DEV Community: Jeffrey Martin

Designing for AI vs. Humans: Are We Ready for the Next Digital User?

Jeffrey Martin — Mon, 10 Feb 2025 21:01:39 +0000

For decades, web design and SaaS development have been laser-focused on human interaction. We optimize for usability, accessibility, and performance, ensuring that users can navigate, engage, and convert efficiently. But as artificial intelligence (AI) agents—large language models, virtual assistants, autonomous bots—take on more roles in decision-making and automation, a crucial question emerges:

Are we doing enough to create the best experience for AI users just by following current best practices?

Or do we need to fundamentally rethink how we build digital products to accommodate an increasingly AI-driven world?

Read the full article here.

AI as a First-Class User: The Next Digital Shift

AI is no longer just a tool; it’s becoming an active participant in the digital ecosystem. From AI-driven search and personalized assistants to autonomous procurement and data-driven decision-making, AI is beginning to engage with digital platforms much like humans do—but in a fundamentally different way.

Here’s what’s happening now:

AI-powered search and discovery (e.g., ChatGPT browsing, Perplexity AI, Google’s Search Generative Experience)
Automated purchasing and decision-making agents (e.g., AI-driven financial trading, procurement bots)
AI-assisted customer service and engagement (e.g., AI-driven chatbots interfacing with websites)
Content aggregation and synthesis (e.g., AI scrapers summarizing and reformatting web content)

Yet, most websites and SaaS platforms weren’t built with AI as a primary user in mind. Instead, AI navigates a landscape designed for humans, often relying on crawling, scraping, and interpreting data in imperfect ways.

Are Current Best Practices Enough?

Most developers and designers follow best practices that happen to align with what AI agents need, but is that enough? Let’s break it down:

✅ What We’re Already Doing Right

Accessibility (WCAG Standards): Structured content, semantic HTML, and ARIA roles already improve machine readability.
Performance Optimization: Fast-loading, well-structured sites benefit both human users and AI crawlers.
SEO & Schema Markup: Structured data like JSON-LD and proper metadata help AI parse content effectively.
API-Driven Architectures: Many SaaS platforms expose structured APIs, making AI consumption easier.

❌ What We’re Still Missing

AI-Specific Content Structuring: AI agents don’t need visual affordances like buttons, but instead require explicit data structures for deeper interaction.
Better Context Awareness: AI often misinterprets intent in digital experiences designed for human intuition.
Multi-Modal Optimization: AI agents consume text, images, and video differently—should we start thinking beyond just traditional HTML structure?
Agent-Specific UX: Should we create parallel UX models—one for humans, one for AI agents?

What Would an AI-Optimized Experience Look Like?

If we shift our mindset toward AI-first design, we might see new principles emerge:

1. Structured, API-First Websites

Instead of AI scraping unstructured content, websites could natively expose AI-readable endpoints, making data retrieval seamless. Think of it as progressive enhancement, but for AI—a layer of structured data meant just for artificial users.

Example:

Instead of an AI scraping a blog post, websites could provide an AI-optimized content feed (e.g., a JSON endpoint designed for summarization).
E-commerce sites could offer AI pricing feeds to allow procurement bots to make informed decisions without scraping raw HTML.

2. Agent-First Navigation & Interaction

Rather than expecting AI agents to simulate human interaction (clicking buttons, navigating menus), we could rethink how AI “browses.”

Example:

Imagine a version of your SaaS platform where AI agents query data in natural language instead of interacting with a visual UI.
Websites could provide hidden machine-readable prompts that guide AI toward correct interpretations of intent.

3. Conversational & Context-Aware Interfaces

Since AI often acts as an intermediary between humans and digital experiences, should we be designing for AI-to-human translation?

Example:

If a virtual assistant books travel for a user, should travel websites provide a structured itinerary format specifically for AI to summarize clearly?
AI-powered voice search optimization might move beyond keyword-based SEO to intent-based, structured responses.

Which Industries Will Get There First?

