How to Build a Target Company Profile, Then Automate a List of Companies That Match It
In short
Start with the customers you already have, not with a tool. Write down what your best ones have in common, split it into must-haves, signals and exclusions, and test it by hand on about twenty companies. Then automate the finding: pull candidates from public sources such as the Companies House register and company careers pages, filter and score them against your written rules, and have a person approve the list. A scheduled workflow does the repeat work. Finding the named person at each company is a separate job, covered at the end.
How do I build a target company profile before using an AI tool to find leads?
Build it from the customers you already have. List your best ten or so, write down what they share, and turn that into rules a stranger could apply. An AI tool can only find companies that match a profile you have written down. If the profile is vague, it will return a long, confident list of the wrong companies.
Here is the order I would use.
1. Pick your best customers, not all of them. Best means they pay on time, stay, and were easy to win. If you have also had a bad customer, note what made them bad. That becomes an exclusion.
2. Write down what they share. Look at the same handful of fields for each one.
| Field | The question to ask | Where you can check it |
|---|---|---|
| Location | Where are they based? | Company register, company website |
| Company type and age | Limited company, LLP, how long trading? | Company register |
| Sector | What do they do? | Register SIC code as a first cut, then the website |
| Size | How big are they? | A tool database or the company's own site. Check which is reliable for your market |
| How they work | What do they sell and to whom? | Company website |
| Trigger | What was happening when they bought from you? | Careers page, news, their own site |
3. Split it into three lists.
- Must-haves. A company without these is out. Keep them few. Location and type are usually must-haves.
- Signals. These raise a company's score but are not required. A recent job post for a role your product helps is a signal.
- Exclusions. Things that rule a company out even when it fits. Existing customers, competitors, and the kind of customer that went badly.
4. Check every rule can be tested from a public source. If you cannot say where a rule would be checked, a machine cannot check it either. "Well-run" is not a rule. "Has a careers page with an open role in operations" is.
5. Test it by hand on about twenty companies. Take twenty, apply the rules yourself, and see how many you would really want. Time yourself and write down what you did at each step. That list of steps is what you will automate. If you cannot get good results by hand, automating the process will only get you bad results quicker.
What does a profile look like written down?
Here is an invented example, only to show the shape. A bookkeeping firm sells monthly accounts to small UK limited companies.
- Must-have: UK limited company, registered office in Yorkshire, incorporated within the last five years, status active.
- Signals: advertising for a finance or admin role, a website that sells services rather than products.
- Exclusions: other accountancy firms, companies that are already clients, dissolved or dormant companies.
That is one paragraph of rules, and every line can be checked.
What if I have no customers yet?
Use your best conversations instead. Write down who you have talked to that was a close fit, and why. Treat the profile as a guess, and expect to rewrite it after the first fifty companies. The point is to write it down so that you can see which rule was wrong.
What's the best way to build a list of companies that fit a specific target profile automatically?
The best way is a short pipeline, not one clever tool. Collect candidates from public sources, filter them against your written rules, check each one against the company's own website, and have a person approve the result. Which source you use matters less than the rules you apply to it.
I would not trust any one source on its own. Each one answers a different part of the profile, so it helps to know what each can and cannot tell you.
What can the Companies House register tell you?
It shows registered company information. Check operating location and trading activity separately, because a Yorkshire registered office does not show that a company operates in Yorkshire. The register holds public information about companies incorporated in the UK, and it also covers limited liability partnerships and community interest companies.
There are three ways in.
- Advanced company search. On the Find and update company information service, the form lets you filter by company name (includes or excludes), registered office address (a full or partial address), incorporation date, company status, a SIC code for nature of business, and company type. A 2023 Companies House blog post says you can download results as a spreadsheet. Check the page for the current limit.
- The free company data product. This is a monthly snapshot of live companies, supplied as CSV files inside ZIP files. It includes company type, registered office address, SIC code, company status, the dates of the last and next accounts and confirmation statements, and previous company names. It is compiled up to the end of the previous month and made available within five working days of month end.
- The API. The Companies House API retrieves information about limited companies. You register a user account and use an API key. Its rate limit is 600 requests in a five-minute period, per application.
What it does not give you is anything about what a company sells or how it works. Nothing on the data product's field list says what a company does day to day. Treat a SIC code as a first cut and check the website. Use the register to build the long list by area, type and age, then use something else to narrow it.
How can job posts help you find matching companies?
A job post is a signal, not a fact about the whole company. It can show that a company is growing, or has a problem your product helps with. It cannot show that it will buy.
There are two ways to use them.
- Read the company's own careers page. This is the plainest source. Fetch the page, store the link, and note the date. Then you can see exactly what the signal was.
- Use a tool that filters on postings. Apollo's API documentation for its Organization Search lists filters for the job titles in a company's active postings, the locations of the jobs, the minimum and maximum number of active postings, and when jobs were posted. Check what your own plan allows before you build around it.
Posts go stale, and a role may already be filled. Treat a post as a reason to look at a company, not as proof of need. A signal with a link you fetched yourself is worth more than a summary of "growth signals" from an AI model with no link, which is the other thing I would avoid.
Where do database tools fit?
