> For the complete documentation index, see [llms.txt](https://help.bizmachine.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.bizmachine.com/en/search-and-filtering/finding-similar-companies.md). # How to find companies similar to your best customers How to find companies similar to your best customers through the BizMachine MCP connector: the assistant finds what they share, then more of the same. *Updated: 2 October 2026* You can find companies similar to your best customers (lookalike companies) through the BizMachine MCP connector in a few steps: you give the assistant a list of your best customers, let it find out what they have in common, and use those traits to find more companies of the same profile in the BizMachine database. It is not a blind "find similar" button but a guided process in which you see, for every company, why it belongs in the selection. *** ### Why BizMachine has no "find similar" button A button that turns one company into a list of similar ones sounds tempting. The catch is what happens underneath it. An algorithm that looks for similarity on its own often connects companies by surface traits (the same NACE code, similar revenue) and misses what matters. The list looks nice, but half the companies do not actually fit your profile. So BizMachine takes a different route. First you find out why your best customers are good, and only then do you use those traits as a compass for the search. You gain two things: a more precise selection, and a reason why each company is in it. When a result does not fit, you know exactly which trait to adjust. You can do all of it in an ordinary conversation with your AI assistant. All you need is a connected [MCP connector](/en/mcp-connector/mcp-connector-overview.md). ### Before you start * **A connected MCP connector.** The assistant (ChatGPT, Claude, Microsoft Copilot Studio or another tool with MCP support) must be linked to your BizMachine workspace. The guides are in the articles [How to connect BizMachine to ChatGPT](/en/mcp-connector/connecting-chatgpt.md), [How to connect BizMachine in Claude](/en/mcp-connector/connecting-claude.md) and [How to connect BizMachine in Microsoft Copilot Studio](/en/mcp-connector/connecting-copilot-studio.md). * **A list of your best customers.** The companies that really work best for you and are similar to each other. Three to five is enough. {% hint style="info" %} If your best customers are very diverse (say, a small manufacturer and a large bank), split them into two or three groups and work through each one separately. With a mixed list, the assistant finds only the most general traits in common, for example the country and the size range, and the selection comes out too broad. {% endhint %} ### Step 1: Prepare your list of customers You can hand your list of customers to the assistant in three ways. Pick whichever suits you best. #### a) By pasting or uploading the list The simplest route. You just paste the company names or their Company IDs directly into the chat, or upload a file (Excel, CSV). It works regardless of what you have set up in Prospector. {% hint style="info" %} We recommend entering the **Company ID**. It is unambiguous and the assistant finds the company without follow-up questions. With names it may ask you to clarify when there are several companies with a similar name. {% endhint %} #### b) With a label in Prospector If you already have your customers marked with a [label](/en/lists-and-export/labels-as-a-workflow.md) in Prospector, you can point the assistant directly to it: "Take the companies from the My customers label." The connector reads labels, so you don't have to paste anything by hand. #### c) From your CRM If you have your CRM (for example Pipedrive) connected to the assistant alongside BizMachine, you can pull the list straight from there: "Take my customers from the CRM and find them in BizMachine." It depends on which tools you have connected to the assistant. {% hint style="warning" %} Shared labels may not always display correctly for colleagues. If the assistant does not see the label, use route a) and paste the list directly into the chat. {% endhint %} ### Step 2: Let the assistant find out what your customers have in common Now you simply ask the assistant to go through your customers in BizMachine and describe what connects them. This produces the profile of your ideal customer, the compass for the next search. Copy this request and add your list: ``` Here is a list of my best customers: [paste names or Company IDs]. Find each company in BizMachine and go through its profile. Then tell me what these companies have in common: - industry and main activity - size (number of employees, revenue) - region and type of organization - technologies they use - business indicators (activity, growth, reachability) - other indicators From that, build the profile of one ideal customer and list 3 to 7 traits that describe it best. ``` The assistant reads verified data from BizMachine about each company (not what it finds on the web or makes up) and returns a summary. Take your time to read it. This is the moment when you check that the profile matches how you know your customers. If something does not fit, tell the assistant and have it adjust the profile. ### Step 3: Find similar companies and refine the selection When the profile fits, you let the assistant find the companies that match it. Continue in the same conversation so that the profile stays in context: ``` Based on this profile, find more companies in BizMachine that match it. Leave out the companies I already have in my customer list. Proceed like this: 1. First tell me how many such companies there are in the database. 2. Then show a sample of 10 companies and, for each one, write why it belongs in the selection. 3. When that fits, prepare a list of 50 companies. ``` With this process you go from a rough estimate to a finished list. First you find out the size of the market (how many companies meet the profile at all), then you check a small sample to confirm that the selection makes sense, and only then do you have the whole list prepared. That way you avoid working straight away with hundreds of companies that you would end up discarding. Go through the sample and refine the selection based on it: * **Too many companies, or too general?