What companies at your stage posted for

Methodology and data

Who is included

• Seed-stage AI companies (124 companies; 105 with a headcount reading). Companies building AI-native products that announced a seed or pre-seed round in the last 12 months, 5 to 150 employees, selling to businesses or developers. Not consultancies, not agencies, not open-source projects without a company. Study: https://humaninthelooptalent.com/research/verticals/seed-stage-ai-companies/.
• Bonfire Ventures portfolio (91 companies; 90 with a headcount reading). Companies listed as current portfolio companies of Bonfire Ventures on https://www.bonfirevc.com/companies. Study: https://humaninthelooptalent.com/research/verticals/bonfire-ventures-portfolio/.
• Vista Equity Partners portfolio (88 companies; 70 with a headcount reading). Companies listed as current portfolio companies of Vista Equity Partners on https://www.vistaequitypartners.com/companies/. Study: https://humaninthelooptalent.com/research/verticals/vista-equity-partners-portfolio/.
• Valor Ventures portfolio (26 companies; 19 with a headcount reading). Companies listed as current portfolio companies of Valor Ventures on https://valor.vc/portfolio/. Study: https://humaninthelooptalent.com/research/verticals/valor-ventures-portfolio/.

Each group is a tracked sample, not a verified complete census. A company without a found posting may be hiring through other channels. Company shares use every tracked company in the group as the denominator, including companies with no posting found.

How the benchmark is computed

1. Each study's fact sheet carries twelve months of job postings per company, read from a job-postings archive and classified into role families; the benchmark reads those fact sheets and nothing else, so every figure here appears on a study page too.
2. A share is stated only when the group has 30 companies or more; below that the table shows counts. A share is the number of companies with at least one posting in the family divided by all companies in the group.
3. Role families are compared within one headcount band only when 30 companies in that band have postings; until then the page lists the size mix and compares the group whole. Where each group stands today: Seed-stage AI companies: the band with the most companies with postings (20–49 people) has 19, so the group is compared whole; Bonfire Ventures portfolio: the band with the most companies with postings (20–49 people) has 10, so the group is compared whole; Vista Equity Partners portfolio: no band has a company with postings yet, so the group is compared whole; Valor Ventures portfolio: no band has a company with postings yet, so the group is compared whole.
4. The family "leaning most into the last quarter" is shown only when its share of postings in the last 90 days beats the 25% a flat year gives, measured among companies below the collection cap, at 5 or more companies.
5. Postings in the last 60 days are counted by posting date within each study's window. The "other" family (roles that say nothing about direction) is left out of every table.
6. Medians, never means. Rules version benchmark@1.1.0 (src/lib/benchmark.ts, tested in test/benchmark.test.ts); the page is rebuilt whenever a study JSON changes.

Windows and limits

Collection limit: each study keeps the most recent postings per company up to its cap, so a company at the cap shows only its newest postings and every count is a floor. Counts describe observed postings, not unique vacancies or completed hires; reposts can represent the same search. 1950 postings across 329 companies, windows from 2025-10-14 to 2026-10-17.

How the studies were made

Original research computed from the signal layer of the recruiting CRM behind Human in the Loop Talent: public job boards, company news, funding announcements and headcount readings, collected continuously for every tracked company and recomputed every Sunday. A figure is published only when the sample supports it; the data behind every page is published as JSON next to it.

1. The category is defined in one sentence (who is in, who is out). A search plan turns it into a structured company-database query (keywords, products, funding stages, founding year, countries, size band) and six to ten web-search angles.
2. Three feeds enumerate the companies: the structured query against a company database; the list pages a researcher would read for each angle, fetched whole and read for every company they name; and bulk firmographic enrichment (headcount, industry, headquarters) for any company that arrived without them. The lists merge by company domain, then by name.
3. Companies that pass the inclusion gate (category, size band, geography) become tracked companies in the CRM, so the signal layer watches the whole category: job postings every weekday, news, funding and leadership changes. The last twelve months of each company's postings are read from a job-postings archive, so the study sees a year of hiring, not only what is open today.
4. After at least a week of observation a fact sheet is computed in code: headcount bands, twelve months of hiring by role family and quarter, open roles, open leadership searches, tools named in job posts, revenue-seat changes, funding and AI initiatives over 180 days.
5. The study is written from the fact sheet only. A second pass checks every number, name, title and date against the sheet and strips anything it does not contain; the claim-check result is shown on the page.
6. A census under ten companies is not published. Studies are rewritten when the census is at least a week old and the last study is older than four weeks.

Limitations

• The groups are not matched to each other: a portfolio spans sizes and sectors, a category does not. Shares are comparable within a group and indicative between groups.
• A company appears in a portfolio study because the investor lists it, and in a category study because it met the category definition when the census was taken; a company can sit in both.
• Postings are read from an archive of public boards; a company hiring only through referrals or agencies posts nothing and counts as not having posted.
• Nothing on this page states that a hire caused growth or that a company will hire; the page records what groups of companies posted over twelve months, not the order in which they posted it, because the archive keeps each company's most recent postings up to its cap.

Sources, licence and reuse

• The posting histories come from a third-party job-postings archive the research reads under licence. Per-posting records and per-company derived rows are not redistributed; this page publishes counts and shares per group, and each study page publishes its fact sheet.
• Study JSON: seed-stage-ai-companies.json (version vertical@1.0.0), bonfire-ventures-portfolio.json (version vertical@1.0.0), vista-equity-partners-portfolio.json (version vertical@1.0.0), valor-ventures-portfolio.json (version vertical@1.0.0).
• Licence for the figures on this page and the study fact sheets: CC BY 4.0. Reuse with attribution.

Cite this

Related: all intelligence, which roles precede growth, Seed-stage AI companies, Bonfire Ventures portfolio, Vista Equity Partners portfolio, Valor Ventures portfolio.