Zendesk just launched their Trends Report for 2025 giving insight into where Customer Experience and support is moving towards this year. Similar to last year they got feedback from thousands of customers across the world to gather information on what they think where customer care is going.
Throughout their study they basically found that they could basically cut their customer base in two. On one side you have the CX Trendsetters, companies that have embraced AI and all the benefits brings, and CX Traditionalists who haven't focused on automation and leveraging AI to deploy their solutions across channels.
It'll be no surprise that I fall into the camp of those that embraced AI. (I'd never date to call myself a Trendsetter though.. blue jeans and black t-shirts rarely reflect that attitude). The benefits in transforming knowledge content to fit the customer and channel, the insights you can get from your tickets and customers by letting an AI model loose on all conversations, and the way we can leverage LLMs to (re)write responses to customers are just some of the benefits.
However, I do wonder if the CX Traditionalists as Zendesk calls them are solely comprised of not yet sayers and it's just a matter before they all follow suit, or if theirs a contingent of companies that love their hand crafted human approach and have zero need – and benefit – for AI and will never shift, without seeing any downsides. For each Starbucks there's a local coffee shop that also sells beans online.
Let's dive in!
Driven by AI copilot success, companies are racing to adopt fully autonomous service for next-level interactions
Trend 1
79% of agents believe having AI as a co-pilot supercharges their abilities, enabling them to deliver superior customer service. Learn how AI tools are helping early adopters drive efficiency and enhance customer experience.
I'd be stating the obvious when I say that most of us probably have some AI Agent / Copilot / Chatbot installed on their phone or laptop. ChatGPT, GitHub Copilot, Slite's Ask X are just a few ways in which we can have an assistant sitting next to us to answer questions, generate code, translate text or summaries long emails.
However, there's a difference between the above tools. Where ChatGPT and GitHub Copilot are world models that use the internet's knowledge to provide answers, tools like Zendesk Agent Copilot or Slite's Ask X use, by lack of a better word, company knowledge to provide answers. This gives your team the benefit of AI while staying in control as a company on content, quality and tone of voice.
Hesitancy to integrate AI has opened the door for YoY increase in shadow AI usage by agents
I think this is where we'll see the biggest benefits and battles being fought over the next year. Recently I had a discussion with a customer who saw better results from Microsoft Copilot than Zendesk's Ultimate or Copilot when generating answers for customers and agents' alike. Deeper study showed that the underlying reason was not the tools themselves but the source of the data.
Every scenario where Microsoft "won", the answer came from sources outside of the company. With no guarantee on the correctness of the sources, or the legality of using that information.
I keep coming back to this quote from an interview with Mark Zuckerberg on The Verge:
Yeah, so our view is that there’s actually going to be a lot of these that people talk to you for different things. [...] let’s say you’re a small business and you want to have an AI that can help you interface with customers to do sales and support. You want to be pretty confident that your AI isn’t going to be promoting your competitor’s products, right?"
Embracing AI and leveraging Copilots for me goes hand in hand with controlling your data, managing knowledge and measuring effectiveness. Deploying technology for the sake of technology will get you nowhere.
Consumer confidence in AI agents hinges on how engaging, friendly, and human-like these agents become
Trend 2
Consumer confidence in AI agents hinges on how engaging, friendly, and human-like these agents become. Traits like accuracy and efficiency are now table stakes for any interaction.
Consumer confidence hinges on creating a sense of connection. AI agents that are friendly, engaging, and embody human-like traits are quickly becoming the new expectation: 64% of consumers say these traits make them more trusting of AI.
Deploying a good AI Agent starts with knowledge. If you want to offer a good experience and automate more and more of your customer interactions, having support articles that reflect actual customer problems is the basis for everything. No knowledge leads to no answers leads to sad customers.
However, even when your AI Agent has all the knowledge, things like tone of voice, conversational design, brand and good instructions are key. Similarly one you start offering solutions to your customers, having some basic understanding of who they are is just as important.
There's easy things like welcoming logged in users with their name. There's date and time to play around with to great customers with good morning, good evening and good night. Or wishing them happy holidays or happy birthday.
But, with personalization comes a greater responsibility. If you don't know it's better to omit than to guess, in order to prevent scenarios like this.
A bad personalization often leads to a worse customer interaction than one where you just show all options.
Similarly, if you want to present your AI Agent is human-like and give it personality, you better make sure it gets the same breath of ownership and knowledge as your agents. No one likes a chatty personal bot that has to say no all the time. And on the flip-side, I rather have a dry generative reply than talking to a chatbot that pretend to be human. Uncanny valley is still a thing.
Personal AI assistant popularity prompts companies to prepare for a future where assistants take the lead
Trend 3
Consumers are ready to reclaim their time and relinquish their to-do lists to Personal AI assistants. Rapid technological advancements are ushering in an assistant-first world: 67% of consumers say they’re eager to offload tasks like handling customer service issues to a personal AI assistant.
"Hey Siri, remind me to upgrade my flight to Relate 2025". It's an actual thing I asked Siri earlier today after rethinking my idea that a 9 hour flight with limited leg room was a good plan.
Reading into the 2025 Trends, this prompt did make me think. How far are we actually from "Remind me to upgrade a flight" to "Hey Siri, can you ask Delta to upgrade my flight to a seat with more legroom?"
Looking at the current state of Siri, probably not that close. But with companies like OpenAI, Claude and Anthropic providing Agents that can interact with other platforms on your behalf, I do wonder what our future will be if we can indeed ask our phone to just do something and let us know the result, instead of us handling the conversation.
With Zendesk Copilot the agent side of the conversation is almost automated and mostly supervised by a human agent. Things like checking eligibility for an upgrade, pulling available options from the booking systems and sending the options to the customer or upgrading the flight after approval are all possible and require the agent to just check the provide replies and approve them.
So what's to stop the customer side from evolving that same way?
Hey Thomas?
Yes Siri?
Delta offers two options with the following price, which one do you want?
Go for the first one.
Done.
The one big hurdle here is accountability I think. If I say yes to Siri's output. And an agent says to Copilots input, how can we be sure that that interaction between those two agentic systems made something important go lost in translation?
Voice AI, with its natural language capabilities, helps consumers feel heard across all touchpoints
Trend 4
The key to a satisfying customer experience starts with consumers feeling heard. Learn why 74% of consumers believe AI that understands and responds to their voice would highly improve their overall experience.
Personally I hate, hate calling any customer care team. Inefficient due its synchronous nature (please hold while we find/execute/approve/...) and time consuming.
However, I love using Siri. "Hey Siri play me the rooftop version of Get Back" is so much faster than looking it up in the Music app and selecting kitchen as the airplay destination.
Similarly, my 7 year old is learning about research and Google at school. Last week he wanted to lookup stuff and asked for my laptop. He got tired from typing and searching for keys, so I showed him ChatGPTs voice mode to ask questions. What followed was 30 minutes of the most random questions about the world (and Minecraft) spoken aloud by him, and with written responses from ChatGPT. I really like that channel switching. He could use what he knows best, his voice, and ChatGPT responded in what's best for knowledge: something to read.
I don't think we're there yet for Customer Care though. I've seen what PolyAI and Babelforce can do but there's still a big gap between the setup experience for the generative Zendesk Bot (just a checkbox) and Voice AI for Zendesk Talk. Do note the yet in my message. Never underestimate what happens in 18 months of technological evolution. 18 months ago Zendesk barely had a generative bot. And look at where we're now.
In a world of AI-driven experiences, personalization yields customer loyalty like never before
Trend 5
88% of CX trendsetters consider personalization critical as they incorporate emerging technologies to meet rising customer expectations. Find out how to harness AI tools to drive trust with your customers.
A company delivers on what it promised is what keeps me loyal, in so far loyalty to a brand can be a thing. Personally I rather have an honest "this repair will take a week" than a bad repair in 2 days, or a promise of a quick repair that takes longer.
Similarly just like everyone I do appreciate that personal touch. A barista that knows my drink (and apologizes when the beans are of the natural variety), Apple Support that allows me to select my Mac instead of copying a serial number, my bike repair show that has an actual photo of my bike on their repair page, it's all small touches that give you some assurance that the end product, the service you get will be good.
I also see this with customers for which I deploy AI Agents and Copilot. Those that do it right have their customer journeys mapped out. They know when they need to ask for information, and they know what's the minimum they should ask to identify and have context, while still providing a good and quick experience for the customer.
Things like Messaging authentication, allowing for multiple conversations so both agents and customers can see ticket history or using metadata in conversations are just a few way how Zendesk make this process of identification and personalization possible.
Conclusion
I have to admit, this article started as an intro for my News Roundup for February but quickly outgrew its allowed length.
Trend reports for me are a weird thing. They show you both where the puck is and is going to be, but doesn't really tell you how to get there.
If you ask me what my favorite of the 5 trends is, it's got to be Trend 3. I truly believe we're going to see three kinds of AI Agents being deployed.
World models like ChatGPT or Claude that give you answers to broad topics but with a big risk of inaccuracies. Company models like the ones Zendesk or Slite allow you to deploy. A narrow scope focussing on your company but with the guarantee that the data in the model is yours and correct. And finally personal models that leverage my data as a user. And which one you use depends on your use case. I'll use my personal models when interacting with my phone or laptop. Asking it about my calendar, photos, and documents. I'll use company models at work, or when I interact with another company's chatbot. And I'll use world models when I code or do research.
And if I see one trend being true or relevant for me it's the way these three model types will interact in the future and handover tasks from one context to the other.
Regardless of these high-level trends and insight, personally, I rather have more a more practical approach to get you where you need to be. So here's my 5 technical trends for CX:
- Knowledge is key. Write down your support questions, answers and procedures
- Give ownership. If you want to automate your CX you need to allow your AI Agents, Copilot and human agents to execute without approval. Trust in your process.
- Gather insights. Keep looking for bad experiences and slow downs in your process and fix them. Every time.
- Be an end-user. Try out your systems as an end-user. Ask your mom to test out your AI Agent and don't just blame the tool when something fails. Also check your process
- Omnichannel Agents only work when those channels offer the same solutions. If you offer a conversational AI Agent, make sure that bot can handle all scenarios you'd normally use email or voice for. This does not mean you have to handle the question the same way, but don't block.
