concerned about data privacy, Zendesk has announced an Advanced Data Privacy and Protection offering, which allows businesses to have greater control and flexibility over how they manage their customer data. Amang the new capabilities this brings, CX leaders can:

• Safeguard their business with stronger encryption, including private keys and admin control of key rotation and revocation.
• Elevate privacy by specifying what data is shown and kept, for example redacting email and shipping addresses, or setting custom rules for how the long the data can be retained and when it should be deleted.
• Strengthen security at scale with greater visibility by logging who accessed and searched for data.

My take

When we spoke to Zendesk at the time of the May announcements, Cristina Fonseca, Head of AI at Zendesk, emphasized the company's ability to leverage its deep experience in CX to tune its generative AI models, contrasting it with the more generic approach of its partner, ChatGPT creator OpenAI. She told us:

OpenAI doesn't have the proprietary data that we have, the CX data that we have ... If you want to build intent detection and sentiment detection on top of OpenAI, that gives you a good advantage. But we go even one level further.

Today's announcements of industry-specific and function-specific intent detection build on those proprietary models and also offer customers the option to further fine-tune the response flows provided. This is where Zendesk has a real strength in the market, but it also needs to recognize the reservations enterprises have about the reliability and trustworthiness of generative AI models. Therefore the enhanced data privacy measures are welcome, but perhaps don't yet go as far as we've seen from other vendors with more extensive AI investments.

There's an important balance that vendors have to strike between on the one hand supporting customers in realizing the productivity and competitive advantages of generative AI, while on the other hand ensuring that trust and data privacy is fully respected. The speed at which generative AI has evolved is putting unprecedented demands on vendors to deliver new functionality while constrained by considerations of data privacy, accuracy and appropriate use. Getting this balance right will test the mettle of every vendor.