The Guide to Consequential Hiring
Johnathan James · Founder, [solved] · 20+ years in talent, most of them leading technical searches for startups · Published October 7, 2026 · Last updated October 7, 2026
About this Guide
This Guide is for founders, CTOs, hiring leaders and talent teams at technology companies, making hires in Engineering & AI, Product, Sales & GTM or Leadership. It is written for whoever owns a hire the company can't afford to get wrong.
It follows one hire that matters from start to finish. The eight decisions, each linked to its chapter:
Eight decisions in four stages
Define
Reach
Prove
- 04
Read evidence when AI writes the application
credit what candidates produce live, and set an AI policy that matches the job.
- 05
Confirm identity
verify every remote candidate after the first real conversation and before your team interviews, with consent and in line with the law, and confirm their history through references you find yourself.
- 06
Structure the interviews
one outcome per interviewer, set questions, written scales, independent scores and a debrief on evidence.
- 04
Hire
Each chapter opens with a direct answer you can act on, then the detail, examples for all four specialties, and the sources behind every figure. Read it in order before a search, or go straight to the chapter for the decision in front of you.
Figures from other organizations are credited inline with their dates. Our own evidence comes from our team's search record from April 2023 to March 2026, defined on our How we measure page, and from our review of founder and hiring-manager posts published between October 2025 and October 2026. Every source carries a label: peer-reviewed research, industry research or guidance, vendor research, government and law, [solved] record, or practitioner review. Reading time is about 40 minutes.
Chapter 1: Define the outcomes the hire must deliver

What should my first engineer, product manager or sales hire own in year one?
Key points
- Write outcomes at 90, 180 and 365 days before the job description.
- Add the decisions the hire owns and what you will stop doing.
- The outcomes set the level, the pay, the interview plan and the first-year review.
Why outcomes come first
Most bad hires go wrong before the first interview, when the role is written. The a16z talent team finds that executive mis-hires usually come from misdiagnosing what the company needs, and advises hiring for what the business must accomplish in the next 12 to 18 months (a16z, 2023, updated 2025). The same pattern runs through consequential hiring in every specialty. In the posts we reviewed, founders and hiring managers describe:
- an AI engineer opening that turned out to be two jobs, model evaluation and production pipelines, filled in the end by two people;
- a first product manager given a goal with no authority to reach it, followed by the founders taking product back;
- a founding designer title that covered several distinct jobs;
- a first sales hire brought in to take over deals the founder had run personally for two years.
In each case everyone agreed on the title, and nobody had agreed on the job. Outcomes settle the job.
Write outcomes at 90, 180 and 365 days
- 90 days: the first result. The person understands the system, customers or pipeline they inherit, and delivers one bounded piece of work.
- 180 days: the main result. Something ships, closes or changes in a way the business can measure.
- 365 days: ownership. The person runs the area, and decisions reach you only when they cross limits you set.
Write each outcome as a result someone else could check: a number, a shipped system, a decision made. Activities such as "manage the roadmap" describe the job. Results describe success.
Worked examples
These are examples. Set the numbers from your own baseline.
From first result to ownership: worked examples
| Role | 90 daysFirst result | 180 daysMain result | 365 daysOwnership |
|---|---|---|---|
| Founding engineer, AI product | Evaluation suite for the core model in production; current quality and cost measured | Inference cost per request reduced by an agreed share at equal quality | Owns model releases, including rollback and monitoring |
| First product manager | 20 customer interviews; activation drop-off mapped | Redesigned activation flow live; day-30 retention moved by an agreed amount | Owns discovery and the roadmap, with founder input at set review points |
| Founding account executive | Ideal customer profile and qualification criteria agreed; first outbound tested | First deals closed that the founder did not source | A sales process documented well enough for the next hire to follow |
| VP Engineering | Delivery and team review complete; on-call and team structure agreed | Team grown to plan with release cadence held | Owns engineering hiring, budget and technical direction |
Show the data
| Role | 90 days | 180 days | 365 days |
|---|---|---|---|
| Founding engineer, AI product | Evaluation suite for the core model in production; current quality and cost measured | Inference cost per request reduced by an agreed share at equal quality | Owns model releases, including rollback and monitoring |
| First product manager | 20 customer interviews; activation drop-off mapped | Redesigned activation flow live; day-30 retention moved by an agreed amount | Owns discovery and the roadmap, with founder input at set review points |
| Founding account executive | Ideal customer profile and qualification criteria agreed; first outbound tested | First deals closed that the founder did not source | A sales process documented well enough for the next hire to follow |
| VP Engineering | Delivery and team review complete; on-call and team structure agreed | Team grown to plan with release cadence held | Owns engineering hiring, budget and technical direction |
Decide what the hire owns
A hire who owns a result but none of the decisions behind it will struggle, and so will you. The accounts we reviewed include product managers held to goals they could not steer, and engineering leaders held responsible for delivery while work was assigned around them. Three questions settle ownership before the search starts:
- Which decisions does this person make alone?
- Which decisions still need you, and how do they reach you?
- What will you stop doing once they start?
