Automate your research and analysis

Outcome

By the end of this lesson you'll have a repeatable way to go from "I know nothing about this" to a short brief you can put in front of a decision-maker. You'll orient quickly, compare options against criteria you chose in advance, and finish with a list of exactly what needs checking before anyone relies on it.

Concept

Research is where AI feels most impressive and is most risky. It can organize a topic in seconds. It can also state something wrong in the same confident, tidy voice it uses for something right. A wrong fact inside a neat brief is harder to spot than a sloppy guess.

So treat the AI as a fast, well-read colleague who sometimes misremembers, and set up your process around that. We'll use a four-stage flow called Orient, Frame, Compare, Verify:

1. Orient — learn the lay of the land and the common traps.
2. Frame — decide the question and the criteria before looking at options.
3. Compare — evaluate every option by the same yardsticks.
4. Verify — check the claims that will actually carry weight.

Two principles run through all four. First, criteria before options. If you pick what matters after seeing the options, you'll quietly pick what favors your early favorite. Second, the decision stays yours. AI lays out tradeoffs; you weigh them against things it can't see, like budget, timing, and office politics.

Also know the limits: depending on the tool, it may not have current information, and it can't tell you what's true about your own company unless you give it that. Whenever your decision hinges on something recent, go to the original source.

Build

Part 1: Orient.

I'm new to [TOPIC] and need to understand it well enough to [PURPOSE].

Give me:
1. A plain-language overview in under 150 words
2. The 3 most common misunderstandings
3. The 5 terms I need to follow a conversation
4. Questions a knowledgeable person would ask first

Mark anything you're unsure about with "(verify)".

Part 2: Frame the decision. Write down, in your own words, the question and your criteria. For example: "Which of these three tools should our team trial? Criteria: cost per person, setup time, data privacy terms, and how well it fits what we do now." Rank the criteria by importance. Do this before Part 3.

Part 3: Compare.

I'm deciding: [DECISION].

Options: [OPTION A], [OPTION B], [OPTION C]

Criteria, in order of importance:
1. [CRITERION]
2. [CRITERION]
3. [CRITERION]

Make a table: one row per option, one column per criterion. In each cell,
give a short assessment and mark confidence as high, medium, or low. Where
information may be out of date or you don't know, say so rather than
estimating. Do not recommend a winner.

Part 4: The decision brief.

Turn this comparison into a one-page brief for [AUDIENCE].

Sections:
1. The decision and why it matters now
2. Tradeoffs between the options
3. What would make each option the right choice
4. What I'd need to verify before deciding, as a checklist

Keep every uncertainty from the comparison visible. Do not smooth over
low-confidence items.

Part 5: Verify. Take the checklist from section 4 and work through it using primary sources: the vendor's own pricing page, official documentation, a person who knows. Mark each item confirmed, corrected, or couldn't confirm. Put those marks into the final brief.

A worked example

The topic, table, and checklist below are invented to illustrate structure. They are not real evaluations of any real product.

Decision (illustrative): Which of three note-taking tools should a five-person team trial?

Option	Cost per person	Data privacy terms	Fit with current workflow
Tool A	Low (medium confidence)	Unclear (low)	Good (medium)
Tool B	Medium (low)	Detailed (medium)	Fair (medium)
Tool C	Unknown (low)	Unknown (low)	Good (high)

Verify checklist (illustrative): Current pricing for all three. Where Tool A stores data and who can access it. Whether Tool C has a team plan.

Look at what this gives you: no winner, a lot of honest "low" markings, and a to-do list. That's a good research draft. A confident ranking built on those cells would not be.

Run

Try it now (5 minutes). Pick a small real question from your work, like a tool to trial or a process to change. Write the decision and three ranked criteria in your own words. Run the compare prompt with two or three options, then read the verify list and circle the one item that would change your answer if it turned out wrong. That's the first thing to check.

Guardrails: what a human must verify.

• Any number or claim that will influence the decision. Prices, dates, statistics, and feature claims get checked at the source.
• Quotes and citations. If the AI names a study, article, or person, find it yourself. Invented references look real.
• Recency. Treat anything that changes often (pricing, policies, product features) as unverified until you've seen it on the current official page.
• Privacy. Don't paste confidential plans, customer information, or internal financials into tools you haven't approved. Describe the situation in general terms instead.
• The call itself. Don't let "the AI said" stand in for your reasoning.
• If the topic touches money, health, or legal matters, this is general education, not professional advice. Check with a qualified person before acting.

Make it a habit. Save the four prompts as a set. Always write criteria first. Keep the "confirmed / corrected / couldn't confirm" marks in your brief so readers can see how much weight it carries.

Checkpoint.

• I wrote my criteria before comparing options.
• My comparison shows confidence levels and gaps.
• I checked at least three key claims at their source.
• The brief lists what remains unverified.
• The final decision is mine, with my reasoning attached.

You now have prompts for inbox, meetings, documents, and research. The last step is pulling them together in Automate your role.