ACTION_ID: get_employees_by_company_using_ai_ark NAME: AI Ark: Find Employees at Company CATEGORY: Lead Sourcing CREDITS: 0.2 per person returned Find people at one company through AI Ark. Returns a structured array of people — name, LinkedIn URL, current title, seniority, location, role start date and company identifiers — capped at the requested count. Its filters go past the native finder's: past titles, excluded title words, department, tenure in current role, spoken language and city-level location. It bills 0.2 credits per person, the same as get_employees_by_company_using_floqer_native — use it when you need one of those filters, not as the default finder. 1. INPUTS Multi-value text fields share one rule: separate values with semicolons; commas also work when no semicolon is present. Location fields take semicolons ONLY, because place names contain commas, and headcount labels contain commas too ("10,001+"). Some input names have changed. The old names (company_name, number_of_people, exclude_current_titles, past_titles, function_department, profile_language, industry, advanced_raw_ai_ark_filters_json, use_personal_api_key) are still accepted; use the names below. A mapped variable that is blank on a row leaves that filter unset. Exception: a variable INSIDE raw_ai_ark_filters_json. ai_ark_api_key (type: radio, optional, default false) false (default): runs on Floqer's AI Ark key, billed in Floqer credits. true: runs on the workspace's own AI Ark key; AI Ark bills 0.5 of its own credits per person and Floqer bills nothing. When the company is found but no employees match, the company check also costs 0.1 AI Ark credit. true with no AI Ark connection fails the row as an error. company_identifier (type: string, required) The company's domain or LinkedIn company URL. A path is ignored ("https://stripe.com/pricing" = "stripe.com"). ⚠ A domain also matches unrelated small companies that list it as their website. The real company's staff crowd them out, so they appear only when your filters exclude those staff — and the cell still reads `complete`. Guard with KEY NOTES before fanning out. A LinkedIn company URL matches exactly. Only linkedin.com/company/ is accepted — /showcase/ pages and percent-encoded (non-Latin) slugs error; pass the domain instead. A stale slug fails as "no company matching", so retry with the domain. number_of_employees (type: string, required) How many people to return, 1-100, as a string ("5"). Values above 100 are capped; a fraction is truncated ("2.5" returns 2). Billed per person returned — set the smallest number that meets the need. job_titles (type: string, optional) Current titles, fuzzy matched on words: "Head of Revenue" also returns "Head of Product Marketing, Revenue Suite". Prune with exclude_job_titles. A title containing a comma ("VP, Sales") needs semicolons between titles. exclude_job_titles (type: string, optional) Words; anyone whose current title contains one is dropped. Example: "product; finance; marketing; intern". past_job_titles (type: string, optional) Titles from the person's previous roles, fuzzy matched. seniority (type: multiDropdown, optional) Any of: Owner, Founder, C-Suite, Partner, VP, Director, Manager, Senior, Entry, Intern, Unpaid (case-insensitive). Configure body: ["VP", "Director"]. From a variable: "VP; Director" or "VP, Director". There is no Head level; reach "Head of X" through job_titles. departments_functions (type: string, optional) Departments or functions, e.g. "sales; product management", in AI Ark's classification — "Sales" can include solutions engineers. person_location (type: string, optional) — SEMICOLONS only Where the person lives: country, state, city or region, e.g. "Germany; Paris" or "Berlin". Matched as exact text, case-insensitive. ⚠ A comma ("Germany, Paris") or a typo ("Germny") does not error: it matches nobody and the cell fails with "none of its employees matched these filters (Person Location)". languages (type: string, optional) Languages the person lists, e.g. "english; german". started_current_role_from (type: date, optional) started_current_role_to (type: date, optional) When the person started their current role, as "YYYY-MM-DD" ("YYYY-MM" and an ISO timestamp also work), typed or mapped from a column. From = earliest start, To = latest; leave To empty for "up to the present". Either can be set alone. Matched by whole months against any role held now, side roles included. Any day of the current month works (a later To date means "up to now"); a later month, another format, or From after To errors. years_in_current_position (type: dropdown, optional) How long the person has held their current role: lt_3m (less than 3 months), 3_6m, 6_12m, lt_1 (less than 1 year), 1_2, 3_5, 6_10 or gt_10 (more than 10 years). Ignored when started_current_role_from or started_current_role_to is set. ⚠ The bucket does not follow the returned current_role_start_date: lt_1 returned a person whose role began three years earlier. Use started_current_role_from / _to instead, and check current_role_start_date downstream when tenure matters. company_headquarters (type: string, optional) — SEMICOLONS only company_industries (type: string, optional) company_type (type: multiDropdown, optional) company_headcount (type: multiDropdown, optional) — SEMICOLONS only company_keywords (type: string, optional) These filter the person's current company — the one company_identifier already names. Leave them empty: a value that fits the company changes nothing, and one that excludes it returns only the lookalikes described under company_identifier. Vocabularies: company_type: Privately Held, Public Company, Self Owned, Self Employed, Partnership, Non-Profit, Educational Institution, Government Agency company_headcount: 1-10, 11-50, 51-200, 201-500, 501-1,000, 1,001-5,000, 5,001-10,000, 10,001+ (the labels contain commas, hence semicolons: "51-200; 10,001+") company_industries names that contain a comma (e.g. "technology, information and internet") are kept whole even in a comma-separated list. raw_ai_ark_filters_json (type: string, optional) Extra AI Ark people-search filters as JSON, merged over the