** Add another trait to the profile: a specific region, technology or [signal](/en/advanced-features/company-signals-and-notifications.md) (say, "only companies that are hiring salespeople right now"). * **Too few companies?** Remove the strictest condition, or add another country to the region. * **Don't forget to exclude existing customers.** So that the companies you already have do not remain in the results, tell the assistant to leave them out, or use the [My labels (excluded)](/en/lists-and-export/excluding-existing-customers.md) filter in Prospector. ### Step 4 (optional): Sort the selection by priority with AI Prioritization The list from Step 3 is ready to use. If you want to know which companies to look at first, AI Prioritization takes you one step further: it gives every company a value from 0 to 100 according to how well it fits your use case, and adds a reason for each one. Prioritization always runs on a segment or a label, so the procedure is this: 1. **Have the assistant write the use case.** From the profile created in Step 2, the assistant writes a use case for AI Prioritization for you right away, that is, a description of your product and your ideal customer. The connector saves it to your workspace. 2. **Have the assistant save the selection to a label.** The connector can create a new label and add all the companies found to it right away. Just ask the assistant. Prioritization will then run on this label. 3. **Run AI Prioritization on the label.** In Prospector, open the label, click **Run AI Prioritization** and choose the use case the assistant wrote for you. 4. **Sort the companies by value.** Every company gets a value from 0 to 100 together with an explanation. You simply start from the highest ones, and you can see at once which companies to start with. The details are in the article [AI Prioritization](/en/ai-prioritization/ai-prioritization-overview.md), and tips for a good use case in the article [How to write a use case for AI Prioritization](/en/ai-prioritization/ai-prioritization-tips.md). {% hint style="info" %} The connector **writes** the use case, but you run the evaluation itself in Prospector. That is on purpose: prioritization works with dozens of data points about every company and runs over the whole label or segment at once. {% endhint %} ### Example: from three customers to a list of fifty Let's say you sell CNC machine servicing and your three best customers are engineering companies in the Moravian-Silesian Region with 50 to 250 employees. The assistant finds them in BizMachine and works out what they have in common: engineering production, medium size, their own production plant, CNC technology and active hiring of production staff. From that it builds a profile. Then it finds another fifty companies in the database that match this profile, leaves out your three customers, and for each one writes why it belongs in the selection. ### Troubleshooting | Problem | Cause | Solution | | ------------------------------------------------------- | ------------------------------------------ | --------------------------------------------------------------------------- | | Should I enter names or Company IDs? | Names can be ambiguous | Enter Company IDs. They are unambiguous, unlike a company name. | | The assistant does not see my label | Shared labels may not always display | Paste the list directly into the chat or upload a file (route a in Step 1). | | The results do not fit or are too general | The profile is too broad | Add a more specific trait: region, technology or signal. | | There are too few companies in the result | The profile is too strict | Remove the strictest condition or add another country. | | The assistant offers to download thousands of companies | The connector is not meant for bulk export | For large volumes, use Prospector or the Prospector API. | ### Frequently asked questions **How many customers should I enter?** There is no single right number. More important than the count is that the companies are similar to each other and really work for you. You can start with a few companies: three to five is enough. More well-chosen examples sharpen the profile. Just watch out for a company that does not fit in with the others. It can skew the result. **How is this different from AI Prioritization?** AI Prioritization sorts the companies you already have in a segment or label by how well they fit your use case. Here you actually start one step earlier: from your customers you first work out which segment to look for at all. Step 4 above shows how to combine the two. **Does it work for Slovakia, Poland, Hungary and Germany too?** Yes. One MCP connector gives access to all the countries your account has rights to. The depth of data differs between countries. **Does the assistant work with current data?** Yes. The connector reads the same data as Prospector: **11.7 million companies in five countries** *(as of June 2026)*, continuously verified from more than 100 public sources. **Can I save the result as a segment?** As a label, yes. As a segment, no. Ask the assistant to save the selection to a label. The connector can do that on its own. Prospector can build a segment only from filters, whereas this selection came from comparing profiles. If you want to keep watching for new similar companies, [build a segment in Prospector](/en/search-and-filtering/creating-segments.md) from the same traits the assistant described for you in Step 2. *** ### What next * [BizMachine MCP connector: what it does and what it is for](/en/mcp-connector/mcp-connector-overview.md) * [How to create a segment in Prospector](/en/search-and-filtering/creating-segments.md) * [AI Prioritization](/en/ai-prioritization/ai-prioritization-overview.md) * [How to exclude existing customers and avoid duplicate work](/en/lists-and-export/excluding-existing-customers.md)