How I use Claude Code to research prospects in minutes

Before Claude Code, proper account research took 20 to 30 minutes per company. LinkedIn for the contact, their website for the positioning, Crunchbase for funding, job boards to understand what they are hiring for, recent news to find an angle. Good research. Slow research.

The bottleneck was not finding the information. It was synthesising it into something actionable. You gather six sources and still spend five minutes figuring out what the actual outreach angle should be.

Claude Code collapses this. It can browse the web, read pages, and synthesise information in one pass. More importantly, it reasons about the information rather than just retrieving it. Ask it the right question and it tells you the angle, not just the facts.

The prompt matters more than anything else in this workflow. A bad prompt gets you a summary of the company's About page. A good prompt gets you a specific, actionable outreach brief.

I start by giving Claude Code the company name, the contact's role, and one sentence about what I sell. Then I ask a specific question, not a general research request.

The question I use almost every time is: what problem is this company most likely experiencing right now that is relevant to what I sell, and what is the single most specific opening line I could use to address it?

This reframe changes everything about the output. You are not asking for a company overview. You are asking for a hypothesis about their pain and a way to open a conversation about it.

I include any signals I already have. If I know they recently hired three SDRs, I include that. If their job posts mention they are switching CRMs, I include that. Claude Code incorporates the signals into its reasoning and produces more specific output.

The output Claude Code produces is a starting point, not a finished brief. Read it critically. It will sometimes miss context that is obvious to you as someone who understands the company's industry or situation. It will occasionally make a logical leap that does not hold up.

Take two things from it. The specific angle you will lead with in your first email. And one question you want answered on the discovery call. Everything else is context that is useful to have but probably will not come up.

Do not paste the output directly into your email. Read it, form your own opinion about whether the angle is right, and write the email yourself. The research is Claude's. The judgment is yours.

One pattern worth building: after a few weeks of using this workflow, you will notice that certain angles come up repeatedly for companies at similar stages or in similar industries. Document those patterns. They become your playbook for researching at scale.

For one-off research before an important call, the Claude Code workflow works perfectly. For researching 50 accounts before a campaign, you need Clay.

Clay has an AI column that runs a prompt against every row in your table. You can write the same research prompt and have it run against each company in your list using their specific data as context. The output lands in a column that you can feed directly into your email personalisation.

The setup is: import your target account list, pull firmographic data through Clay's enrichment integrations, then add an AI column with your research prompt. Reference the company name, headcount, industry, and any signals you have enriched as variables in the prompt.

The quality of the AI column output in Clay is not as high as a focused Claude Code session on a single account. But it is dramatically faster for list-level research. Use Claude Code for your top 10 accounts. Use Clay for the rest.

Deep research only makes sense when the outreach warrants it. A personalised cold email to a senior decision maker at a target account deserves 5 minutes of research. A broad awareness campaign to 500 mid-market companies does not.

Know when to use the workflow and when to skip it. If the deal size and seniority justify the time investment, do the research properly. If you are running volume outreach to a broad ICP, write a strong template and rely on good targeting rather than deep personalisation.

The mistake is treating research as always necessary. It is not. Relevance matters more than personalisation. A well-targeted email with a relevant offer to the right person beats a deeply personalised email to the wrong one every time.