AI Ark Similarity Search

根据种子公司的资料，使用AI Ark的人工智能相似度匹配来查找类似公司。提供一个公司域名或名称，即可获得按相似度得分排名的具有相似企业画像的公司列表。当用户说“找到与...类似的公司”、“寻找相似公司”、“找到更多像...这样的公司”、“像Stripe这样的公司”，或者希望从种子公司或最佳客户那里扩展目标账户列表时，请使用此功能。这是AI Ark的独特卖点——Apollo没有与此相当的功能。每次搜索大约需要2.5个积分。最多接受5个种子域名。

Available Context

@_platform-references/org-variables.md

AI Ark Similarity Search

Goal

Find companies similar to a seed company using AI Ark's AI-powered similarity matching engine. This uses the lookalikeDomains parameter — AI Ark's unique differentiator with no Apollo equivalent.

Credit Cost

Each similarity search costs ~2.5 credits (uses the company search endpoint with lookalikeDomains). Use preview_mode: true to fetch 5 results first (same cost) before running a full search. Always warn the user before executing.

Required Capabilities

AI Ark API: Similarity search via ai-ark-search edge function (action: 'company_search' with lookalike_domains)

Inputs

seed_company_domain: Domain of the target company (preferred). Up to 5 domains for multi-seed matching. Examples: 'stripe.com', 'notion.so', 'hubspot.com', 'figma.com', 'salesforce.com'
seed_company_name: Company name fallback if domain unknown
match_count: How many lookalikes to find (default: 25, max: 100)
preview_mode: Fetch 5 results first for quick validation

Execution

Identify the seed company domain (ask user if unclear)
Warn the user: "This search will cost ~2.5 AI Ark credits. Proceed?" (or offer preview first)
Optionally run with preview_mode: true to show 5 results
On confirmation, call ai-ark-search with seed domain(s)
Present results as a table for review

Output Contract

Return a table with columns:

Company Name, Domain, Industry, Employee Count, Employee Range, Location, Description, Technologies, LinkedIn, Website

Show total_count to indicate how many lookalikes exist in the database.

Key Differentiator

Uses AI Ark's lookalikeDomains parameter on the company search endpoint. Accepts up to 5 seed domains or LinkedIn URLs. AI Ark matches by firmographic profile similarity (industry, size, tech stack, growth signals) — not keyword matching.

Common Patterns

Seed from best customer -> find 100 lookalikes -> people search for VPs -> outbound
Seed from closed-won deal -> lookalikes -> pipeline expansion
Multi-seed: pass 3-5 top accounts as seeds -> find companies matching that ICP profile