ข้ามไปยังเนื้อหาหลัก

Turning AI into purposeful platforms for customers

Zendesk CTO Adrian McDermott gives an update on the company’s latest artificial intelligence (AI) innovations

เผยแพร่แล้ว 14 กันยายน 2565

Over the last 70 years, AI has progressed dramatically, especially in customer service. We have all experienced this improvement, from crappy conversations with a bot that never understood what we needed, to much more seamless interactions with AI today that often instantly provides help. There continues to be massive potential for how AI can better serve customers across all service interactions, and each year our Zendesk CX Trends research shows more businesses and customers expect to use and interact with AI.

We continue to invest in building AI at Zendesk and most recently, launched new customer sentiment and intent functionality powered by machine learning that can provide more personalized experiences. Customers no longer need to repeat themselves, as agents are automatically assigned tickets based on their expertise, and given relevant context and recommended solutions for a quick resolution.

This approach is key to delivering better experiences. Most customers don’t want to be wowed, they just want service to work really well — in a perfect world, it should all be invisible to them. Triaging requests is a foundational element for efficiency and helps businesses drive faster resolutions as well as foster customer loyalty. This customer loyalty is tied to repeat business, which all company leaders can agree positively impacts the bottom line.

An approach to extending AI
In customer experience (CX), AI works particularly well today in three key use cases: automation, recommendation, and prediction. Zendesk has been innovating in these areas with Answer Bot, Content Cues, suggested macros and more. Now, we are really leaning into expanding AI’s impact even further with accessible AI for companies of all sizes applied through a vertical lens for better personalization.

We combine insights from our trillions of data points (tickets) and then apply a vertical lens to create a unique model for each customer. In fact, 80% of interactions can fall within 20 outcomes, so we identified the typical outcomes for any industry before it gets put to use by a customer. For example, the retail industry sees the same types of questions related to returns, shipping, order status, etc. We have trained our software models on these interactions in order to establish confidence and eliminate any guesswork, allowing customers to focus on bigger picture tasks.

With our latest tools, Zendesk is taking the burden off CX teams through pre-trained machine learning tools that do more faster and with less guesswork involved. These models come ready to use instantly, and continue to learn over time and become customized to each company’s operations by continuously incorporating feedback.

We are excited to introduce this new way of deploying AI at scale and democratizing access to these kinds of solutions. In a time of rising customers’ expectations, companies are facing pressure to get these digital experiences right, and those that can immediately identify intent, extract sentiment and categorize customer segments are better poised to succeed in the long term.

The future of AI…for now
With just a few more years of AI development, we expect there to be many new use cases that will impact the CX industry. Not just customer-facing AI such as front line customer interactions, but also on the business operations side for admins, developers and more.

Here’s a scenario: imagine never struggling to get the right, clean, high quality data to analyze (if you have not personally spent time scrubbing data, this is a huge burden). Looking ahead, we will be able to specify the characteristics we want in data based on things we know about real data, and then generate the perfect set to use very quickly with synthetic data – this will be transformative.

Currently, too many AI models require each business to feed in their own data to entirely train a model. However, for every industry, there are consistent support needs that are easy to identify – like a return for e-commerce. Businesses will no longer need to generate their own data to represent common issues AI can solve. With Zendesk’s Intelligent Triage and Smart Assist, businesses will have a foundational understanding from one solid, robust data set out-of-the-box that all businesses in that industry can use. I believe we will see more like this as AI vendors adjust their offerings.

There might still be some fine tuning a business may want to do for very unique business needs, but the foundational understanding of support concepts will already be built into that AI out of the box. Generating a bit of data to customize further for a unique business will be so much easier with synthetic data.

As AI extends further, there will be more impacts to CX, and Zendesk will continue to evolve our products to ensure our customers benefit from the top AI R&D trends, and as the technology continues to scale dramatically.