They are the quickest way to filter by size, sector and location in one go. A database such as Apollo lets you do the first cut in minutes. The weakness is that you are limited to the fields it holds, and you do not see how it decided a company fits. I compare the options in Apollo alternatives for UK B2B teams, and in whether to use a tool or a custom workflow.
Where do the company's own pages fit?
They are your check. Once you have candidates from the register or a tool, read each company's own website against your rules. This is where an AI model earns its place. It can read a page and answer each rule with yes, no or unknown, and quote the line it relied on. It must be allowed to say unknown. If it is forced to answer, it will guess.
How do I set up an automated system that finds companies matching our customer profile?
Set it up as a loop with five stages: profile, collect, score and check, approve, and hand on. Build it by hand for twenty companies first, then put it on a schedule. A workflow tool such as n8n, which we use, or a script run on a timer, will do the repeat work. The profile and the approval stay with you.
Here is what each stage does.
- The profile. Your written must-haves, signals and exclusions, kept in one place the workflow can read. When the profile changes, the workflow changes with it.
- Collect. The workflow pulls candidates from your sources. For example, new companies from the register that fit your area and SIC code, and companies from a tool's filters that have an open role in your target function. Keep the rate limits in mind. The Companies House API allows 600 requests in five minutes.
- Score and check. Apply your must-haves first and drop anything that fails. Remove exclusions, including companies already in your customer list or CRM. Then read each remaining company's website and mark each signal yes, no or unknown, with the quote and the link. Add up a simple score.
- Approve. A person looks at the list before anything is used. Show the score, the reasons, and the links, so approving takes seconds per company. A wrong company that slips through costs far more later than it does here.
- Hand on. The approved companies go into your CRM or sheet, ready for the next step.
Then put it on a schedule. Weekly is plenty for most teams. Because the loop remembers what it has already seen, you review only the new companies each time.
Log why each company was in or out. When a company turns out to be wrong, that log tells you which rule to change.
What will go wrong?
- The profile is too wide. A list of thousands is usually a sign the rules are not rules. Tighten the must-haves.
- The source is out of date. The company data product is a monthly snapshot, so a very new company will not be in it yet.
- A signal is stale. Check the date on any job post.
- The model guesses. Make unknown an allowed answer, and require a quote and a link for every yes.
- Nobody reviews. An automated list that nobody reads is a faster way to contact the wrong people.
None of this is clever. It is the same work an intern would do, written down so a machine can repeat it, with the judgement kept where it belongs.
What should happen after you have the list of companies?
A list of companies is not yet a list of people. The next step is finding the named decision-maker at each one, and checking their work email. That is a separate job with its own checks, and I cover it in how to find the named decision-maker at a company, not a generic inbox.
When should you do this yourself, and when is Aigentic the right choice?
Do it yourself when the profile is simple, the list is a one-off, or you enjoy the work. A Companies House download, a spreadsheet and an afternoon will get you a long way, and a standard tool will cover most of the rest. Our consultancy is the wrong choice if you have not yet worked out who your best customers are, because nobody can automate a profile that does not exist.
We do two things in this area. Aigentic Intelligence is done for you: it finds companies that match a target profile, identifies the named decision-maker at each, and returns a verified work email. It is built. If you want your own rules built into your own workflow, that is a custom build, priced per proposal after we have seen what you need.
I would be wary of anyone who promises results before they have seen your profile. We are a young consultancy, founded in September 2025, and I will not quote client results here. I explain how to check that claim in how to check an AI consultancy has real client results.
Key takeaways
- Build the target profile from your best existing customers, and write it as must-haves, signals and exclusions.
- Every rule needs a place it can be checked. If you cannot say where, a machine cannot check it.
- Test the profile by hand on about twenty companies before you automate anything.
- The best way to automate the list is a short pipeline: collect from public sources, filter against your rules, check the company's website, and have a person approve.
- The Companies House register gives registered details such as registered office, type, status and SIC code. It does not tell you what a company sells.
- A job post is a signal, not proof. Fetch the page, store the link and the date.
- Let an AI model answer yes, no or unknown, and require a quote and a link for every yes.
- Finding the named decision-maker at each company is a separate step.
Sources
Fetched on 12 November 2026. Pages and limits change, so check them before you rely on them. This is general information, not legal advice.
- Companies House advanced company search form, for the filters (name includes or excludes, registered office address, incorporation date, company status, SIC code, company type): https://find-and-update.company-information.service.gov.uk/advanced-search
- Companies House blog, "Using the Companies House advanced search function" (26 April 2023), for what the register holds and the spreadsheet download: https://companieshouse.blog.gov.uk/2023/04/26/using-the-companies-house-advanced-search-function/
- GOV.UK, "Companies House data products", for the free monthly company data product, its fields and timing: https://www.gov.uk/guidance/companies-house-data-products
- Companies House Developer Hub, "Get started", for registering and API authentication: https://developer.company-information.service.gov.uk/get-started
- Companies House API, "Rate limiting", for 600 requests in five minutes: https://developer-specs.company-information.service.gov.uk/guides/rateLimiting
- Apollo API documentation, "Organization Search", for the job posting filters: https://docs.apollo.io/reference/organization-search
Start with the Sample Five, five leads in your niche with a named decision-maker and a verified work email.