From Generalists to Purpose-Built: AI is Coming for Your Favourite Interfaces
AI tools are becoming more specialised, rewriting the rules of UI design as they go.
Software used to be a blank slate—flexible but demanding. Over time, software specialised, making tasks easier by structuring workflows around user needs. The more niche the focus, the more effort is saved for users.
AI is now taking the same journey—and in the process, it will rewrite the rules of digital interaction. The interfaces we rely on today? They might not survive the shift.
Specialised AI tools aren’t just automating work—they’re reshaping the very way we interact with software. The interfaces we’ve relied on for decades are about to change in ways we can’t yet predict.
The history of technology is a story of refinement: from open-ended tools that require expertise to specialised solutions that work out of the box. The most powerful technologies don’t stay open-ended forever—they evolve into something more focused, structured, and useful.
The first personal computers greeted you with a blinking cursor and a silent expectation: Tell me what to do. Then came graphical interfaces, drop-down menus, toolbars—a structure to hold all that potential. Over time, software specialised, adapting itself to our needs rather than demanding we adapt to it.
AI is the next frontier, and it's following the same arc. Right now, AI still feels general-purpose, a blank canvas waiting for the right prompt. But that won’t last. The best AI tools won’t be the ones that do everything—they’ll be the ones that make smart decisions about what users actually need.
The Shape of a Tool Decides How You Use It
There’s something satisfying about picking up the right tool for the job. A chef’s knife, sharp and balanced, glides through an onion with precision. A Swiss Army knife, on the other hand, has a blade too small, a handle too awkward—but it’ll get the job done in a pinch. The difference? A chef’s knife is purpose-built. A Swiss Army knife is prepared for anything but perfect for nothing.
Software has followed the same trajectory. In the early days, digital tools were like Swiss Army knives—broadly capable but demanding expertise. Whether it was early computing interfaces that required command-line fluency or the first wave of professional software suites filled with cryptic menus and configuration panels, users had to do the work of shaping these tools to their needs.
Over time, software evolved toward more structured, more opinionated tools—ones that made assumptions about what users wanted and streamlined the path to getting there. Photoshop became Canva. Spreadsheets became budgeting apps. Adobe Premier became Descript.
And now AI is making the same transition. We began with general-purpose AI models, capable of doing just about anything—but only if you knew the right prompts, the right tweaks, the right mental models to coax them into usefulness. Now, we’re seeing something different: AI tools that are deeply specialised, tightly integrated, and designed for specific jobs.
This isn’t a step backward. It’s progress. Specialisation isn’t about limiting possibility—it’s about making AI more useful, more immediate, and more aligned with how people actually work.
The more niche a tool becomes, the more it does for you. And that’s a good thing.
The Benefits of Opinionated AI (And Why It’s a Net Positive)
When software makes assumptions about what users need, it isn’t limiting them—it’s freeing them from unnecessary decisions. Every tool has an opinion, whether implicit or explicit, about how it should be used. The more specialised a tool becomes, the more useful it is, because it does more of the thinking for you.
We’ve seen this pattern before.
Early UX design was all about control. Customisable interfaces, endless settings menus, fine-grained adjustments—users had to shape the tool themselves.
Then came opinionated design. Apple’s iOS restricted certain customisations to optimise for usability. Google Search simplified interfaces to prioritise speed. Figma stripped away clutter to focus on collaboration.
These tools weren’t less powerful—they were more useful because they were designed around clear, intentional constraints.
AI is now undergoing the same transformation.
Descript—a video and audio editing tool has eliminated the traditional timeline interface entirely. Instead of scrubbing through clips, you edit video like you would a Word document. It’s opinionated: it assumes that users want to edit based on spoken content, not raw footage. This removes flexibility—but for many users, it also removes friction.
Look at Flora—a creative AI tool that rejects the “one-prompt, one-output” model. Instead of treating AI like a vending machine for assets, Flora structures the creative process into an evolving workflow, giving artists more control over iterations. It assumes creativity is non-linear—so it builds for that.
