Practice15 min read

Writing Up What You Found

Research doesn't matter until it's a recommendation

You've done the work. You've asked good questions. You've stress-tested your thinking. Now you need to turn it into something someone will actually read and act on.

Most research write-ups fail in the same way: they front-load the process ("I looked at 7 vendors, ran a matrix, talked to 3 references...") and bury the recommendation. The reader scrolls. They never get to your conclusion.

The recommendation-first memo

Flip the order. Three components, in this exact sequence:

The recommendation — one sentence. "We should adopt Vendor B for our Q3 customer support migration."

The reasoning — 3-5 bullets, each tied to a criterion that matters. Include confidence levels: high / medium / low.

The trade-off — one paragraph. The strongest argument against your recommendation, and why it doesn't win.

That's it. Three components. The audience knows what you're recommending in 5 seconds. They have your reasoning in 30 seconds. They know you've considered the counter-case. They can either approve or push back specifically.

For longer-form audiences (a board, a steering committee), add an appendix with the underlying detail. But the memo itself is short.

Confidence levels are the most overlooked detail. Marking a bullet "high confidence" vs. "medium confidence" tells your reader where to push back. It also protects you when an assumption turns out wrong — you flagged it. Recommendations that pretend everything is equally certain age badly.

Practice Exercise

You just spent a week researching whether your team should adopt a new project management tool to replace your current setup. You've narrowed it to two options (Linear and Asana). Your research says Linear is the better technical fit but Asana is more familiar to your team. Your director is busy and won't read more than half a page. The decision affects 30 people. Write the prompt that turns your raw research into a recommendation memo your director will actually read.

No pressure — just give it your best shot! Write a prompt for the scenario above and our AI will give you friendly, specific feedback on how to improve.

0 / 5,000

(try writing your own prompt first!)

You did it.

You've finished "AI for Research & Analysis." Here's what you can now do that you couldn't 90 minutes ago:

Tell search and research apart — and use AI accordingly

Ask 5-dimension research questions that produce defensible answers

Compare options without AI's hedging bias — force the call

Spot where AI is most likely to make things up and verify the right things

Source-ground your prompts — feed AI what it needs, don't trust its memory for specifics

Build a decision matrix that makes your reasoning visible and changeable

Stress-test your own thinking by asking AI to argue against you

Write a recommendation memo that gets read and acted on

The most important thing now: use it on something real. Pick one decision you're working through this week — a vendor, a hire, a strategic call. Run the research, build the matrix, stress-test it. The next time you walk into a meeting with a recommendation, it'll be the most defended one in the room.