.avif)
.avif)
The EA hires that fail don’t fail at the job description — they fail at vetting. A polished profile and a good first call tell you almost nothing about judgment, follow-through, or discretion. Here’s the 12-point process that does, with a weighted scorecard, real work-sample prompts, and the security bar every candidate must clear.
The Direct Answer
Vetting a remote executive assistant means testing four things a résumé can’t show: judgment (scenario tests, not trivia), communication (written and live, under mild time pressure), follow-through (a real work sample with a deadline), and discretion (security posture, references, tenure pattern). Run the 12-point checklist below over two structured rounds plus a paid work sample, score it on the weighted scorecard (hire bar: 80/100 with all three gates passed), and budget honestly — done properly, the process takes 10–15 hours per serious candidate in our placement experience. And know the base rates: on open marketplaces you are the vetting layer — even Upwork’s own resources warn buyers about freelancer red flags and the risks of hiring freelancers. A managed service does this screening before you ever see a candidate — Oceans Talent accepts roughly 1% of applicants (first-party figure; see Sources & methodology), and that gap is exactly what you’re absorbing when you vet alone.
Why EA Vetting Fails (When It Does)
Executive assistant is a judgment role evaluated with task-role methods. Most founders screen for tools, typing, and pleasantness — then hand over their inbox, calendar, and confidential business context and hope. In our placement experience the miss rarely shows up in week one; it surfaces over the first month or two as dropped follow-ups, context that doesn’t accumulate, and the quiet realization that you’re now managing the person who was supposed to be managing the chaos.
The cost of a mis-hire isn’t the fee — it’s the quarter you lose to re-onboarding, re-explaining, and re-trusting. That’s why every step below is designed to surface how the person operates over time, which is the thing interviews are worst at.
Six Signals to Investigate Before You Interview
None of these is automatically disqualifying — each is a prompt to dig before you invest interview hours. Treat them as questions to resolve, not verdicts:
- Instant full-time availability. Sometimes it’s a contract that just ended — a legitimate and common reason. Investigate: ask what wrapped up and why. “Start tomorrow” plus “currently supporting 30 executives” is the combination that doesn’t add up.
- Vague ownership language. “Assisted with,” “supported,” “helped manage” — press for specifics. If the actual scope was task execution, that’s fine for a task role; for an executive operator you want demonstrated “I owned.”
- Rates far below market. Could be a geography difference (often legitimate for global talent) — or scope confusion, or a portfolio of simultaneous clients you don’t know about. Ask how the rate works and what it assumes about your share of their attention.
- Client overlap opacity. A freelancer who won’t say how many concurrent clients they carry is telling you your response times will be someone else’s priorities. Directness here is the signal; the number itself matters less than the honesty.
- No referenceable executives. Short careers and NDAs can explain a thin reference list — but anyone who has genuinely run an executive’s operating layer for years has a former principal willing to say so. Ask why if they don’t.
- Perfect-polish applications with generic substance. In 2026, fluent paragraphs are free. Specificity is the signal left — look for real names, real systems, real outcomes. Generic polish alone isn’t disqualifying; generic answers under follow-up questioning are.
The pattern to act on: multiple unresolved signals after you’ve asked. One good explanation is a candidate; three deflections is a decision.
The 12-Point Vetting Checklist
1. Structured scenario interview. Present three real situations (“two board members request the same slot while I’m in the air — walk me through it”) and score the decision process, not the answer. Improvisation reveals judgment; rehearsed answers reveal preparation.
2. Written communication test. Have them draft a decline-but-preserve-the-relationship email and a three-line status update from a messy paragraph of context. This is the daily job; test it directly.
3. Live coordination simulation. A 30-minute exercise: reprioritize a conflicted calendar with incomplete information and explain the tradeoffs. Watch what they ask before they act — in our screening experience, the quality of a candidate’s questions is one of the strongest predictors of on-the-job judgment (a heuristic from our placement work, not a measured statistic).