These tools don’t just use AI. They shape it. They provide scaffolding—structured pathways that remove the burden of knowing how to interact with AI, letting users focus on what they want to accomplish.
The Impact of Opinionated AI Tools
We’re not heading toward a future where AI takes away choice. We’re heading toward a future where AI tools compete by offering different types of structure, different assumptions, and different ways of working.
Instead of one giant, general-purpose AI system, we’ll have an ecosystem of specialised AI tools, each optimised for different industries, workflows, and user preferences.
More diversity in AI solutions: Instead of one-size-fits-all AI, we get tools tailored to different user needs.
Less cognitive load: No need to engineer the perfect prompt—tools work with users, not against them.
More automation, less effort: The best AI tools anticipate what users want and remove unnecessary steps.
Higher quality standards: More competition means better-designed tools, better interfaces, and more refined experiences.
As long as AI tools remain completely open-ended, they require users to adapt to them—learning prompting strategies, troubleshooting quirks, and figuring out what works through trial and error.
But as AI becomes more structured, more opinionated, and more specialised, it shifts the burden away from the user. Now, the AI adapts to the user instead.
That’s not a loss of power. That’s progress.
The AI Abstraction Spectrum: More Specialisation, More Utility
All software sits somewhere on a spectrum between flexibility and usability. The more a tool can do, the more effort is required to use it effectively. The more it assumes about what you need, the easier it is to use—but at the cost of some flexibility.
AI is no different. We’re seeing the rapid emergence of layers of abstraction that move AI from being a raw, general-purpose capability to something more structured, guided, and purpose-built for specific jobs.
Let’s map it out:
Base AI Models (Maximum Flexibility, Maximum Effort)
Examples: OpenAI’s or Anthropic’s API, Hugging Face models
Raw power, no guardrails. These are general-purpose AI systems that can do just about anything—but only if you know how to prompt them correctly, fine-tune parameters, and manually validate outputs.
Who benefits? AI researchers, developers, and power users who are comfortable shaping the tool themselves.
General-Purpose AI Tools (Flexible, but Still Requires Effort)
Examples: ChatGPT, MidJourney
A little easier, but still requires user effort. These tools allow more natural interactions than base models, but they still demand some level of user expertise. You need to know how to phrase a prompt, iterate on responses, and experiment to get the best results.
Who benefits? Early adopters, professionals, and those willing to experiment with AI’s quirks.
Structured AI Tools (Guided Workflows for Common Tasks)
Examples: Jasper AI, Flora
More structure, less friction. Instead of making users figure out how to get good results, these tools offer templates, workflows, and constraints that guide users toward effective outcomes. You don’t have to “prompt” an AI model—you just select an option and refine from there.
Who benefits? Marketers, designers, and professionals who want AI assistance without the learning curve.
Domain-Specific AI (Deeply Integrated into an Industry or Workflow)
Examples: Descript, Otto Finance, Canva’s Magic Tools
AI disappears into the workflow. These tools don’t feel like AI interfaces at all. Otto Finance, for example, doesn’t expose AI as a feature—it is the product, working behind the scenes to summarise, offer insights and recommendations, and automate data entry. The AI is embedded seamlessly, abstracting away complexity so users don’t have to think about it.
Who benefits? People who want results, not an AI sandbox.
Passive AI (No Interaction Required—It Just Works)
Examples: Spotify radio, spam filters, Grammarly’s passive corrections
Passive AI operates silently in the background, enhancing experiences without requiring direct interaction. Unlike AI tools that demand commands, prompts, or manual adjustments, Passive AI detects patterns, anticipates needs, and takes action automatically.
Who benefits? Everyone. Whether it’s preventing email overload, auto-adjusting content, or fine-tuning recommendations, Passive AI reduces cognitive load and makes everyday tasks smoother and smarter.
Why This Matters
Each level in this spectrum represents a trade-off between flexibility and usability—but crucially, it also represents a shift toward making AI more useful for more people.
