Ideas lab
Designing AI agents is becoming a new workplace superpower
Article
3 Oct 2025
By Ismail Amla, Senior Vice President of Kyndryl Consult
You are chatting with ChatGPT about potentially moving to London and spending a six-month trial period there. Mid-conversation, it surfaces a map showing rental properties in your price range. No app switching, no new tabs, just an interactive map that appears in your chat. You ask about neighborhoods, and it overlays commute times. You mention needing flights, and suddenly you are booking with Expedia, still in the same conversation. A nice restaurant? ChatGPT connects to OpenTable and recommends the most highly rated restaurant near your hotel. You book it for a celebratory dinner with your spouse.
This is not demoware. As of October 6, 2025, ChatGPT became an application platform where services like Booking.com, Spotify, Figma, and Coursera live directly inside conversations. With 800 million weekly users, OpenAI just standardized a new way of interacting with software. You speak your intent, and the right tools appear.
For enterprise applications, this is the iPhone moment all over again. When that device launched in 2007, it introduced consumer expectations for elegant, intuitive interfaces that eventually forced enterprise software to modernize. ChatGPT is doing the same thing, but the change will happen faster and cut deeper. The iPhone changed how we accessed software. Now, tools like ChatGPT are starting to change how we think about accessing it at all.
In the case of OpenAI’s new Apps SDK, something clever is happening. It utilizes the Model Context Protocol, now adopted by both OpenAI and Anthropic, to enable applications to render interactive interfaces directly in chat while maintaining full access to their backend systems. Users are not getting a stripped-down chat experience. They are getting the actual Zillow map, the actual Coursera video player, and the actual Figma canvas, but they are invoked conversationally and contextually aware of what they are trying to accomplish.
More importantly, they are getting them without thinking about them. Zillow is surfaced when you discuss moving. Spotify is activated when you mention a party. The cognitive load of “which app do I need for this?" disappears. A workforce trained on this interaction model will reject enterprise software that forces them back into the old world of hunting for the right module, clicking through navigation trees, and filling out forms in a predetermined sequence.
The gap between consumer AI experiences and enterprise tools grows daily. While advanced chatbot users casually summon Figma to turn a sketch into a diagram or use Coursera to learn while discussing concepts, enterprise workers still log into Salesforce to manually update opportunity stages through dropdown menus. They navigate Workday’s interface to request vacation days. They switch between Concur, Slack, and email to approve expense reports.
Natural language is no longer a novelty feature. It is becoming the expected input method. Consumer behavior has always previewed trends in professional software. The consumerization of IT movement, which followed the iPhone, proved this. History is repeating itself.
Enterprises must get ahead of this or lose out on the next great UX shift.
For enterprise applications, the question becomes existential. Will your users interact with your application directly, or will tools like ChatGPT become the intermediary? Will employees say, “Open Salesforce and update this opportunity,” or will they say, “update this opportunity,” and let the AI tools handle which system that touches?
Companies that move quickly to adopt conversational, agentic interfaces for their applications will maintain control over their user relationships and data. Those who wait risk being relegated to backend infrastructure. Present, but invisible, in a ChatGPT-mediated workflow.
This transformation from GUI-first to conversation-first is happening in real time with real users. ChatGPT has 800 million weekly users learning right now that software should work like a conversation, not like navigation. They experience applications that appear when needed, understand context, and accomplish tasks through intent rather than instruction.
Those expectations are not staying in their personal lives. Enterprise applications have perhaps 18 months before this becomes table stakes. Every day employees use tools like ChatGPT’s app platform, their tolerance for complexity in enterprise software shrinks. They are already asking why internal tools cannot work the same way. What’s more, they are creating workflows using consumer and business-lite AI tools to build their own agentic flows and app integrations outside of security constraints.
The agentic future is here and now. Consumer expectations are moving. Enterprises must get ahead of this or risk losing out on the next great UX shift — and potentially losing employees and customers who no longer have the patience for clicking through menus and moving cursors in outdated interfaces.