Some industries are already making moves toward AI-first design, either out of necessity or opportunity:

1. E-Commerce & Procurement

AI-driven purchasing is already happening. B2B procurement bots, price comparison agents, and AI-generated shopping assistants need structured, real-time product feeds.

2. Finance & Investment

Algorithmic trading and AI-driven financial advisors don’t navigate bank websites like humans. Structured financial data APIs will become more critical than ever.

3. Healthcare & Research

AI is consuming vast amounts of medical data for diagnostics and treatment recommendations. Should medical websites and patient portals have an AI-friendly layer?

4. Search & Content Discovery

Google’s AI-powered search evolution means searchable content needs to be AI-optimized, not just SEO-optimized.

The Big Question: Is It Time?

While AI isn't a primary "user" of most websites yet, the shift is happening faster than we think. The real question is:

Do we wait until AI becomes a dominant digital consumer before adapting?
Or do we proactively rethink web design for both human and AI users now?

Much like accessibility wasn’t always a priority in web design—but became an essential part of inclusive, ethical development—AI-first design might soon follow.

The web is no longer just for people. AI is here, and it’s navigating our digital world in ways we’re only beginning to understand.

Are we ready?

What Do You Think?

Should designers and developers start optimizing for AI consumption alongside human users? Or is it still too early? Let’s discuss.

#AI #WebDesign #SaaS #FutureOfTech #UX

Calling All Parents in Tech: Help Shape the Future of Accessible Learning

Jeffrey Martin — Fri, 24 Jan 2025 21:39:07 +0000

As developers and creators, we know that building a great product isn’t just about the code—it’s about understanding the people who will use it. That’s why I’m reaching out to you, fellow developers, creators, and especially parents, to lend your voice to something I’m deeply passionate about: The Unbound Initiative : Infinite Stories Unfolding.

The Vision: Education for Everyone, Everywhere

The Unbound Initiative is more than just a project; it’s a mission to reimagine how we learn. We’re building a platform that combines interactive storytelling with real-world learning, making education not just engaging but accessible to youth and lifelong learners around the world.

As a parent, you know the value of tools that enrich your child’s education. And as a developer, you understand how vital user feedback is to the success of any product. This is where you come in.

TLDR; 👉 Fill out the questionnaire here!

Why Your Voice Matters

Every feature we build, every design choice we make, and every line of code we write is meant to create a platform that helps young minds thrive—regardless of disabilities, economic barriers, or learning styles. But we can’t do this without your input.

Your unique perspective as a parent gives us the insight we need to ensure The Unbound Initiative becomes a product that genuinely enhances educational journeys. What challenges does your child face in learning? How do you balance screen time with meaningful engagement? What features would make this platform a valuable part of your family’s routine?

These are the kinds of questions we’re exploring, and we want to hear your answers.

The Power of UX Feedback

As someone in tech, you already understand the importance of UX research. Feedback isn’t just a checkbox—it’s the cornerstone of creating a product that truly resonates. With your help, we can uncover what matters most to parents and students and make informed decisions that lead to an impactful, inclusive platform.

Think of it this way: your feedback isn’t just helping shape a product—it’s contributing to a larger mission of making education a journey of curiosity, creativity, and accessibility for learners everywhere.

How You Can Help

I’ve created a short UX questionnaire on Google Forms to gather insights from parents like you. It takes just 5 minutes, but your input could make a lifetime of difference for learners around the world.

👉 Fill out the questionnaire here!

Let’s Build This Together

As a developer, I know how hard it can be to get a product off the ground. But with the support of a community like this—filled with people who understand both the technical and personal sides of creating—we can build something extraordinary.

So, if you’re a parent, please take a few minutes to share your thoughts. And if you know someone who might be interested, share this post with them, too. Together, we can help The Unbound Initiative thrive and ensure that learning is accessible, enjoyable, and empowering for everyone.

I value each and every one of you, and I thank you in advance for participating to help this project move forward. Feel free to share the link with teachers, students themselves, or other parents.

Let’s create something that makes a difference.