4. Paid work sample with a deadline. A real, bounded task delivered against a clock — prompts, time limits, payment guidance, and grading criteria are in the work-sample kit below. You’re testing follow-through and finish quality, which no conversation reveals. Gate: a failing work sample ends the process regardless of interview performance.
5. Judgment-under-ambiguity probe. Give a brief with a deliberate gap. Do they flag it, ask, or barrel through? You want flag-and-ask — silent assumptions are how confidential mistakes happen.
6. Tool and AI-fluency check. Not a software list — a demonstration: “Show me how you’d use AI to turn this transcript into an action list, and what you’d verify before sending.” The bar for modern EA skills includes orchestrating AI, not just tolerating it.
7. Security and confidentiality posture. Run the full security bar below — MFA, password manager, device controls, least-privilege access, AI confidentiality, offboarding deletion, incident escalation. Gate: pass/fail for anyone touching board, comp, or personnel material.
8. Reference checks with specific questions. Skip “were they good?” Ask former principals: “What did you stop checking?” “What broke when they were out?” “Would you hire them again at their new rate — and why the hesitation, if any?” Gate: at least one former principal (or direct manager for in-house candidates) must be reachable and substantive.
9. Tenure and trajectory pattern. Multiple sub-year engagements can indicate the context-loss pattern you’re hiring to escape — or a string of contract roles that ended on schedule. Ask about each departure; the explanations matter more than the dates.
10. Availability architecture. Concurrent clients, working hours against your timezone, and coverage when they’re sick or away. A solo freelancer’s honest answer to the backup question is “there isn’t one” — decide if you can live with that.
11. Context-retention test. In round two, reference details from round one without prompting. An EA’s core value is compounding context; some people demonstrably accumulate it and some don’t.
12. Motivation fit. “Why executive support, and why long-term?” You’re funding a multi-year context investment. Someone treating EA work as a waypoint to something else will take your operating knowledge with them.
For the interview-question bank to run inside this structure, use our executive assistant interview questions.
The Scorecard: Weights, Gates, and the Hire Bar
Score every candidate on the same instrument or the process degrades into vibes. Each checkpoint is scored 1–5 and weighted by how well it predicts on-the-job success (weights reflect our placement experience — adjust to your context, but adjust deliberately):
| # | Checkpoint | Weight | Gate |
|---|---|---|---|
| 1 | Scenario interview (judgment) | 12 | — |
| 2 | Written communication | 10 | — |
| 3 | Live coordination simulation | 10 | — |
| 4 | Paid work sample | 15 | Score ≥ 3 required |
| 5 | Judgment under ambiguity | 8 | — |
| 6 | Tool & AI fluency | 8 | — |
| 7 | Security & confidentiality | 10 | Pass/fail — fail ends the process |
| 8 | References | 10 | ≥ 1 substantive former principal |
| 9 | Tenure & trajectory | 5 | — |
| 10 | Availability architecture | 4 | — |
| 11 | Context retention | 5 | — |
| 12 | Motivation fit | 3 | — |
| Total | 100 |
Weighted score = sum of (score ÷ 5 × weight), for a 0–100 total. The hire bar:
- 80+ with all gates passed — hire-quality candidate.
- 70–79 with all gates passed — borderline: run a second, different work sample before deciding.
- Below 70, or any gate failed — pass, regardless of how likable the interviews were. The gates exist precisely because charm survives interviews and gates don’t.
Download the scorecard: ea-vetting-scorecard.xlsx (auto-scoring: enter 1–5 per checkpoint and the gates, and the verdict computes)
The Work-Sample Kit
The paid sample is the highest-weight checkpoint because it’s the only one that tests real work. Three prompts we recommend, sized so no candidate spends more than about two hours:
Sample A — Research brief. “Prepare a one-page brief on [company relevant to your business]: leadership changes in the last 12 months, product or pricing moves, and the three questions I should be able to answer before meeting their CEO. One page maximum.” Deadline: 48 hours. Expected effort: ~2 hours.
