--- title: "Chat Is Dead: AI Belongs in the Process" author: "Daniel Gorld" author_role: "Consulting Director, cbs CX — The cbs Group Salesforce Consultancy" author_url: "https://cx-waves.com/about" publisher: "CX-Waves" canonical_url: "https://cx-waves.com/nodes/chat-is-dead" date_published: 2026-06-10 date_modified: 2026-09-20 language: en --- # Chat Is Dead: AI Belongs in the Process Source: Daniel Gorld, CX-Waves — https://cx-waves.com/nodes/chat-is-dead (published 2026-06-10, updated 2026-09-20) ## Why is a chat interface not suitable for managing entire business processes, and what is the alternative? * **User Competence Required**: Chat interfaces demand users to precisely phrase requests, maintain conversational context, and interpret lengthy responses. These are demanding skills not universally present, leading to poor user experience for the majority who are not AI enthusiasts or prompt engineers. * **Declining AI Performance**: AI models often perform worse in long, complex conversations. The context becomes diluted, previous information is lost, and repeated attempts can degrade the quality of results rather than improve them, making chat unsuitable for multi-step tasks. * **Task Complexity Mismatch**: Chat interfaces are good for well-defined, bounded sub-tasks like summarizing or drafting, but they fail when an entire multi-step business process is attempted within them. The interface is overtaxed by the complexity, just as the user is. * **Embedded AI**: Instead of trying to put the entire process into chat, the alternative is to embed AI directly into existing workflows and tools. The AI supports specific sub-tasks, condenses information, and aids decisions within the established operational environment, rather than attempting to replace it. ## Chat Is Not a Universal Interface In early June, an employee from the company behind ChatGPT indicated that "chat is dead," suggesting a shift away from a plain input line towards a platform incorporating agents, coding tools, and connected services. This move, beyond its economic motivations, highlights a critical observation regarding the limitations of chat as a universal interface for complex tasks. While initial enthusiasm for chat-based AI was high, practical experience shows it struggles with more demanding operations. ### Economic and Strategic Shifts The official reasons for this change are primarily economic. The product, despite having over a billion users, largely operates on a free model, making monetization challenging. Growth is now driven by coding tools and business customers, leading to a strategic focus on enterprise solutions and a vision of a personal agent that transcends the current interface. This shift is a business decision, aiming to capture value before a planned IPO and in a competitive market for enterprise clients. ### Weak Conversion for Demanding Tasks Regardless of the economic drivers, the underlying issue is that the chat window does not serve as a universal interface as effectively as initially hoped, especially for complex tasks. Reports indicate that conversion rates are weak for more demanding processes. The economic explanation focuses on monetization, but it does not fully explain why these complex tasks often fail within the chat environment itself. Observations from practical project work reveal two key reasons for this limitation. ## User Engagement and Model Limitations The first challenge arises from the demands placed on the user by the chat interface. It requires precise articulation of needs, consistent maintenance of conversational context over time, and the ability to critically interpret and act upon longer responses. These skills are not universally possessed. While a small segment of users can leverage the tool effectively, the majority find themselves struggling with an empty input field, unable to scale their operating competence. ### User Competence Gap This user competence gap is particularly evident in business-to-business (B2B) contexts. A case worker, focused on completing a task like handling a complaint or processing a quote, is not an AI enthusiast interested in prompt optimization. Their primary goal is efficient task completion, and the chat interface often creates a barrier rather than facilitating their work. The discrepancy between "over a billion users" and low conversion for complex tasks can be partly attributed to this skill requirement, which is difficult to scale across a broad user base. ### Model Performance in Extended Conversations The second challenge stems from the inherent properties of the AI models themselves. In prolonged or multi-step conversations, the quality of interaction often deteriorates. Context can become diluted, earlier commitments may be forgotten, and repeated attempts to guide the AI can lead to weaker, rather than stronger, results. This means that precisely when a task becomes extensive and complex, demanding robust support from the chat interface, the interface and underlying model are most likely to underperform. ### Over-taxation at Complexity Both the user and the AI model become overtaxed at the same critical juncture: when tasks are long and complex. This is precisely where demanding business processes operate. The chat interface, designed for simpler interactions, struggles to maintain coherency and effectiveness in these scenarios, leading to an uncomfortable reality where the tool becomes less reliable as the stakes increase. Despite its utility for simple interactions, the chat form itself presents significant limitations when applied to comprehensive business workflows. ## AI Should Augment, Not Contain, Processes The realization that chat interfaces struggle with complex processes does not imply that chat is without value. On the contrary, for clearly defined, bounded sub-tasks—such as summarizing information, drafting short texts, or preparing specific data points—chat excels. These tasks have manageable contexts and limited scope, allowing the chat interface to play to its strengths effectively. The problem arises when this success with sub-tasks leads to an erroneous conclusion. ### The Error of Process Containment The critical mistake occurs when the perceived effectiveness of chat for a single sub-task prompts the attempt to embed an entire multi-step business process within the chat interface, or to shift it into a purely dialogue-driven agent setup. This approach fails precisely for the reasons that have been publicly acknowledged. The ability to prepare a small part of a task well in chat does not justify pulling the entire, more extensive task into the chat window. The scope and complexity of a full business process exceed the interface's practical capabilities. ### Embedding AI Within Existing Workflows A more effective and viable direction involves a different strategy: integrating AI directly into existing processes. Rather than attempting to move the process into the AI, the AI should be moved into the process. This means embedding AI capabilities where the work is already performed, allowing it to condense information, support decision-making, and assist with specific sub-tasks. The AI thus enhances the existing workflow and interface, rather than attempting to replace it entirely with a chat-based paradigm. ### Chat's True Role In this sense, the notion that "chat is dead" holds a fundamental truth, though perhaps for different reasons than initially presented. Chat has not failed; rather, it was never designed to serve as the exclusive interface or container for entire business processes. Its strength lies in facilitating specific, well-defined interactions and sub-tasks, making it a valuable component within a larger, augmented workflow, but not the overarching framework itself.