#UXResearch #EdTech #DevCommunity #Accessibility #Learning #Parenting

My First Attempt at Building a Retrieval-Augmented Generation (RAG) Model

Jeffrey Martin — Sun, 19 Jan 2025 01:48:25 +0000

Hi, I’m just starting out with AI development, and today I took on the exciting challenge of building my first Retrieval-Augmented Generation (RAG) system. The idea of combining retrieval-based techniques with generative AI fascinated me, and I wanted to give it a try. Spoiler: it’s been a mix of confusion, small wins, and a lot of learning!

This journey has been an incredible learning experience, and I’m thrilled by how much I’ve learned about building intelligent AI-driven systems. If you’ve worked on RAG models or have tips on using tools like LlamaIndex or SingleStore, I’d love to hear your advice!

I plan on developing this and "messing around" more tomorrow and this week as I plan to use this in a project that I am working on.

If anyone has any advice at all I would welcome any suggestions, tools, etc.

Welcome to Your Constellations of Purpose.

Jeffrey Martin — Fri, 17 Jan 2025 23:22:52 +0000

This is a submission for the GitHub Copilot Challenge : New Beginnings

What I Built

Constellations of Purpose: Tasks and goals are represented as stars. Users connect these stars to form constellations of meaning, revealing the bigger picture of their life’s direction. I wish I had more time, but it's a on day build - so I stopped. Ideally I would love to add Authentication and an archive of goals completed, that are visualized in an interactive method.
The MVP I am submitting uses LocalStorage for data - so if user deletes cache the data is lost.

Demo

DEMO LINK

Repo

JeffreyScottMartin / constellations-of-purpose

Constellations of Purpose

Constellations of Purpose is a personal goal management tool that helps you visualize and achieve your dreams. Create goals, break them down into manageable tasks, and watch as your progress lights up the sky.

Entry for the GitHub Copilot 1 Day coding challenge Demo Link

https://constellations-of-purpose.vercel.app/

Prompt: New Beginnings

Features

Add and manage goals
Break down goals into tasks
Track task completion
Visualize progress with completion percentages
Onboarding flow for new users

Getting Started

Prerequisites

Node.js (>= 12.5.0)
npm or pnpm

Installation

Clone the repository:

git clone https://github.com/JeffreyScottMartin/constellations-of-purpose.git

cd constellations-of-purpose pnpm install pnpm dev pnpm build pnpm start

├── .gitignore
├── .next/
├── eslint.config.mjs
├── next-env.d.ts
├── next.config.ts
├── package.json
├── pnpm-lock.yaml
├── postcss.config.mjs
├── public/
├── README.md
├── src/
│   ├── app/
│   │   ├── goal/
│   │   │   └── [goal]/
│   │   │       └── page.tsx
│   │   ├── goals/
│   │   │   └── page.tsx
│

…

View on GitHub

Copilot Experience

This was my first attempt at really using Copilot. I used it for a lot of auto completion, but also realized the power of Copilot for help in generating a full component, after playing around with the prompts and realizing that ChatGPT-4o produces better output for the project if you prompt in detail of what you are looking for. Also, when I used @workspace - it took longer but the responses were better as Copilot examined the entire workspace. I found Claude 3.5 Sonnet was my goto for smaller assistive functions within a file. eg prompt for claude "animate the task stars orbiting around the goal star" and bang!, there it was a perfect little animation for my canvas context. I used Copilot Inline mostly to quick fix type errors. I also had Chat Gpt4o write my README at the end. I also used Claude to see if there were any possible Linting errors prior to running a build.

GitHub Models

I initially developed a GitHub model to respond as if from another planet in the cosmos, however didn't really see anything that was clearly like that

Conclusion

I definitely has changed my work process moving forward. By the end of the day it was like having another developer right there with me. The best part I feel is that it allows a CREATIVE person like me to be creative and worry less about implementation as I have an assistant that can provide answers and guidance quickly. The the creative create!

All in all such a fun project. And to see how far one can get in a day with a new way of developing was a challenge. Congrats to everyone who made a submission.