Use AI for the matching and drafting work around hiring: spotting new roles across the portfolio, matching candidates to them, drafting outreach, summarizing interviews and keeping the pipeline current. Keep a person on every message that goes out and every judgment about a candidate. Start with one workflow, prove it works, then add the next.
When I talk to talent leads at VC firms, the hard part is rarely finding good people. They know good people. The hard part is the bookkeeping around them: which companies are hiring right now, which roles never got posted, who you already sent to whom, and which candidate from last spring would be perfect for the role that opened this morning. Once that context exists, most of it is a matching problem, and matching is something AI does well.
Speed is most of the value here. The talent help founders remember is a shortlist the day they post a role, not a polished search six weeks later. AI can make same-day the normal pace for a small talent team. But only if the data underneath is in order.
Get the data in place before the AI
AI matching on messy data produces confident nonsense: candidates sent to roles that closed last month, or to a company that just paused hiring. Before you build any workflow, make sure two lists stay current without anyone updating them by hand.
Open roles. Careers pages and LinkedIn posts cover the public ones. But many of the best roles are never posted. They show up in a board deck ("we need a VP Sales by Q3"), in a monthly update or in a partner's call notes. If your role list only reads careers pages, you're missing the roles where your help matters most.
Candidates. Resumes arrive in email, referrals come through partners and people introduce themselves at events. Keep one candidate database that the firm owns and can export, with consent and a note on where each person came from. If it lives in one recruiter's inbox, it leaves when they do.
With both lists current, the chain is short: new role → match against the database → draft shortlist → a person reviews and sends.
The first list mostly builds itself if you read what founders already send. New roles and hiring struggles show up in updates, board decks and public posts well before anyone asks you for help.
Five recruiting workflows worth automating
Here are the workflows I'd look at, roughly in order of how fast they pay off:
| Workflow | What AI does | What a person does |
|---|---|---|
| Role discovery | Watches careers pages, posts and founder updates, and flags new roles, including unposted ones | Confirms the role with the founder |
| Candidate intake | Reads resumes from email, tags skills, seniority and location, files them | Spot-checks the tags each week |
| Shortlists | Matches each new role against the database and drafts a ranked list with reasons | Picks who to send and adds their own read |
| Outreach drafts | Writes personal drafts to candidates and founders | Edits and sends every message |
| Interview follow-up | Summarizes feedback and reminds you to close the loop with candidates | Owns the relationship and the hard calls |
Start with role discovery or shortlists. They're where speed shows up first, and founders notice right away.
Drafts, not sends
This is the rule I'd hold hardest. AI drafts every message, and a person sends it. Candidates talk to each other, and so do founders. One clumsy automated message, like a senior engineer pitched for a sales role or a company name spelled wrong, costs more trust than a hundred good messages earn.
Drafts still save most of the time. Here's what a draft to a founder might look like in Slack:
Saw you're hiring a founding AE. Three people from our network look like a fit:
Candidate A: closed mid-market deals at a dev tools company, wants an early-stage role, based in your city.
Candidate B: first sales hire at a seed company that went on to raise a Series A, open to remote.
Candidate C: leads an SDR team today, ready for a closing role, strong references.
Want intros to any of them? I'll check they're still looking first.
The person sending adds what AI can't, something like "B is the one I'd call first, I've seen her work up close." That line is why founders trust your shortlists over a list from a recruiter they've never met.
Run a two-week pilot
Don't roll out five workflows at once. Pick one and measure it:
- Before you start: write down a baseline. For example, the time from a role going live to the first candidates landing in the founder's inbox.
- Week 1: run the workflow with every step visible. Review all the AI output before using it, and note what it gets wrong.
- Week 2: keep what held up. Fix the inputs behind bad matches, which usually turn out to be stale data rather than the model.
- Review: compare against the baseline. If time-to-shortlist dropped and founders responded well, keep it and pick the next workflow.
Track the misses too: wrong matches, outdated roles, awkward drafts. They tell you where the data needs work. (In my experience the fix is almost always upstream of the AI.)
What AI shouldn't do in recruiting
Keep people on anything that judges a person or depends on trust:
- Rejections and final decisions. A model shouldn't decide who's out.
- Screening on things unrelated to the job. Check what your matching actually uses, and make sure it isn't standing in for age, gender or background. Rules on automated hiring tools differ by location, so check them where your companies hire.
- Closing senior candidates. Executive hires close through relationships and real conversations.
- Delicate moments. A candidate who was laid off last week deserves a message written by a person.
What belongs in a lean VC talent stack?
Four layers are enough for most firms: a live view of open roles across the portfolio, a candidate database you own, an assistant that drafts matches and outreach from that data, and the email and calendar where the work already happens. Add an applicant tracking system only when volume demands it.
| Layer | What it's for | Examples |
|---|---|---|
| Role view | Every open role across the portfolio, kept current | A portfolio job board, a shared table |
| Candidate database | People you know, with skills, interests and source | Airtable, your CRM, a spreadsheet you own |
| Assistant | Matches, summaries and drafts from the two layers above | Claude or ChatGPT, connected to your data |
| Where work happens | Sending, scheduling and talking to founders | Email, calendar, Slack, WhatsApp |
Notice what's missing: a candidate portal that founders have to log into. Founders want candidates in their inbox, not another login. An applicant tracking system makes sense once you run many searches at once, but plenty of small talent teams don't need one yet. For which services to offer in the first place, see what portfolio companies need from VC talent teams.
How should talent leaders adapt hiring in an AI-driven market?
Help founders hire for judgment and AI fluency, not only headcount, and test both with real work samples. Expect more AI-written applications, so put more weight on conversations and references. Move faster than before, because strong candidates now get found and approached within days.
The profiles founders need are shifting. With smaller teams, each hire covers more ground, and working comfortably with AI tools is becoming part of most roles. How AI is changing startup leadership teams covers what that means at the executive level.
Applications are also noisier. Many resumes and cover letters are now at least partly AI-written, so they tell you less than they used to. Help founders lean on work samples, structured conversations and references. Coach them to ask candidates how they actually use AI day to day, and to listen for specifics.
And speed matters more than ever. Strong candidates get approached quickly, so a slow process loses them. A talent team that can send a shortlist the day a role opens has a real edge. That speed comes from the data and the workflows above. Longer hours won't get you there.