Context
Credits
This Research was done at Salesforce in collaboration with my colleague Ralf Josef, Strategic Innovation Designer.
July - September 2025
Traditional chatbots often frustrate users by offering limited, repetitive responses and failing to grasp context or nuance.
Users frequently encounter dead ends and are forced into tedious loops of irrelevant conversation, which undermines trust and wastes valuable time.
In contrast, advanced AI Agents understand intent, adapt to conversation flow, and deliver personalized interactions that genuinely solve user problems.
While chatbots are rigid and superficial, AI Agents excel by leveraging deeper intelligence to create meaningful, satisfying user experiences.
Still, a long way to go, despite growing interest in AI, many customers still confuse AI Agents with chatbots. They expect the same scripted, shallow interactions and assume every interface is just another glorified FAQ.
This confusion leads to underuse, skepticism, and missed opportunities. Here are the core principles shaping CX in this emerging agentic era.
Quit hiding the agent in the dusty bottom-right corner, where users instinctively ignore it — a relic of every clumsy chatbot they’ve ever clicked and regretted. That corner placement signals “support widget,” not “intelligent partner.”
Instead, bring the agent into the light: position it at the center-bottom or weave it directly into the main navigation, where its presence feels intentional, visible, and central to the experience. This shift isn’t cosmetic — it’s symbolic. It tells users the agent is part of the product, not a bolt-on accessory. Likewise, burying it deep in a site map or behind optional menus hides its potential and slows down adoption.
The faster people see it, tap it, and test it, the faster it learns, adapts, and proves its worth. Early and frequent interaction is the oxygen that helps an agent evolve from an invisible widget into a trusted co-pilot — visible, capable, and central to your brand’s new way of engaging customers.
AI Agents should break free from the conventional chatbot look — the speech bubble, the avatar in the corner, the “How can I help?” cliché. Those visuals carry baggage from years of limited, scripted interactions that failed to deliver real value.
To signal a true leap forward, the design must evolve.
Think dynamic forms, fluid motion and living interfaces that express intelligence and adaptability rather than containment. Modern design codes — minimal, reactive, and context-aware — help communicate that this is not a chatbot waiting for input, but an autonomous system capable of perceiving, reasoning, and acting.
The moment users see it, they should feel that difference: not another assistant to chat with, but an intelligent presence woven into the experience — one that moves with purpose, not prompts.
Chatbots wait — agents act. Unlike their passive predecessors, agents don’t linger for instructions; they read the room. By interpreting individual needs and subtle behavioral cues, they know when to step in and why.
Their goal isn’t to interrupt or annoy, but to assist with precision — surfacing the right insight, offer, or action exactly when it’s most useful. This timing turns engagement into relevance, transforming what once felt like an intrusion into a moment of genuine help.
The result: users feel understood, not managed — guided, not prompted.
Chatbots speak in text — and only text.
AI Agents, on the other hand, are multimodal communicators. They can express ideas through images, links, videos, data tables, or even voice, choosing the most effective medium for the moment. This versatility transforms communication from a linear exchange into a richer, more intuitive dialogue. Instead of forcing users to read long paragraphs, agents can show, illustrate, or demonstrate — making complex information effortless to grasp.
The result is not just better comprehension, but a more human-like interaction that feels natural, adaptive, generous and alive.
A chatbot is chained to a website — and that’s a problem when your users rarely visit it. Today, people live on WhatsApp, Instagram, and every platform in between, expecting brands to meet them where they already are.
AI Agents break those walls. They can orchestrate conversations seamlessly across channels, maintaining the same intelligence, tone, and context whether it’s text, image, or voice. This continuity builds familiarity and trust: users don’t have to “find” the agent — it finds them, adapting effortlessly to their preferred mode of interaction.
The result is a truly omnipresent experience — consistent, personal, and natural — where engagement feels like a dialogue, not a destination.
Deploying an agent to replace simple interactions that can otherwise be accomplished with 2 clicks will cause more, not less, frustration with the end user.
A good (first) use case should significantly simplify the experience for the user, therefor demonstrating value - as well as be deployable quickly, showing speed to value.
We’re entering a new paradigm — one where AI doesn’t just respond, it acts. Yet, to truly unlock the promise of Agentic AI, we must unlearn some of our old reflexes. Wrapping an autonomous, decision-making agent in the familiar skin of a chatbot might feel safe, but it actually limits how users perceive and engage with it. It keeps them anchored in the past — expecting conversation, not collaboration; replies, not results.
This is perfectly natural: we design and use technologies based on what we already know. But progress requires a gentle rupture with habit. To help customers grasp the real value of Agentic AI, we need to present it through new principles — interfaces, metaphors, and experiences that express initiative, proactivity, and partnership.
Only by reframing how it looks, feels, and behaves can we help people see it for what it truly is: not another assistant waiting for input, but an intelligent counterpart moving things forward on their behalf.
