Conversational LLMs (used now by billions of people) have set a new standard for what’s possible; it’s no longer about simple automation of tasks, but instead producing a new standard of work once only achieved by human intervention. One of the industries where AI has had the most immediate impact is within customer experience.
Despite this remarkable improvement in technology, most companies deploying AI in their CX operations are still falling short. The legacy SaaS approach of deliver and deploy doesn’t work anymore. We need a new way of bringing AI to CX, one centered on human attention, to unlock the full potential of LLMs.
The Legacy Software Trap
For generations, Silicon Valley has built software to, 1. Solve a problem, 2. Deliver value, and 3. Scale. This has created the high-margin SaaS model that we know today. The value of Silicon Valley has hinged on this model, and many are trying to apply it now to AI and LLMs.
The issue lies in application; LLMs aren’t traditional productivity software. In fact, LLMs create work successfully that once required human intelligence, empathy, and adaptability. Old SaaS followed rigid, pre-programmed workflows; LLMs excel in dynamic scenarios. They can engage with customers in ways that feel natural, aligning with how people communicate and interact. LLMs aren’t here to refine the same old automation; they’re here to rethink what CX can be.
The Missed Opportunity: LLMs as Creators, Not Tools
LLMs are not software, and CX cannot treat them as such. LLMs can generate conversational experiences that rival those of human agents, adapting to the nuances of customer needs in real-time, creating an experience that stands on its own.
Instead of bolting AI onto existing software frameworks, companies should refactor the entire customer experience operation — software and people together. This shift recognizes that LLMs can perform work once reserved for humans, but that realizing their value requires rethinking how customer operations are designed.
Rather than building from the software up, organizations can design AI-native CX systems that prioritize customer-aligned interactions over rigid automation. The focus should be on use cases where LLMs excel: complex, open-ended interactions that demand creativity, empathy, and context.
Attention Over Automation
Perfecting LLM-based systems looks nothing like traditional software development. The familiar playbook starts to break down when the “code” itself can think. Instead of long build cycles and rigid release plans, you can stand up a working prototype in minutes and refine it through rapid experimentation. Write a prompt, observe the behavior, and adjust. Each iteration teaches you something new.
LLMs don’t self-learn in the old machine-learning sense, but evolve based on human attention and intent. In this model, humans play a crucial role. The best results come when prompts are treated as living expressions of business policy: precise, intentional, and aligned with what the organization stands for. It’s less about writing code and more about teaching a system how to embody judgment, empathy, and brand integrity.
Attention as the Core of AI-Native CX
Attention is the key to unlocking the potential of LLMs, and it never diminishes in importance. Attention manifests in how we improve our systems. For example, when analyzing large volumes of conversation transcripts, those insights are distilled into precise modifications to the business policy that informs our instructions.
When performance falls short, we don’t just feed the data into a machine learning black box. We use human intelligence to evaluate the issue, suggest changes to tone, brand, or constraints, and refine the business policy. This process demands buy-in from the business, as it’s the people who make the strategic decisions that shape outcomes.
When designed well, LLMs reveal friction points humans might overlook. Consider a simple example: a company running a high-touch loyalty program discovered that its “all-inclusive” event invitations didn’t actually include parking, a source of irritation for customers.
Analyzing conversations through an LLM-based system could surface that mismatch, raising the question of whether to change the offer or clarify the messaging. The decision itself still required human judgment, but the AI provides the visibility to make that judgment possible.
That’s the deeper promise of this technology — value in cultivating awareness. LLM systems, when deployed thoughtfully, draw attention to the gaps between intention and experience. They invite leaders to reexamine policies, pricing, and promises and how they play out in real conversations. The technology becomes not just a tool for service delivery, but a mirror for strategic reflection, helping organizations listen closely and respond intelligently to customers.
Re-Imagining CX
For AI to revolutionize CX, we must reimagine the entirety of customer experience from the ground up, using LLMs not as tools but as capable, human-like workers. To do this, startups need to abandon the old SaaS playbook and embrace a new model: one that integrates people and technology, prioritizes attention over automation, and builds from the customer out.
By focusing on attention, both in how we develop our systems and how we empower our clients to make informed decisions, we’re building the next generation of AI-native CX technology. Attention isn’t just all you need — it’s the only way to get it right.