Sample B — Meeting-prep pack. Give a sanitized calendar excerpt and two or three email threads. “Produce a prep note for Thursday’s board call: attendees and context, open items, decisions needed, and a draft agenda. One page.” Deadline: 24 hours. Expected effort: ~90 minutes.
Sample C — Inbox triage simulation. Provide 15–20 sanitized emails. “Categorize by urgency and required action, draft replies to the three most important, and list what you’d escalate to me immediately and why.” Run live in 60–90 minutes, or async with a 24-hour deadline.
Payment. Pay every candidate who completes a sample — including the ones you don’t hire — at a flat rate that respects roughly two hours of their market rate (as a rule of thumb, $25–$75 depending on seniority and market). Unpaid “test tasks” filter out exactly the experienced candidates you want, and paying is what makes a demanding sample fair to ask for.
Time limits. Cap expected effort at two hours and say so explicitly. The deadline tests follow-through and honesty about capacity — not speed. A candidate who asks for a reasonable extension before the deadline is showing you exactly the behavior you want on the job.
Grading. Score each sample 1–5 on four criteria, averaged: (1) Followed the brief — scope, length, format as specified. (2) Judgment — did they prioritize what actually matters, and flag gaps rather than guess? (3) Communication economy — could you forward this to a board member as-is? (4) Complete and on time. A 5 is forwardable without edits and shows a judgment call you didn’t ask for but are glad they made. A 3 is competent but needs your polish. A 2 answered a different brief than the one you wrote — and a 2 or below fails the gate.
The Security Bar (Pass/Fail)
A remote EA touches your inbox, calendar, documents, and often board and comp material. Screen the security posture like it’s a systems decision, because it is. Every item below is a direct question in round two — and for anyone handling sensitive material, this checkpoint is pass/fail:
- MFA everywhere. Multi-factor authentication on every account they’ll touch for you — email, password manager, file storage. “I use it when required” is a fail; “I use an authenticator app by default” is a pass.
- Password manager, shared-vault workflow. Credentials move through a shared vault (1Password, Bitwarden or similar) — never over email, chat, or spreadsheets. Ask what tool they use today; the answer should be immediate.
- Managed or dedicated work device. Work happens on a device with disk encryption, auto-lock, and current OS updates — ideally one your company manages or a device dedicated to work, not the family laptop. Ask what device they’d use and who else has access to it.
- Least privilege by default. Delegate access instead of shared passwords wherever the platform allows (Google delegation, calendar sharing tiers, view-only links). A strong candidate should prefer scoped access — enthusiasm for holding your root credentials is itself a signal to investigate.
- AI confidentiality rules. They should be able to state a clear line: no confidential client or company material into consumer AI tools; AI-assisted work happens in approved tools under your policy. Ask “what would you never put into a chatbot?” and listen for a real answer.
- Offboarding data deletion. At the end of the engagement: access revoked, local copies deleted, and a written attestation. Agree to this in the contract, not at the exit.
- Incident escalation. The scenario question that matters most: “You realize you’ve sent a sensitive document to the wrong person. What do you do in the next ten minutes?” The only passing answer starts with telling you immediately. Candidates who lead with containment-and-hope fail — discretion means escalating fast, not hiding well.
The Freelancer Trap (Run the Math Before You Run the Process)
None of this checklist is free. Executed properly it’s 10–15 hours per serious candidate in our experience — sourcing, two rounds, a paid sample, references — and on open marketplaces you’ll run it repeatedly, because the platform’s screening is a search filter, not a vetting layer. That’s the freelancer trap: the rate looks like the cost, and the vetting, management, and replacement risk are the invisible line items. (The engagement-model math — freelance and fractional vs. full-time dedicated — is in the executive assistant cost guide, and the generalist comparison is covered in virtual assistant vs. freelancer.)
