Module 5 of 8 · Intermediate · 10 min read

Research and summarising

Find, read and synthesise sources with AI research tools, and keep every claim tied to something you can check.

Key takeaways

1Use search-connected or deep-research tools for anything current or factual.
2Ask for sources on every claim, then open and check the important ones.
3Summaries should be shaped by your purpose: decision, learning or reference.
4Look for what's missing, contradicted or uncertain, not just a neat answer.

01Where AI helps

Deep-research modes in many assistants run multi-step searches and write cited reports. They save hours, and still need checking.

Scoping a topic: key questions, terms, people and debates.
Searching widely and reading many sources quickly.
Summarising long documents, papers, transcripts and reports.
Comparing sources and surfacing disagreements.
Turning findings into briefs, tables and literature reviews.

02The workflow

This is the orchestrator-worker pattern from Multi-agent patterns, with you as the orchestrator.

Write the research question and what you will do with the answer.
Ask the AI to map the topic and propose sub-questions.
Run searches or deep research; collect sources with links.
Read and summarise key sources for your purpose.
Synthesise: agreements, disagreements, gaps and confidence.
Check the most important claims in the original sources.

03Tools to reach for

A model without search has a knowledge cutoff and can't see recent events. Use search for anything that may have changed.

Chat assistants with web search or deep-research modes, for cited reports.
Answer engines that show sources inline.
Academic search tools that find and summarise papers.
Document chat tools for long PDFs and your own files.

04Prompt: scoping

A good map of the topic makes the searches much better.

Example

I'm researching {{topic}} to {{purpose, e.g. decide whether to enter this market}}. Map the topic: key sub-questions, the main schools of thought or debates, important terms I should search for, the types of sources most reliable here, and what would change my decision. Don't answer the questions yet.

05Prompt: purposeful summary

Tell the model why you are reading.

Example

Summarise the document in <doc> for {{audience}} who need to {{decision or use}}. Structure: the main claim in one sentence; key findings with page or section references; methods and their limitations; what the document doesn't address; and how confident we should be, with reasons. Quote exact figures; don't round or estimate.

06Prompt: synthesis across sources

Ask for disagreement as well as consensus.

Example

Here are summaries of {{n}} sources on {{question}}, each with a link: <sources>{{summaries}}</sources> Synthesise them: where they agree, where they disagree and why (different data, methods or dates), what remains unknown, and the overall answer with a confidence level. Cite sources by number for each point. Flag any source that looks low quality or out of date.

07Pitfalls

The goal of AI research is a faster path to sources you've checked, not a substitute for them.

Fake or wrong citations: open the link and check it says what's claimed.
Out-of-date answers: check dates on sources, especially for prices, laws and technology.
One-sidedness: ask for the strongest counter-arguments.
Summary drift: summaries of summaries lose nuance. Go back to originals for key claims.

Finished reading?

Mark it complete to track your progress, then test yourself with the AI at Work: Business & Professional quiz.

Next: Business plans and strategy →