Beyond UX/UI: Preparing for the AI Agent era

As Artificial Intelligence (AI) continues to advance, a paradigm shift is emerging in how we interact with software. AI agents, capable of autonomously completing tasks on behalf of users, are poised to render traditional user experiences and interfaces obsolete. The focus of software development must pivot from designing intuitive graphical interfaces to creating robust, flexible, and extensible Application Programming Interface (API)s that empower these AI agents to seamlessly interact with applications.

This shift is not merely a technological evolution; it’s a revolution in software engineering that calls for a new discipline – let’s call it Agent Interaction Design (AID). AID represents the framework through which developers can architect software systems that are optimised for AI agent collaboration rather than direct human manipulation.

Why UX/UI alone won’t suffice anymore

For decades, User eXperience (UX) and User Interface (UI) design has focused on facilitating human interaction with technology. However, AI agents bypass traditional interfaces entirely, operating through machine-to-machine communication. Here’s why this shift is both inevitable and essential:

• Efficiency gains: Humans require intuitive designs to reduce friction, but AI agents can process data and perform actions without visual cues, leading to unprecedented speed and accuracy.
• Complexity management: AI agents can navigate dense feature sets and execute intricate workflows without overwhelming human users with unnecessary details.
• Scalability: Traditional UX/UI design scales poorly for users who need highly personalised workflows. AI agents can tailor interactions programmatically via APIs.

Examples: the future is already here

1. Excel vs. Autonomous Data Analysis
   In the traditional model, a user would interact with Excel via its interface, manually entering formulas, formatting cells, and generating charts. However, with tools like OpenAI’s Codex, AI agents can directly manipulate spreadsheets via APIs – automatically generating insights, drafting reports, and even suggesting optimal business decisions. The end-user never touches the spreadsheet itself.
2. Photoshop vs. Autonomous Content Creation
   Photoshop, with its intricate toolset, exemplifies the power of UI-driven workflows. However, cutting-edge generative AI models like Stable Diffusion XL and Adobe Firefly have revolutionised content creation by autonomously producing and manipulating high-quality images based on descriptive prompts. For example, an AI agent could generate marketing assets such as banners or social media graphics by understanding campaign goals, brand guidelines, and target audiences. It could create unique visuals, refine existing images, and deliver fully realised designs ready for deployment, all without the need for direct manual editing in Photoshop.

   Saw this on Twitter (X) – a day after writing this article.

   RIP Photoshop.

   Gemini 2.0 now can edit image by talking to it , it's insane.

   Here’s how it works: pic.twitter.com/gzWdkJYs2q

   — Pearl AI (@heypearlai) January 1, 2025

3. E-commerce websites vs. Conversational Agents
   Why browse e-commerce websites when an AI shopping assistant can query APIs to compare prices, check stock, apply discounts, and place orders? Already, platforms like Amazon Alexa and Google Assistant are moving toward API-driven commerce, fundamentally altering the role of traditional product interfaces.

The imperative for flexible APIs

To enable this transition, APIs must evolve beyond their current capabilities:

• Modularity: APIs must expose granular functionality that AI agents can compose into complex workflows.
• Context awareness: APIs need to support stateful interactions, enabling agents to perform actions based on past exchanges and user preferences.
• Security and governance: With AI agents accessing sensitive systems, robust authentication, authorisation, and audit trails are critical.
• Interoperability: APIs should adhere to standards that enable cross-platform and cross-application agent collaboration.

The death of traditional software?

Will applications like Excel or Photoshop disappear? Not entirely. These tools may remain for specialised users who prefer manual control, but their primary usage will likely shift toward serving as backend engines for AI agents. Traditional interfaces will become secondary to APIs, much like command-line interfaces in today’s GUI-driven world.

Embracing Agent Interaction Design (AID)

To prepare for this AI-driven future, product teams must adopt AID principles:

• Agent-First thinking: Design APIs and systems with AI agent interaction as the primary use case.
• Simplicity over aesthetics: Focus on clean, consistent data structures and predictable behaviour rather than visual polish.
• Human collaboration: Develop APIs that facilitate collaboration between human users and AI agents, such as approval workflows or insight explanations.

A new era of software development

The rise of AI agents heralds the dawn of a new software engineering discipline. By shifting from UX/UI to AID and prioritising API design, developers can unlock a future where humans benefit from the power of software without the burden of traditional interfaces. This transformation won’t just enhance productivity – it will redefine the very nature of how we interact with technology.

As the lines between human-driven and AI-driven interactions blur, it’s clear: the future isn’t in what we see but in what’s seamlessly connected beneath the surface.

Here is a Google Gemini podcast on this article.

First dropped: January 1, 2025 | Last modified: January 02, 2025