Not everyone wants to prompt a base model. Not everyone wants infinite control. Most people just want a tool that understands their needs and helps them get the job done.
And that’s exactly where AI is headed.
The Future: An AI Ecosystem of Purpose-Built Tools
If history is any indication, we’re not moving toward a single, all-powerful AI that does everything. We’re moving toward an ecosystem of specialised, deeply integrated AI tools—each designed to fit naturally into how people already work.
This is a good thing.
The early internet was a chaotic sprawl of general-purpose websites. Then came focused platforms—Amazon for shopping, LinkedIn for networking, Spotify for music. Each took a massive, unfocused space and structured it around user needs.
We’re seeing the same thing with AI.
AI will become deeply embedded in existing workflows.
Instead of forcing users to prompt a chatbot, AI will be natively integrated into their existing tools.
Descript isn’t an AI video generator—it’s a rethinking of video editing itself.
Instead of requiring financial advisers to learn AI, Otto Finance adapts to their existing workflows, rethinking how advisors interact with their work—not through spreadsheets, but through AI-driven orchestration.
AI tools will compete on specialisation, not raw power.
We won’t just have “one” AI for writing—we’ll have different tools for different styles and industries.
Jasper AI is for marketing. Lex is for creative writing. ChatGPT is a generalist.
This mirrors traditional software: Figma and Photoshop both edit images, but they serve different needs.
AI tools will be defined by their constraints.
The best tools will be those that don’t try to do everything but instead define clear boundaries.
Just like Notion blends structure with flexibility in note-taking, future AI tools will offer purposeful constraints that guide users toward the best outcomes.
The future of AI is opinionated.
Instead of users having to shape AI to their needs, AI tools will be designed for specific users, industries, and workflows.
We’ll see a rise in UI innovation (like Descript’s script-based video editing) that challenges conventional software models.
The best AI products won’t feel like AI at all—they’ll just feel like the best tool for the job.
Wrapping Up: Rewriting the Rules of UI
Right now, many AI products still cling to legacy interface patterns—drop-down menus, buttons, text fields—because that’s what users are familiar with. But as AI becomes more embedded in our workflows, we’ll see entirely new ways of interacting emerge, breaking free from old software paradigms.
I’m bullish about tools like Descript: removing the timeline made text the primary interface. That’s a radical rethinking of how video editing should work in an AI-first world.
The best AI tools won’t just bolt AI onto old paradigms. They’ll rethink the entire experience around what users actually need—without making them learn complex new skills.
That’s the real frontier: not just making AI-powered versions of existing tools, but inventing entirely new ways of working.
The shift toward specialised, opinionated AI tools isn’t just about automation or reducing friction—it’s about unlocking entirely new ways of working that we can’t fully predict yet.
The history of software gives us clues:
Early word processors imitated typewriters before evolving into modern collaborative tools like Google Docs and Notion.
Early mobile apps mimicked desktop interfaces before discovering swiping, pinching, and gestural navigation.
Early AI tools still feel like chatbots and dropdown menus—but that won’t last.
Right now, we’re in the “replication phase” of AI interfaces—where most tools still feel like traditional software with AI bolted on. But as AI becomes more embedded and designers rethink what’s possible, we’ll start to see entirely new UI paradigms emerge.
Just like touchscreens made us rethink mobile interfaces, AI will force us to rethink digital tools from the ground up. The real innovation in AI won’t come from better models alone. It will come from the design breakthroughs that make AI feel effortless, invisible, and deeply suited to the use cases we’re finding for it.
Signal Path
AI is reshaping the way we design—our products, our tools, our jobs. Signal Path is a weekly exploration of the challenges and opportunities in making AI products intuitive, trustworthy, and a little more human. Written by Andrew Sims, a design leader working in Fintech, it’s for designers, product thinkers, and technologists grappling with AI’s impact on their work and the products they make. I’m not claiming to have all the answers—this is a way to think through ideas, articulate challenges, and learn as I go. Thank you for joining me as I navigate this path.