fields above, e.g. {"contact":{"location":{"any":{"include":["germany"]}}}}. Invalid JSON errors. The company always comes from company_identifier; an `account` domain filter here is ignored without a warning. ⚠ A variable inside the JSON that is blank on a row FAILS the row. Put per-row values in the dedicated fields, where blank = unset. ⚠ A raw key the fields also build replaces their filter: {"contact":{"experience":…}} dropped the job_titles filter (other titles came back), while {"contact":{"location":…}} kept it. A raw experience.current.duration window was not applied; filter tenure with started_current_role_from / _to. 2. OUTPUTS Some outputs have been renamed: no_of_people is now no_of_employees and people is now list_of_employees, for every action. Per person, country_name is now country; actions that stored the old column name keep country_name (Get Action's outputs lists the names an action uses). References written with the old names ({{.no_of_people}}, {{.people.country_name}}) still resolve; use the new names in new references. The responseStructureIds are unchanged, so running workflows are not affected. ai_ark_id (AI Ark's person id) is returned only by actions added before this change, and {{.people.ai_ark_id}} still resolves on them. no_of_employees (type: number) People actually returned; below number_of_employees when the filters narrow the pool. list_of_employees (type: structured_array) — one entry per person: - first_name, last_name, full_name — last_name can be a bare initial ("H.") on limited profiles, and a middle initial can sit in first_name. Names can be in native script (Japanese, Arabic). - linkedin_url — the person's profile. - job_title, headline — as written on the profile, any language. - company_name — verbatim, including native-script names. - company_website — bare domain ("stripe.com"), no scheme. - company_linkedin — full company URL; the field to guard on. - country (country_name on older actions) — Title Case. - location — "City, State, Country, Continent". - seniority — AI Ark's lowercase output labels (c_suite, vp, head, director, manager, senior, mid_level, entry, founder, unknown…), NOT the input vocabulary: head and mid_level cannot be fed back into the seniority filter. - current_role_start_date — month precision, "YYYY-MM-01". 3. HOW TO CONFIGURE Configure Action body: { "inputs": { "company_identifier": "{{input.company_domain}}", "number_of_employees": "5", "job_titles": "VP Sales; Chief Revenue Officer", "exclude_job_titles": "assistant; intern", "seniority": ["VP", "C-Suite"], "person_location": "United States; Canada", "started_current_role_from": "2025-01" } } Dropdown values are not checked when you configure — a bad value or separator surfaces only as a cell `error` at run time. Copy them from §1. 4. KEY NOTES - Guard against lookalike companies whenever you pass a domain AND a narrowing filter: keep only people whose company_linkedin matches the account's LinkedIn URL (or company_name, when you have no URL) before outreach. Unfiltered pulls are filled by the real staff, so the guard only bites when filters exclude them. Parse recipe in JS EXTRACTION REFERENCE. - No ranking and no paging. People are not ordered by seniority or title-match quality, and the same inputs return the same people on every run. To get more people raise number_of_employees; to get different people change the filters. - Cell outcomes (each sets row_status to has_failures, and halts the row's chain by default — continue_workflow_if_action_fails, concepts.txt): error An input value is invalid: a dropdown value, number_of_employees below 1 or not a number, a LinkedIn URL that is not a /company/ page, raw JSON that does not parse, a malformed start date, one in a later month, or From after To. The message names the field and lists the allowed values. error No AI Ark connection with ai_ark_api_key true, or the workspace's AI Ark key was rejected, is out of AI Ark credits or is rate limited. failed "AI Ark has no company matching ." — unknown domain, or a renamed or closed LinkedIn page: retry a URL with the domain. failed "AI Ark found , but none of its employees matched these filters ()." — loosen the named filters; check location and industry spelling first. With no filters set it ends "matched the search." failed "Provider AI Ark had no data for this input.": no one matched and the company check could not run; run the row again to learn which. failed "Required input is missing for …" — the mapped ROW value is blank. The action itself is configured, whatever the message says. 5. WHERE IT FITS IN A WORKFLOW input (account list — domain + company LinkedIn URL per row) -> get_employees_by_company_using_ai_ark -> push_data_to_sheet (one row per person; carry the account's LinkedIn URL along as a column) -> drop rows whose company_linkedin differs from the account's -> person_verified_work_email_waterfall -> outreach. 6. WHEN TO USE - You need a filter the native finder lacks: past titles, excluded title words, department / function, tenure in current role, spoken language, or city-level location. 7. WHEN NOT TO USE Title, seniority and country filters are enough -> action-detail/get_employees_by_company_using_floqer_native.txt Same job, same 0.2 credits per employee. One specific person, a conditional fallback search (C-suite, else VP, else Director), or an ICP that standard B2B databases under-index -> see §7 WHEN NOT TO USE in action-detail/get_employees_by_company_using_floqer_native.txt JS EXTRACTION REFERENCE `list_of_employees` injects into a `formula` in the responseId pattern (action-detail/formula.txt §8.1) with SEMANTIC responseStructureIds (first_name, company_linkedin, …), like the native finder. This parses it and applies the lookalike guard: (() => { const norm = u => String(u || "").toLowerCase() .replace(/^https?:\/\/(www\.)?/, "").replace(/\/+$/, ""); const people = Object.values({{.list_of_employees}} || {}) .map(fields => Object.fromEntries( fields.map(f => [f.responseStructureId, f.value]))); return people.filter(p => norm(p.company_linkedin) === norm("{{input.company_linkedin}}")); })() Last updated: 2026-11-04.