The AI Tools Grew Up Too

AI tools have shifted from shiny new features to daily infrastructure in just three years.

An early adopter doesn't just get ahead of the market. They get ahead of a moving target, because the tools themselves keep changing underneath them. Looking back at the AI tools I was actually using in 2022 and 2023, next to what those same tools do now, the distance is bigger than I expected.

ChatGPT: From a Clever Assistant to an Actual Working System

Four years ago, I was using GPT-3 and remember hesitating over moving to GPT-4.  At the time, the information was over a year old in GPT-3 and was moving to around 6 months delay in real-time information. Yes, GPT-4 could do some new, cool stuff like reading a diagram or a document and reasoning about it in plain language. I remember asking myself if it was really worth it to have more current data in my “Wikipedia on Steroids.”

The picture shifted again when I tested GPT-5 in 2026. Those earlier models answered from a frozen snapshot of the internet, sometimes badly out of date. GPT-5 now decides, in real time, whether to answer from what it already knows or pull in a live search. OpenAI's own testing found this cut the rate of factual errors by roughly 45 percent compared to GPT-4o.

I was also evolving with the market. In the 2023 to 2026 period. I also began testing and using tools like Bard, Gemini, and Copilot.

The Assistant Horse Race: ChatGPT, Copilot, Gemini, and Bard

I've used all the major assistants at different points — ChatGPT, Copilot, Gemini, and Bard before it was retired and folded into Gemini in early 2024. From 2022 to probably mid-2025, ChatGPT was my clear preference.

That's shifted in the past year, and not in a straight line. Gemini has been solid, but I still find it harder than it should be to pull a clean table or a specific block of text out without dragging the whole response along with it. ChatGPT, the tool I trusted most, has started to feel canned when I ask it for ideas or editing help. Copilot is capable but generic, though I hold that judgment loosely.

My bar for a good interface is a lot higher now than it was when any of this felt new, and that's worth admitting before blaming the tool. Also worth noting is that I am using it as part of a small, niche consulting business.

Grammarly: From Proofreader to Judgment Call

I started using Grammarly in 2020 for editing. It was the best tool on the market at that time and I believe is still best today.

Somewhere in the 2023 or 2024 timeframe, Grammarly hopped on the train of promoting itself as an AI tool.  I was skeptical because everyone seemed to start offering AI-enabled tools at that time. I watched Grammarly grow its capabilities in 2024 and 2025, but essentially it just remained the company’s editing tool.

Today, Grammarly reads the document type itself and adjusts its judgment accordingly. It also now checks whether text was AI-written and can "humanize" AI output for different audience types. It has solid fact-checker and citation tools, too.  Grammarly has moved from a spell-checker with new tricks in 2024 to what is now closer to an AI-driven editorial judgment platform.

Canva: From a Design Shortcut to a Full Studio

The Canva I have used and written about for nearly 6 years was useful for exactly what a small firm needs. It helps create fast, affordable visuals without a design department. That was also its Magic Studio's early promise. And like Grammarly, it’s the best visual design tool on the market, especially for small businesses.

The 2026 version of Magic Studio consolidated four separate AI tools into one conversational interface. It's advertised as cutting the time to generate a usable design from roughly eight seconds to under two. It can turn a spreadsheet into a chart, expand a photo's frame, or lift a subject out of a background entirely.

What I was using to save time on visuals has become something closer to a design collaborator that also saves time.

Midjourney: From Novelty to Speed

Midjourney was actually the first of these tools I used in practice. Back in May 2023, I generated the header image for an article on AI and project management in under ten minutes. It felt like a small miracle at the time.

When I wrote about leveraging AI visualization tools like Midjourney the following year, the honest pitch was that it could produce a meaningful image when you didn't have one, without compromising a client's confidentiality. That's still true.

What's changed is that I don’t use Midjourney anymore. Canva can do all we need now, and occasionally I get a visual generated by Gemini. Consolidating tools for your operating context is a needed and natural evolution.

The Buzzword Test: Summarizing and Document Tools

That same 2023 article on AI for project managers included a section I titled "Standard Promises from Software Companies." I was blunt about it then — plenty of vendors were relabeling ordinary automation as "AI (aka, our software)" and selling it as something new. Budget analysis, schedule optimization, report generation. Useful features, dressed up in a buzzword.

The tools I actually trusted at the time were narrower and more honest about what they did: Wordtune Read for document summaries, Google's Natural Language API for entity and sentiment analysis, and Live Transcribe for accessible transcription.

Wordtune Read as a standalone product is gone now, discontinued by AI21 Labs in 2025 as the company hit financial trouble. The AI assistants and Grammarly can do what we were using Natural Language API for “back in the day” (a.k.a. about three years ago). Like Midjourney, a natural progression of tool consolidation for our context and the changing market.

Live Transcribe remains the best accessible transcription tool on the market.

The Pattern Underneath It All

None of these tools got smarter in a straight line, and none of my preferences or our small consulting firm’s context held steady either. Each AI tool crossed a threshold where it stopped being a feature and became something you plan a workflow around. I think that’s a sign of a market maturing and maturing users. The AI tools stopped being interesting and started being assumed as part of the process. The tradeoff is that your patience for each AI tool’s rough edges gets shorter.

Which version of these tools are you actually using?  Are they something that you occasionally reach for or are they something you’ve incorporated into your workflow?  Are you still fascinated by what AI can do, or have you matured into a critic of each tool’s limitations?