AI Research Assistants

~490 Words | ~2min Read

I recently had to prep for a client call. The problem? I didn’t know their market well, and I didn’t have time to research it myself. In consulting, you’re expected to speak the client’s language. You have to understand their market position, current initiatives, competitive headwinds. And all that just to know how your engagement fits their broader strategy. That kind of research takes hours. Hours I didn’t have.

So I tried something different. I gave my AI agent a list of questions: What’s their market position? What initiatives are they pursuing? What headwinds are they facing? What’s their recent public history? I also told it to use public sources, and to cite it’s sources. Then I let it work.

The AI transformed my questions into search queries. It executed those searches and extracted semantically relevant snippets. Then it used that context to create structured research document with citations. What would have taken me 4-6 hours took 30 minutes. I showed up to that call prepared.

This brought to mind a shift in thinking I’ve been exploring. AI isn’t about automation—it’s about augmentation. The question isn’t “What can AI do for me?” It’s “What would I do if I had a team working with me?” If I had a research assistant, I’d delegate: “Here are my questions. Go research these and brief me before the meeting.” That’s exactly what I did.

AI excels at transformation tasks. That is, it’s good at changing data from one shape to another. In this case: questions → search queries → search results → relevant snippets → synthesized research document. That’s a sequence of transformation tasks. And internet research? That’s entry-level knowledge work. Searching and condensing. AI puts entry-level research assistant capabilities in your hands. With some obvious caveats around credible sources, obviously!

Tools like AI internet search has opened entire worlds of transformation tasks! You’re not limited to what the AI knows from training, nor even from the context you can directly give it! It now has the ability to gather current information, and cite sources! You can use this to build context you can verify, much in the same way you’d start trying to learn something!

The workflow I used isn’t consulting-specific. You could use this approach for design research. You can get context on unusual uses of Azure resources!. Or preparing business cases for leadership. Or collecting industry best practices for your team. Any time you need to gather and synthesize information quickly, this pattern applies.

I’ve documented the workflow and made it available. Note: This is not about replacing your thinking! It’s about compressing the information gathering stage. That way you can focus on synthesis and decision-making. The research assistant does the grunt work. You do the knowledge work.

If you’re curious, the workflow is here: https://github.com/djscheuf/non-dev-agentic-toolkit/blob/main/README.md#research–learning. Try it with your own AI agent. Adapt it to your domain. See what happens when you delegate the research and keep the thinking.