• BLOG

38 Prompts for People and Company Research

· Autumn AI

These examples cover eight research workflows. Adapt the inputs and requested output to your project.

Sales prospecting

Build account lists from what is on the web right now: who holds the role today, who is hiring, what is in the stack.

  1. Build a list of 50 mid-market companies (200 to 2,000 employees) running Snowflake without dbt, identify the data-platform owner at each, and return a verified work email plus the evidence behind the stack claim.

  2. Find the VP of Data or Head of Analytics at every logistics company that raised a Series B since January, rank the accounts by data-team hiring momentum over the last two quarters, and attach the job posts you counted.

  3. Take these 200 accounts and flag which posted RevOps or data-engineering roles in the last 30 days, which champions changed jobs, and which started evaluating a competitor, citing every signal.

  4. Find every company that lists Segment in its security trust center or subprocessor page but has no CDP engineer on staff, then identify the most likely buyer for an analytics tool at each and explain why.

  5. For each account in this CSV, tell me whether the current CTO was in the role six months ago, who replaced whom, and link the announcement or profile change that proves it.

Startup discovery

See incorporations, stealth teams and fresh rounds the day they happen, months before the databases catch up.

  1. Find every Delaware C-corp incorporated in the last 14 days whose name or registered agent suggests AI infrastructure, resolve each to its founders, and pull their prior employers, LinkedIn profiles, and any live domain or landing page.

  2. Track stealth startups whose domains went live in the last 30 days where at least one founder left Stripe, Ramp, or Brex within the past year. Include DNS records, careers pages, and the earliest archived snapshot as evidence.

  3. List every seed round announced this quarter in fintech infrastructure with the lead investor, round size, participating funds, and founder backgrounds, citing the announcement or filing behind each cell.

  4. Show me companies in the current Y Combinator batch that have not yet announced publicly, using accelerator pages, founder profile changes, and new domains as signals. Rank by confidence and show the evidence.

  5. Which companies posted their first engineering job in the last 60 days, have fewer than 10 employees, and have a founder who previously worked at a company that exited for more than 500 million dollars?

Recruiting and talent

Source from real work, not keyword matches: repositories, papers, talks, launches, and who people have built with before.

  1. Find staff-level infrastructure engineers in the Bay Area who have contributed to a distributed database or storage project on GitHub in the last two years and are not currently at a company that raised in the last 12 months. Link the contributions.

  2. Identify machine learning engineers who published at NeurIPS or ICML in the last three years, now work at a company smaller than 200 people, and have a public project that shipped to users. Return the paper, the employer, and the project.

  3. Build a shortlist of 30 product designers who have led a zero-to-one launch at a developer-tools company, with the launch page or announcement as evidence for each.

  4. For this list of 40 candidates, tell me who has worked with each other before, at which company, and during which years. Show the overlapping employment records.

  5. Find former founders whose company shut down or was acquired in the last 18 months and who have not announced a new role, filtered to people with a backend or infrastructure background.

Investors and market maps

Map a category from primary sources: who is building, who is funding, who is leaving to start something new.

  1. Map every company building agent observability tooling. For each, give founders, founding date, total raised, lead investors, headcount trend over the last six months, and the two most recent product announcements with links.

  2. Which engineers have left OpenAI, Anthropic, or Google DeepMind in the last 90 days, and which of them show signs of starting a company: a new domain, a Delaware filing, a profile change to 'stealth', or co-founder pairing?

  3. For these 25 portfolio companies, report hiring velocity, executive changes, new customer logos found on their sites, and any pricing-page changes since last quarter, each with a source.

  4. List every investor who led a seed or Series A in developer infrastructure in the last 12 months, ranked by number of deals, with the companies they backed and their co-investors.

  5. Find companies in vertical AI for legal that have not raised institutional money, have at least five employees, and have a live product with pricing. Include founders' prior companies.

Risk, fraud and KYC

Resolve who is really behind a company or an application, and check every claim against a source.

  1. Verify this business: confirm incorporation, registered agent, officers, operating address, live domain age, and whether the listed executives have a public footprint consistent with the claimed roles. Flag anything that does not line up.

  2. For this applicant, check whether the stated employer exists, whether the stated title is consistent with public profiles, whether the employment dates overlap with other claimed roles, and cite the source for each check.

  3. Find every company registered to this address or this registered agent in the last two years, group them by shared officers or directors, and highlight clusters that look like shell structures.

  4. Given this list of 500 merchant applications, identify which have a domain registered in the last 30 days, no employees findable on the web, and an officer who appears on other recently formed companies.

  5. Cross-check this person's claimed education and work history against public sources, and return a field-by-field table with a confidence score and the evidence for each field.

Expert and vendor sourcing

Find the people who have actually done the thing, and the vendors that actually serve the segment.

  1. Find 20 people who have run payments infrastructure at a company processing more than a billion dollars a year, are no longer in that role, and have written or spoken publicly about it. Link the talk or post.

  2. Identify former regulators or compliance leads from the FDA or EMA who now advise or consult for medical-device startups, with their current affiliations and evidence.

  3. Which vendors provide SOC 2 audit services to companies with fewer than 100 employees? Build the list from customer trust centers and case studies, and count how many customers each vendor appears on.

  4. Find practitioners who have migrated a large codebase from Python 2 to Python 3 at a company with more than 200 engineers, using blog posts, conference talks, and repositories as evidence.

Founder and network research

Warm paths, shared history, and the relationships that do not show up in a CRM.

  1. Find technical founders who recently left infrastructure companies, appear to be building a new startup, and have a warm path to our team. Return the evidence behind every match.

  2. For these 100 target accounts, find the shortest warm path from anyone on our team to a decision-maker at each account: shared employers, shared schools, co-authorship, shared investors, or public interactions. Show the path.

  3. Build a profile of this founder: every company they have started or joined, every co-founder and the years they overlapped, investors across their companies, public writing, and open-source work. Cite each fact.

  4. Who are the people that appear most often as early employees across companies backed by this fund? Rank by number of portfolio companies and show the employment records.

Agents and API workflows

Prompts written for the API and agent skills: structured inputs, structured outputs, and evidence on every field.

  1. Enrich this list of 1,000 email addresses into people records with current employer, title, location, LinkedIn URL, and GitHub URL. Return a source for each field and a confidence score per row.

  2. Given a company domain, return a company profile with founders, founding date, headcount, funding history, technology stack, pricing page summary, and the three most recent public announcements, with sources on every value.

  3. Monitor this list of 300 companies weekly. Emit an event whenever one changes CEO, posts a first sales role, publishes a pricing change, or appears in a new funding announcement. Include the source URL in every event.

  4. For each row in this table of people, find any public talk, paper, or blog post from the last 12 months, summarise it in one sentence, and link it. Leave the cell empty rather than guessing if nothing is found.

  5. Resolve these 5,000 company names to canonical entities. Merge duplicates, attach the primary domain and legal name, and flag any name that matches more than one plausible company.