Claude Just Learned to Find Your Next Customer

You do not need a bigger sales team. You need one more skill loaded into the AI you already pay for.

It is 9 p.m. in Lagos. Tomiwa runs a four-person software consultancy that builds internal tools for mid-size retailers. Tonight he is doing what he does most nights after the paying work is done: scrolling LinkedIn, copying names into a spreadsheet, guessing at email addresses. Tomorrow he will spend two hours verifying that half of them are wrong.

Multiply Tomiwa by every small agency, consultancy, and service business on the planet, and you get the real cost of prospecting. Not the fee for a tool. The hours nobody bills for.

Why finding customers still eats a whole evening

Here is how finding your next ten customers usually works, whether you sell software in Lagos, run a marketing shop in Manchester, or do consulting out of Austin.

• You scroll LinkedIn and copy names into a spreadsheet by hand.
• You buy a list from a directory and discover half the emails bounce.
• You hire a part-time assistant to do the scrolling, then spend time checking their criteria were right.
• You pay for a prospecting tool, log into yet another dashboard, and export a CSV into yet another tool.

None of that is prospecting. That is data entry wearing a sales title.

The tool was never really the bottleneck. The chat window was, because until now, the AI you talked to could describe a plan. It could not go and execute one.

What changed: Claude went from answering to acting

Ask a chatbot for prospects and it will explain how you might find some. Ask an agent that is actually connected to a live contacts database, and it goes and gets them, then tells you which ones are worth your time.

That distinction is the whole story of what shipped this week. On August 18, 2026, the sales intelligence company Seamless AI announced a new integration with Claude built on Anthropic's Model Context Protocol, or MCP. Jake Phillips, Seamless's Chief Product and Technology Officer, put it plainly in the announcement: "With the latest Seamless MCP release, Claude users can now connect Seamless and start putting B2B sales intelligence to work."

MCP itself is not new. Anthropic introduced the protocol in late 2024 as a shared, open way for an AI model to reach outside its chat window and use real tools: a database, a calendar, a CRM, a search engine. What is new is how fast serious B2B platforms are shipping MCP connectors instead of building yet another dashboard. Seamless is one of a growing list, alongside integrations from tools like HubSpot, Notion, and dozens of others now searchable directly inside Claude.

Isn't this just another CRM with extra steps?

No, and the difference matters for a small business. A CRM is a place you have to visit, log into, and learn. MCP is a bridge, not a destination. It lets the AI you already use for everything else, drafting emails, summarizing calls, writing proposals, also reach into a live prospecting database when you ask it to. You are not managing a new tool. You are asking one more thing of the assistant you already trust.

MCP was built to be model-agnostic on purpose. The same protocol is why a growing number of business tools now ship one connector that works across Claude, ChatGPT, and coding assistants like Cursor, instead of building a separate integration for each. For a small business owner, that matters more than the technical detail suggests: whichever AI assistant you settle on, the tools you connect to it are increasingly not locked to one vendor.

A demo Seamless ran to walk sales teams through the new connector shows exactly how thin that layer is. The presenter opens a plain chat with Claude and types a request in ordinary language: find a set of marketing leaders at software companies within a certain size range. Claude passes that request to Seamless, gets matches back, and keeps working the problem: "the initial query returned thousands of matches, I'll tighten the role criteria, the final set are actually on target," narrating its own filtering as it goes.

How does prospecting inside Claude actually work?

Based on that walkthrough and the public announcement, the flow breaks into two cleanly separated stages, and knowing the difference will save you money.

1. Discover. You describe the customer you want in plain language, for example, “operations managers at logistics companies with 50 to 200 staff.” Claude searches Seamless's database and returns matches. This stage does not use Seamless's paid research credits.
2. Rank. You ask Claude a follow-up question, such as which prospects on the list look strongest and why. It answers with the reasoning behind the list, because it can see the same company and title data you can.
3. Approve. Only once you pick the ones worth pursuing do you tell Claude to research them. This step reveals verified emails and phone numbers, and it is the one stage that spends Seamless credits.
4. Organize. The approved contacts get organized into a list inside Seamless, ready for the next step.
5. Act. From there you can draft a campaign, write outreach copy, and either send through Seamless's own connect tooling or push the list into whatever CRM or email tool you already run.

The best-practice advice from that same walkthrough is worth repeating for any small business testing this for the first time: start with small batches. Ask for ten or twenty prospects before you ask for a hundred, get comfortable with how the agent interprets your criteria, and only then scale up. It is the same discipline you would use with a new hire, except this one does not need three weeks of onboarding.

What this replaces on a normal week

What you have after this

Day one, you connect the integration and run a test batch of ten prospects, just to see how it interprets your criteria.

A week in, you have a real shortlist, a drafted campaign, and possibly your first reply from someone who was a stranger seven days ago.

A month in, the question stops being “how do I find leads” and starts being “which of these three ideal-customer profiles converts best,” because you can afford to test more than one.

Same team. Same subscription you were already paying for. A completely different pipeline.

That is also, not coincidentally, the exact kind of shift the 14-Day AI Readiness Sprint exists to find inside your business: one workflow, one week, one measurable change, before you touch anything else.