Deep Research with WorkBuddy

When to Use It, and When Not To

By the end of this article you'll be able to turn a casual "look this up for me" into a research report that's structured, sourced and checkable.

Good for: entering a new industry, evaluating suppliers, background research before writing a proposal, and a one-page competitor overview for your boss. Not for: work that depends on unpublished internal company data, requires interviews with real people, or needs a legal or investment conclusion — in those cases it can supply material but can't make the call for you.

Get the Question Right First

Eighty percent of research quality comes from the question. Before you start, write down four things:

• Subject: a specific company, category or region — not something vague like "this industry".
• Time frame: the last two years, or a five-year trend? State the year range explicitly.
• Dimensions: market size, key players, price bands, channel structure, regulatory environment — pick three to five, that's enough.
• Audience: internal decision-making or a client? That determines the level of detail and the tone.

Launching the Research in One Instruction

Research the domestic residential energy storage market for 2025-2026, covering market size, key players, price bands and channel structure. Output a structured report of about 3,000 words, with a source and date for every key figure. Where something can't be found, write "not available" explicitly — do not estimate.

That last sentence matters. Letting it say "I couldn't find it" is far safer than pushing it to invent a good-looking number.

What the Output Usually Looks Like

Typically it includes a summary, a body organized by dimension, a comparison list and a source list. Spend two minutes on the summary and sources before reading the detail — that's more efficient than reading start to finish. If the source list has only two or three entries but covers five or six dimensions, be suspicious.

Sources Must Be Verified

Don't skip this step. Do at least three things:

1. Open every source link and confirm the page actually says that, not just something adjacent.
2. Check the year of the data. Plenty of "market size" figures are three years old but look current after being cited repeatedly.
3. Rephrase a key number and ask again; see whether the conclusion holds. A large gap between the two runs means that item isn't reliable.

Public material is itself biased. It stitches sources together; it doesn't judge what's true. Before citing anything externally, try to find the primary source.

Producing a File You Can Actually Use

Once the content is confirmed, have it convert formats: a PPT deck for your boss, a Word report for a client, keeping source footnotes and consistent heading levels. For a colleague version, add "remove the process narration, keep only conclusions and evidence".

Three Common Mistakes

• Asking ten dimensions at once: every item comes back shallow. Do several rounds and merge at the end.
• Citing numbers straight from the report: for anything external you must trace back to the primary source.
• Assuming it has read your internal material: it only reads files you've authorized and public information. What you didn't give it, it doesn't know.

Once you've done this a few times you'll notice the real saving is in gathering and structuring — the judgement is still yours. To build that skill systematically, start with the methodology in AI learning.