How I used AI as a designer this week

At this point, none of us can go 3 minutes without hearing AI mentioned, so let it be known this isn’t my first time interacting with an AI-based tool. Over the past few months, I’ve been casually replacing administrative tasks with AI & using various vibe-coding platforms to iterate quickly. Most recently, I tried using the deep research feature. My first attempt involved digging FAR into the fascinating world of HVAC, but believe me, you don’t want to hear about it. Instead, let’s dive into how I used ChatGPT and Perplexity to run a round of deep research on chatbots.

Feeling behind? Start here.

Building The Prompt

From what I’ve seen, learning the right way to prompt is one of the biggest learning curves. Initially, I tried crafting a prompt myself, & the results weren’t detailed enough. No matter what I added, I didn’t include enough details to generate a thorough study.

To sidestep this, I used a prompting plug-in for ChatGPT. Very meta, right? I asked the plugin to help me draft a prompt for deep research. Actual footage:

Promptable UI feature

From there, ChatGPT returned a set of clarifying questions for me to answer about what specifics I was looking for. Here’s a look at how I answered:

Generating the prompt

Once it had my clarifying responses, here’s the prompt it outputted:

Conduct a comprehensive deep-dive research study on the current feature landscape of AI-powered chatbots and conversational assistants (e.g., ChatGPT, Claude, Gemini, Copilot, Perplexity, Pi, Meta AI, Mistral-powered assistants, and other notable entrants).

Goals

Provide a holistic overview of what features leading AI assistants are offering.

Analyze UI/UX, business, and technical features in detail.

Identify common trends, differentiators, and emerging innovations.

Highlight gaps and opportunities for designing a new AI chatbot.

Structure of Output

Executive Summary

Key findings and top-level trends across all assistants.

Emerging themes shaping the AI chatbot landscape.

Feature Landscape Report (Long-Form) Break down features into three main categories:

UI/UX Features: personalization, memory, multimodal input/output (voice, images, video), conversation history, explainability, tone control, adaptability, plugins, integrations with productivity tools.

Business Features: pricing models, enterprise plans, analytics dashboards, compliance/security, integrations with CRMs, team collaboration features, marketplace ecosystems.

Technical Features: context window sizes, fine-tuning options, custom instructions, API access, latency, offline/local options, extensibility, architecture strengths.

For each, describe what the major players are doing, noting similarities and unique differentiators.

Competitive Feature Matrix (Table Format) Rows = feature categories Columns = major AI assistants ✅ = implemented 🚧 = partially available / in beta ❌ = not available Notes = important caveats or differentiators

Trend Analysis

What features are becoming standardized?

What features are new and differentiating?

Where are users still underserved?

Predictions: what’s next in AI chatbot features (e.g., deeper personalization, agent ecosystems, emotional intelligence, proactive workflows).

Opportunities & Gaps

Identify areas where existing chatbots are falling short.

Suggest strategic opportunities for building a new chatbot that competes effectively.

Style & Tone

Long-form, detailed, and structured.

Use tables, bullet points, and clear headings for readability.

Running the Prompt

Looks great to me. Here, I shifted tools to use Perplexity because they graciously gave me a free month trial. Since I still use the free version of ChatGPT, I tend to hoard the credits. I copied the prompt verbatim into Perplexity & chose the “research” option.

Then Perplexity went to work. You can watch as it scans the internet, finding sourcing, analyzing content from articles, & giving you a play-by-play of how it’s gathering data. Research results started populating in under 5 minutes.

Outputs

The end result? Shocking. Here’s a snippet:

Visual comparison table generated in Perplexity

Perplexity offered a few file types to download. My original plan was to export as a PowerPoint for read outs. When exporting to a .ppt, I found that a lot of research information which was included in the response was being dropped/summarized. I decided to take a manual approach instead & copied out of the pdf download (which was seemingly a word-for-word copy of the output).

Now, I will say my first take read like Anna Karenina. It was over 40 slides, all packed with 150 words. Not exactly set up for top tier engagement. My export required a lot of human massaging to make it digestible. Since AI had done the bulk of the work for me, I had extra time to devote to this step.

What went well

1. using AI to write the right prompt — once I had AI write a prompt for me, the output was incredible
2. depth of research conducted — over 104 sources were scanned
3. visual elements produced — the competitive feature matrix (seen above) was a downloadable image to insert anywhere

What didn’t go well

1. human-produced prompt — the output of my first human attempt was laughable
2. had to run pre-research to find a starting place for which bots to compare — I’ve noticed at a certain point, AI starts to forget. I included my own list of bots to compare to try & anticipate this
3. exporting to .ppt file from Perplexity — it needed a human editor anyway, but it was a bummer to complete this manually
4. no healthcare examples were included — not surprising with the type of protected data offered as a base

Summary

In summary, days were reduced to hours to complete tasks. While it wasn’t all perfect — I still had to refine the export and add a human touch, the time saved was huge. It’s exciting to see how AI is becoming part of our daily workflow at athenahealth.

At athenahealth, we’re developing a strong culture around working AI into our daily responsibilities. Follow along as some of the athenahealth designers continue to share our AI-oriented learnings.