How AI has made me more productive

The last few months have seen mounting speculation that we’re in the throes of an AI bubble. Much of the market’s surge comes from AI companies, with valuations tied to expectations of major productivity gains.

Yet the results so far are sobering. As The Atlantic recently observed, even companies investing heavily in AI have seen “virtually no impact on their bottom line.” One study suggests coders may be 20 per cent slower with AI than without it.

My own experience points in another direction. As a law professor and lawyer, I’ve recently crossed a threshold in my use of AI that feels like a genuine leap in productivity.

After experimenting with AI off and on since 2022, I’ve begun relying on it daily over the past two months. I now use Perplexity, Claude, and ChatGPT as near-constant companions. They’re research assistants, sounding boards, and editors I turn to at every stage of my writing.

The effect has been transformative, less like adopting a new app and more like moving from typewriter to word processor, or from paper-based research to the internet.

For me, the biggest gains have come in three areas central to legal and academic work: conducting research more efficiently, drafting with fewer roadblocks, and editing to a tighter finish. Each has changed my workflow in ways that, while difficult to measure in hard numbers, are impossible to ignore.

Research, transformed

Before chatbots gained web-search capabilities, researching a legal issue meant looking up lawsuits, cases, or scholarship on Google, CanLII, or SSRN.

It was a multi-stage process that saw me run several queries, sift through links, waste time on irrelevant material, and only gradually sketch the contours of a problem.

But as Perplexity and other models have improved at gathering and summarizing sources, they’ve brought about a radical shift in how I conduct preliminary research. 

Results still vary, but these tools are now consistently good at producing summaries accurate enough to get me started, usually with links to a few sources that let me quickly gain my bearings.

More often than not, one initial source that GPT or Perplexity surfaces will contain a trail of footnotes or references leading straight to the essential material.

Rather than poking around SSRN, Springer, or Google, an AI summary with links gets me off the ground much faster and provides a reliable lay of the land.

Within minutes, I can judge whether an argument or idea is worth pursuing or should be reframed. That simply wasn’t possible before, and it’s impossible now to give up.

Writing with a new rhythm

An insight offered soon after ChatGPT appeared still rings true: AI won’t replace skilled writers, but it will enable those who can’t write well to write well enough.

Until recently, my rule of thumb was that I didn’t need AI to write—I did it well enough myself. I would only occasionally drop in a sentence, paragraph, or title and ask GPT or Claude for alternatives, and I accepted its suggestions sparingly.

But I now consult AI routinely, dipping in and out more fluidly with shorter snippets. 

The value lies less in offering consistently better wording than in seeing a problem from a new angle. That fresh perspective is often all I need to move forward.

AI also spots gaps in my approach. For instance, when I share opening paragraphs, it flags the lack of a hook or suggests a stronger structure.

AI as an editor

After finishing a draft, I can spend days—even weeks—chiselling away at each sentence, trying to cut the fat and improve the flow. 

AI has changed this process altogether. I now ask it to review full drafts for typos, tighter wording, smoother flow, and bold suggested changes. The edits are often so subtle that they would take me several read-throughs to catch on my own. 

For the first time, I’m adopting more of AI’s suggested edits than not, and reaching a tighter draft faster than before.

AI is also invaluable for handling citations. I keep a thread open where I drop links, titles, or full references, and it formats the McGill citation instantly without prompting each time. 

Measuring the impact

It’s hard to measure AI’s overall impact on my workflow. 

Research is where it makes the most significant difference, as gathering sources faster has enabled me to start more projects and finish them sooner.

The editing magic doesn’t boost output, but improves quality in ways that wouldn’t have been possible otherwise.

While AI may currently slow down coders, its impact—raising quality and increasing output—feels undeniable for those who primarily read and write.