The first mobile application user interfaces were often scaled-down versions of what was already available on the web. Then, user experience (UX) designers recognized that the different smartphone form factor created new business opportunities and greater utility compared to what people were doing on their desktops. UX designers created mobile-first experiences tailored to the job to be done and other design thinking principles. The underlying agile development practices, along with the emergence of app stores, paved the way for explosive growth in smartphones and mobile applications.
Today’s AI experiences seem to be following a similar path, with basic, sometimes bolted-on user experiences.
- First-gen chatbots appeared as pop-ups with text entry-and-response user interfaces (UIs) overlaid on the application’s screens.
- The primary UI for large language models (LLMs) is often a text box that accepts a prompt followed by a response that includes text and other media.
- Early AI agents were embedded in workflows, allowing users to prompt for information rather than point and click.
In a recent Coffee With Digital Trailblazers LinkedIn Live event, we discussed AI-first UX and planning for the evolution of customer journeys. Joanne Friedman, CEO of ReilAI, remarked, “A UX must be tailored to the persona and the role of the human being. Intent, perspective, authority to make decisions, and even judgment are elements of the human context that surround a role. That also means that context is now tied to security. What a person can and can’t see, or what they have access to now, makes it a design consideration and one that agentic AI is well-suited to enable.”
What is AI-first UX?
I expect that the evolution of how SaaS embeds AI agents will provide a model for how AI-first UX should look and behave. You can also see how AI is embedded in ecommerce experiences by selecting “I’m looking for ideas” in the Marriott Homes and Villas AI search experience, reviewing Perplexity Shopping, or trying out Rufus, Amazon’s new shopping AI.
“AI-first UX is the collapse of the app sprawl that’s defined enterprise software for the last decade,” says Vishal Sood, president of R&D at Typeface. “We’re moving from users bouncing between disconnected tools to orchestrated systems where agents carry context across workflows, and canvases and editors let humans steer the output. The winners will be hybrid environments that blend conversational interfaces, visual workspaces, and agentic orchestration into a single coherent experience.”
SaaS sprawl is a real issue. Large enterprises average more than 600 SaaS applications and spend $280 million annually on SaaS. Some SaaS solutions are embedding data fabrics and zero-ETL capabilities, enabling their AI agents to use data outside their environments. The results are not just a shift from clicks to conversations; it’s an evolution toward integrated experiences.
Hector Ouilhet Olmos, vice president of design for AWS Solutions at Amazon Web Services, says agentic AI demands a fundamental shift from the “desktop metaphor” to designing interfaces that mimic human collaboration dynamics rather than physical objects. “Instead of forcing fluid, conversational intelligence into rigid buttons and chat panels, we must dismantle traditional user interfaces and build human ones. These will translate millennia-old human collaboration patterns like negotiation, interruption, and escalation into native digital experiences where AI functions as a teammate, rather than a tool,” says Olmos.
Many of today’s traditional user experiences can be deconstructed into forms, reporting dashboards, and workflows. Let’s consider how AI-first UX may evolve away from these structures.
Conversations and interviews replace forms
Will entering data into forms and using them to make edits become obsolete? AI-first UX will provide alternatives when people must enter information into systems of record to get their work done.
“Instead of navigating screens or filling out static forms, users simply describe what they want to accomplish,” says Chris Mayor, vice president of architecture at Coupa. “Forms evolve into adaptive conversations that prefill known information and dynamically gather the rest. This transforms enterprise software from systems of record into systems of action.”
One UX metaphor, based on conversations, works when the user has a job in mind. A second occurs in reverse, where an AI agent prompts users for recommended actions or decisions. “Every platform shift begins by replicating the old model, but real transformation happens when workflows are redesigned,” says Preetpal Singh, group managing director and global head of product and platform engineering at Xebia. “In this model, forms evolve into adaptive interviews that prefill known data, ask contextual follow-up questions, and accept natural language or images, while still preserving clarity and compliance.”
AIs generate reports and dashboards
Many IT departments used to have reporting functions with teams developing dashboards and writing custom SQL queries to retrieve data. Much of that work shifted out of IT, as many CIOs promoted citizen data science, established data visualization best practices, and deployed advanced analytical solutions to help departments build dashboards and move away from manual spreadsheets.