The alternative is to buy the vetting layer instead of building it. A managed service runs this entire funnel before you see anyone: Oceans Talent screens applicants down to a roughly 1% acceptance rate through structured assessments, work samples, and live simulations, then matches in about two weeks with an 86% first-match success rate and a 90-day integration program (all first-party figures — definitions in Sources & methodology; the full process is documented here). For what the managed route looks like inside a real company, the Ever + Other case study has the EA+ describing her own scope: “My job is everything for the company except the financial services part — mainly operations, hiring support, and go-to-market setup, plus personal tasks for the founder.” If you’re comparing providers rather than people, the provider-selection side — agency vs. marketplace vs. managed service — is covered in how to hire a remote executive assistant; this page vets the person.
Frequently Asked Questions
How do you vet a remote executive assistant?
Two structured rounds plus a paid work sample, testing judgment (scenarios), communication (written + live), follow-through (deadline work), and discretion (security posture + references). Score it on a weighted 100-point scorecard with three pass/fail gates; in our experience the full process takes 10–15 hours per serious candidate done properly.
What are the biggest red flags when hiring a freelance EA?
The signals worth investigating: instant availability, vague ownership language (“helped with”), rates far below market, opacity about concurrent clients, no referenceable former principals, and polished-but-generic applications. None is automatically disqualifying — availability can just mean a contract ended — but multiple signals that stay unresolved after you’ve asked directly should end the conversation.
Should I test candidates with real work?
Yes — a paid, bounded, deadline-bound sample capped at about two hours of effort. It’s the only step that tests follow-through and finish quality. Pay every candidate who completes one; unpaid “test tasks” filter out exactly the experienced candidates you want.
What should a candidate scorecard include?
Every checkpoint you test, weighted by how well it predicts on-the-job success, scored on the same 1–5 scale for every candidate — plus hard gates that can’t be averaged away: the work sample, the security screen, and at least one substantive reference from a former principal. Set the hire bar before you meet anyone (we use 80/100 with all gates passed) so a charming interview can’t move it.
What’s the difference between vetting a freelancer and using a managed service?
With a freelancer, you are the vetting layer — sourcing, testing, references, and replacement risk are yours. A managed service pre-screens (Oceans Talent accepts roughly 1% of applicants) and carries the match risk: about two weeks to a vetted match, an 86% first-match success rate, and a replacement path if it’s wrong.
How long should vetting take?
Run solo and disciplined, plan on two to four weeks per hire in our experience — often longer in practice, because sourcing and scheduling stretch. Through a managed service the vetting is amortized before you arrive; Oceans Talent typically matches in about two weeks.
Sources & Methodology
First-party data. The Oceans Talent figures on this page — the ~1% acceptance rate, 86% first-match success rate, and approximately two-week matching time — are first-party operational metrics from our internal hiring and placement data. Working definitions: acceptance rate is the share of EA applicants offered a place on the Oceans Talent bench after the full assessment funnel; first-match success is the share of placements where the client continues with their first matched EA beyond the 90-day integration program; matching time is the typical elapsed time from client kickoff to a confirmed match. Our screening and matching process is documented at how we hire.
Heuristics, labeled as such. Figures presented as “in our experience” — the 10–15 hours per candidate, the two-to-four-week solo timeline, the question-quality observation, and the scorecard weights — are practitioner heuristics from Oceans Talent’s placement work, not measured statistics. Use them as starting points and calibrate to your own process.
External sources. Upwork, “Red Flags To Consider When Hiring a Freelancer” and “The Risks of Hiring Freelancers and How To Navigate Them” (updated July 2026). Case study. Ever + Other, quoted above.
Next step: See exactly what a 1%-acceptance vetting funnel looks like at how we hire — or skip the 10–15 hours per candidate and meet a pre-vetted executive assistant.
