# Cut due-diligence research from 4 hours to 30 minutes: the Companies House AI build Source: https://getstartedwithai.net/tutorials/ai-companies-house-due-diligence-build Site: Get Started With AI Tagline: Build AI automations that actually run your business. Audience: solo professionals, freelancers, consultants and AI implementers in training Contact: hello@getstartedwithai.co.uk This Markdown companion is generated from the prerendered page for AI crawlers. # Cut due-diligence research from 4 hours to 30 minutes: the Companies House AI build Forty-five minutes per prospect is what a manual Companies House check costs and is the reason most UK practices skip it. The build below pulls the filings via the free Companies House API, runs a Claude or ChatGPT review pass against a six-point rubric and produces a one-page brief with the risk signals flagged. Each new prospect now takes 5 minutes and about 10 pence of API calls. Built once in 60 minutes; runs every time you hand it a company number. The brief stays internal as a working paper, not for sharing back to the prospect. ## What you need - A Companies House API key. Free, issued in 5 minutes from developer.company-information.service.gov.uk. Rate limit is 600 requests every 5 minutes. - A Claude Pro or ChatGPT Plus account. - A Zapier, Make or n8n account on a paid plan. n8n self-hosted works well here because the API is rate-limited and easy to call directly. - A standard prospect template: company number, contact name, the engagement type you are pricing. - 60 minutes uninterrupted to build, then 5 minutes per prospect on every check after. ## Step 1: Define the rubric (15 minutes) Before any code, write the checklist the AI runs on every company. Skip this and the output reads as a generic AI summary. Done well, the rubric is the value. - Trading status: active, dissolved, in administration, in liquidation. Date of last confirmation statement. - Filing pattern: are accounts and confirmation statements filed on time, or do late filing warnings repeat year on year. - Director changes in the last 24 months: appointments, resignations, disqualifications. - Charges register: floating charges, debentures, restrictive covenants. Note the lender and the date. - PSC (people with significant control) changes and any opaque ownership chains through other UK or offshore entities. - Linked entities through shared directors. Useful when the prospect is a small trading company connected to a larger group. ## Step 2: Build the API pull (20 minutes) The Companies House API is one of the cleaner UK government APIs. Three endpoints cover this build. 1. Get the API key. Set it as an environment variable in your automation tool. 2. Endpoint 1: GET /company/{number} returns the registered office, company status, incorporation date and the list of officers. 3. Endpoint 2: GET /company/{number}/filing-history returns the last 100 filings. Pull the 24 most recent. 4. Endpoint 3: GET /company/{number}/persons-with-significant-control returns the PSC list. 5. In Make or n8n, add three HTTP modules in sequence. Each takes a basic auth header with the API key as username and an empty password. 6. Pipe all three responses into a single text blob with section headings. ## Step 3: Build the Claude review prompt (15 minutes) Claude is the better tool here because the input is long and the rubric needs to be applied consistently. The prompt below is the production version. 1. Open Claude Pro. Create a Project named UK Company Due Diligence. Upload the rubric from Step 1 and one fully-completed example brief. 2. System prompt: you are reviewing a UK company filings extract for a prospective client engagement. UK English. Apply the six rubric checks. Output: one-page brief with a risk rating (Green, Amber, Red) and 3 to 6 bullet points per section. 3. Add the rule: do not infer beyond the data. If a rubric check has no data, say so. Never invent a director, a charge or a date. 4. Test against three companies: a clean active trading company, one in voluntary liquidation, and a dormant company. Confirm the output reads correctly for each. ## Step 4: Wire the brief to your inbox (10 minutes) 1. In Make or n8n, after the Claude module add an email or Notion action. 2. Email subject: DD brief for {{company name}} ({{company number}}) - {{rating}}. The rating in the subject is what makes this useful when 10 prospects come through in a week. 3. Body: the Claude brief plus a link to the raw API JSON in a Drive folder for audit. 4. Trigger: a Google Form or Slack message with the company number. ## Step 5: First production run Run the build on the last 5 prospects you accepted. Compare the AI brief with what you knew at the time of acceptance. The first run usually surfaces something missed by the eyeballs-only check, often a late filing pattern or a director who resigned a fortnight before incorporation of a new entity. ## What you ship A three-call API pull, a Claude rubric prompt, an inbox brief and a Drive audit folder. Five minutes per prospect for a HMRC-quality first read; 10 pence in API calls. Resold to corporate finance practices the same build prices at £75 to £150 per check; bundled into accountancy onboarding it pays back in declined engagements that would have absorbed bad debt. ## What to build next Layer in the AI Contract Review build for the engagement letter stage, then the MTD quarterly review build for the live client work. Or move into the WIRE Method to package the DD plus contract pair as a productised onboarding service. ## Related builds More tutorials in Industries. Each one runs in 60 to 90 minutes and ships a finished asset. ### Reclaim 8 hours a week as an accountant: the AI build for practice work 30 to 60 minutes. Solo Pro / Implementer. [Open tutorial](/tutorials/ai-for-accountants-practice-build) ### Cut 12 hours a week of legal admin: the AI build for solicitors and small firms 30 to 60 minutes. Solo Pro / Implementer. [Open tutorial](/tutorials/ai-for-lawyers-firm-build) ### Replace 6 hours a week of session admin: the AI build for coaches and therapists 60 to 90 minutes. Solo Pro. [Open tutorial](/tutorials/ai-for-coaches-therapists-build) ### Reclaim a day a week as a property agent: the AI build for letting and sales 60 to 90 minutes. Solo Pro / Implementer. [Open tutorial](/tutorials/ai-for-real-estate-agents-build) ## Questions people ask before they start ### Is the Companies House API free? Yes. Sign up at developer.company-information.service.gov.uk, request a key, use it within the published rate limits. There is no paid tier and no API contract; treat it as a free public service. ### Does this replace a full corporate finance due diligence? No. It replaces the first 45-minute screen that decides whether full DD is worth the spend. The full DD layer (forensic accounts review, KYC, sanctions screening) sits above this build, not inside it. ### Can I run this on overseas companies? No. Companies House is a UK register. For Irish companies use the CRO; for US companies, EDGAR plus the relevant state register. The same build pattern works on each, but the API and the rubric change. ### What about KYC and AML checks? Different layer. KYC and AML need a vendor like SmartSearch, Onfido or Veriphy with regulator-recognised data sources. This build sits before KYC; it decides whether to bother running KYC on the prospect. ### Is the brief good enough to share with the client? Internal use only. Companies House data is public, but the AI commentary is for your engagement decision, not for sharing back to the prospect. Treat the brief as a working paper.