As organizations deployed more self-service business intelligence tools, they adopted practices for applying design thinking in data science and integrating analytics into workflows. But there was a significant challenge: Designers, data scientists, and engineers had to anticipate users’ questions about customers, finances, and other business functions and then implement data visualizations to answer them.
AI-first user experiences will turn reporting and dashboarding around. Instead of people generating relevant data visualizations, AIs will.
“The definition of ‘user-friendly’ has changed forever, and ‘AI over UI’ has become a new calling when it comes to building enterprise products, says Maksim Ovsyannikov, chief product officer at SugarAI (formerly SugarCRM). “This emphasis on conversational user experience focuses on a user’s ability to ask questions and compose prompts rather than their ability to understand workflow and build reports. Users now simply ask for the report or insight they need instead of spending hours building a report or a dashboard that becomes stale and outdated in a matter of days.”
Singh of Xebia adds, “Reporting [is shifting] from static dashboards to narrative copilots that explain what changed, why it matters, and what actions to consider next, combining visual metrics with interpretation and foresight.”
Workflows become agentic AI collaborations
Simple workflows live in one system of record and connect people through a linear process. Examples include editing website content in a content management system, recording new information about a prospect in a customer relationship management program, or performing basic accounting functions in an enterprise resource planning system. More complex workflows are non-linear, involve multiple departments performing different responsibilities, and require integrating several systems of record. Examples include employee onboarding, quote-to-cash processes, and contract management.
Now imagine all the underlying systems are API-enabled, have AI agents in place to perform basic functions, are integrated with Model Context Protocol servers, and have an AI orchestration platform to facilitate work. What was a workflow becomes a collaboration between people and AI agents that can perform multiple steps through a single user interface.
“An AI-first UX replaces rigid, screen-driven workflows with intent-driven interaction, where users query the system, and AI agents orchestrate the underlying processes,” says Avi Greenfield, vice president of digital enterprise products at Quadient.
A typical employee onboarding process involves steps performed by people in HR, finance, and IT across many systems. In an agentic AI experience, HR initiates the process, and work is coordinated through AI agents, with decisions and approvals sent to the appropriate managers.
“Agentic workflows begin to resemble coordinated teamwork, where AI agents execute multistep processes across systems, surface their reasoning, and escalate to humans when judgment is required,” says Singh of Xebia. “The goal is augmentation over replacement and choosing the right interaction model at the right moment, grounded in strong UX discipline, transparency, and trust.”
Enterprise platforms are enabling the transformation from the workflow they support to agentic AI experiences. For example, Workday recently announced Sana with a new AI user interface and over 300 skills to automate many HR and finance workflows. “Most AI projects today live in pilots and browser tabs. They look impressive in demos, but they don’t change how work actually gets done,” said Gerrit Kazmaier, president of product and technology at Workday. Another example is Anthropic’s release of Claude Cowork with plug-ins for legal, marketing, and other business functions. Example workflows include contract reviews, product documentation, and financial journal entries.
How AI-first UX impacts development
The opportunity to create scalable mobile user experiences drove devops teams to build APIs, use low-code mobile development platforms, and expand continuous testing to cover mobile applications. Vibe coding and other code-generation tools are just the start of what will support the development and testing of agentic AI experiences.
“AI-first UX is moving beyond chatbots into embedded, ambient intelligence where AI becomes an invisible layer that anticipates customer needs and translates intent into action,” says Amit Patel, senior vice president of consulting services at Consulting Solutions. “This requires a shift from feature-driven design to intent-driven experiences, supported by strong data foundations, APIs, and governance so AI can act responsibly. The companies that win will treat AI not as a bolt-on assistant, but as a core experience layer that reduces friction, personalizes at scale, and builds trust through measurable value.”
In larger companies, developing AI experiences will require orchestrating work across multiple AI agents, both from SaaS providers and from internally developed systems. Developers will need to update observability standards, automate AI agent testing, define release-ready criteria, review multiagent frameworks, and consider orchestration platforms.
Andrew Filev, CEO and founder of Zencoder, says, “Just as platforms exist to coordinate human teams, AI demands a new orchestration layer: interfaces designed not to do the work, but to visualize, steer, and direct outcomes across multiple agents.”
We’re only in the early stages of how people and AI agents will collaborate, so expect to see evolutions in platforms and capabilities. Looking to the future, expect that voice, augmented reality/virtual reality, and other physical AI will further transform how we develop AI-first user experiences.
Go to Source
Author:
