CX 2028

When ChatGPT was released, the whole world started talking about what we might achieve with such technology. One of the first ideas that came to mind was customer support. It’s a classic case: written communication that mostly relies on knowledge the representatives have and the customers don’t.

Companies have spent a lot of resources maintaining this knowledge and hiring and training those agents. The moment we can start communicating with customers in their natural language, this challenge can be solved.

At first, AI showed up as different co-pilots to build trust and confidence. Then these solutions started interacting directly with real users and began reducing the need for repetitive “tier 1” cases.

Basically, everything that could once be solved with a simple templated reply.

After replacing the old chatbots that only provide you basic chips style selection , the next in line is the search bars which mostly turned into “Ask” bars. Organizations are implementing more and more knowledge engines across multiple departments, channels, and interfaces. 

People’s patience for searching for answers is shrinking -> look at the old google experience compared to perplexity, chatgpt etc..

And this directly affects customer expectations. Their patience with traditional help centers is fading.

The use of AI becames the standard for  “self service” offering for external users or internal teams, and specifically Customer Experience (CX) teams are now setting up AI-powered tools to deflect* tickets and provide direct, instant solutions.

^{*deflecting is a counterintuative term which is affecting Support orgs negatively. Deflecting rarely mean ‘resolving’, which is the main goal of a Customer Support team.}

^‍

For years, automations were built on workflows, IFTTs and rule-based engines - support operations, chatbots, and more.

The limits of this approach are clear. While strict and predictable, it requires heavy lifting to set up, and the real challenge is maintaining and updating it as operations and procedures evolve.

So far, it’s been hard for LLMs to match that predictability, precision, and consistency to simply replace those flows. On top of that, the tools for setting up and understanding what an LLM is doing or “thinking” still lack enough context.

For a long time, research has been exploring neurosymbolic methods - technology that combines the strengths of LLMs with more structure and control.

We believe companies will find ways to overcome LLM inconsistencies, allowing them to move away from strict workflows without losing the predictability and structure sometimes needed. This will close the chapter of rule-based setups and shift toward policy-based approaches - more like how we train and manage people - focusing on true resolutions rather than simple deflection.

For the first time, AI will start taking on most of the actual workload in support operations - not just handling the biggest buckets of simple chats or tickets, but covering tasks across all domains.

This shift will lead companies to stop seeing AI agents as simple tools and start treating them as part of the team. A new type of team will emerge: Hybrid Teams - made up of AI agents and human agents working side by side. These AI agents will act proactively, suggesting tickets, asking questions to learn more about their roles, and continuously improving.

We believe this change will also bring a wave of personal assistants. AI will not only represent companies in support inquiries but could become an additional service the company provides. On the other side, customers may begin using their own AI assistants to handle interactions on their behalf. This will create an entirely new reality where AI entities communicate directly with one another.

After shifting most of the workload to AI teammates, companies will realize that the current standard of CX is no longer enough. With more resources freed up, they will aim to deliver best-in-class service. To achieve this, CX teams will stop waiting for customers to open tickets. In fact, they will move beyond “answering tickets” and start focusing on Signals.

A Signal can be any trigger you define. It’s the foundation of true proactiveness. Signals might include signs of user frustration, a backend error log, a user idling with items in their shopping cart, or any number of meaningful indicators. By recognizing these signals, you can empower your AI to take timely and relevant action before the user reaches out.

This shift will also redefine how success is measured. Instead of focusing on traditional metrics like FTR (First Time Resolution) and TTR (Time to Resolution), teams will track Proactive Resolution Rate. This measures the percentage of relevant signals detected and addressed before they impact the user experience. It highlights how effectively your support operation can anticipate and resolve issues before users ever need to reach out.

To unlock this level of service, AI agents will need broader context - not only through third-party integrations, but also through deep product understanding and real-time visibility into user sessions.

All of this will lead to a completely different day-to-day for CX teams and trigger a real market shift. Companies will stop hiring for traditional Customer Support or Customer Success roles. Instead, we’ll start seeing new roles emerge, such as Customer Expert and Customer Engineer, and these roles will only continue to grow.

For the first time, customer-facing teams will become company drivers - true builders rather than just responders.

These departments will merge more closely with engineering and product teams, gaining the power and freedom to design and improve how “bad flows” in the product are handled.

Just as we see companies dogfooding their own products today, we will soon see AI agents that understand customers so deeply they will become the first users, beta testers, and A/B testing audience inside companies.

This means we will start asking these AI entities how we should improve our product, based on their complete understanding of customer behavior, needs, and goals.

As soon as AI agents become our beta users, they will likely also become part of our actual audience. Support agents will find themselves helping not only customer agents but also internal AI agents.

Just as companies today have “ask-support” Slack channels where anyone can ask questions and learn, we will see similar interactions with Sales AI agents, Marketing AI agents, and many more.

And of course, some customers will no longer reach out to support directly. Instead, their personal AI assistants will contact us on their behalf and expect full support.

The function we know today as support has always been focused on scaling. But in the past three years, the world has changed completely. There is no longer “support” as we knew it - Customer Engineers and Customer Experts can now achieve infinite scale. From now on, the paradigm shifts: there is no connection between the number of customers and the number of customer experts, because the real “hiring” happens on the AI teammates’ side.

This is the final piece of the puzzle - the true potential that LLMs are unlocking.