The third question carries the most weight. Hiring into a function you have run yourself means handing over decisions that used to be yours. a16z notes that recognizing someone else should own a function is easy with finance and much harder when it is the founder's own area, such as product for a technical founder (a16z, 2023).
Choose the level from the work
| Role | Fits when | The person owns |
|---|---|---|
| Founding engineer | The product is still taking form and the founder is the main builder | Core product, built alongside the founder, with broad scope |
| Senior engineer | The codebase exists and one area needs depth (data, infrastructure, models) | Delivery in a defined domain |
| CTO | The company needs one person accountable for technical direction, the engineering team and technical credibility with customers and investors | Technical strategy and the team |
| VP Engineering | The team is growing and delivery, hiring and process need an owner | Running the engineering organization |
| Founding account executive | Founder-led sales has repeatable wins and needs a second closer | Selling, with the founder still setting strategy |
| First sales leader | The sales motion repeats and a team must be hired and run | The sales team, targets and process |
Choose by the outcomes, then name the role. a16z describes hiring a strong executive at the wrong stage as a common failure (a16z, 2023). A title ahead of the work draws candidates who expect a team to lead. A title behind the work draws candidates who cannot carry the outcomes.
Name the trade-offs
- Speed or durability: ship to test demand, or build for scale.
- Depth or breadth: one specialist domain, or several areas at lighter depth.
- Revenue now or the playbook: close the deals in front of you, or build the process the next hire runs.
Write down which way each one goes for this role. Candidates and interviewers each assume an answer, and they rarely assume the same one.
Signs the outcomes are still unclear
- Two interviewers describe the role's main result differently.
- The description lists ten or more responsibilities across several levels.
- Nobody can say what success looks like at month six.
- The title came first, and the outcomes are being written to fit it.
- The role combines two jobs that rarely sit in one person, such as model research and production reliability.
To turn these outcomes into interview stages, explore the Interview Process Planner.
To write these outcomes into a brief any recruiter can work from, explore the Recruiter Brief.
In a [solved] search
Before any outreach, one of our senior talent partners works through these outcomes, decision rights and trade-offs with you in a live session. The result is written down as your Signal Spec: the one document that sets the level, the pay and the interview plan for the whole search. Start a Search, or Talk Through a Role first.
Sources (3)
Sources for this chapter
- Industry research or guidance: a16z (Jeff Stump and Shannon Schiltz). The First Principles of Executive Hiring. Published June 14, 2023, updated February 12, 2025.
- Practitioner review: Hacker News. Comment on Ask HN: Who is hiring? (June 2026) describing one AI engineer opening filled by two hires.
- Practitioner review: our review of founder and hiring-manager posts on LinkedIn and Hacker News, October 2025 to October 2026.
Chapter 2: Test the role against the market

How do I set salary, equity and scope before I start hiring?
Key points
- Check pay and equity against current data for the function, stage and region, and note how old each source is.
- For sales roles, test quota and ramp as well as pay.
- Before outreach, test the role on 10 to 15 people doing it at companies one stage ahead of yours.
Read pay data by function, stage and date
Pay data is only useful when it matches your role, your stage and the current year. Three published sources cover most early-stage technology roles.
Current startup medians (Carta, 2026). Across startups on Carta valued at $1 million or more, the median salary for new hires in Q1 2026 was about $177,000 in product and $167,000 in engineering. From January 2024 to February 2026, median individual-contributor salaries rose 6.4% and median initial equity grants rose about 11%. AI and ML engineers moved fastest: salaries up 9.1% and equity grants up 31%, and up 64% at startups valued between $1 million and $10 million (Carta, May 2026; over 900,000 salary and equity data points).
Seed-stage salary ranges (Kruze Consulting, 2024). From payroll records at more than 450 seed-stage startups:
| Role at a seed startup | Bay Area | Other areas |
|---|---|---|
| Senior engineer | $180,000 to $235,000 | $160,000 to $210,000 |
| Product titles | $130,000 to $185,000 | $110,000 to $175,000 |
| Mid-level sales (base) | $80,000 to $110,000 | $70,000 to $100,000 |
These ranges date from late 2024. Carta's 2026 data shows median startup salaries have risen about 5% since, so treat them as a dated reference point and check current data before you set a range.
Equity for the first five hires (Carta, 2024). From more than 8,000 initial grants, vesting over four years:
Equity for the first five hires
First
Second
Third
Fourth
Fifth
Show the data
| Hire | Typical range | Median |
|---|---|---|
| First | 0.5% to 4% | 1.49% |
| Second | 0.3% to 2% | 0.85% |
| Third | 0.21% to 1.2% | 0.50% |
| Fourth | 0.18% to 1% | 0.44% |
| Fifth | 0.13% to 0.8% | 0.34% |
Median equity grants have grown since this data was published (Carta, 2026). At Series A, the median company reserves 15% of fully diluted equity for its option pool, with a wide spread: 8.3% at the 10th percentile and 26.9% at the 90th.
Read every source with four questions. Which roles and stages does it cover? Which region? How old is it? Is the figure base salary, cash including variable pay, or equity as a percentage? Founding roles carry more scope than the median role in any dataset, so expect the upper half of a range, or more equity.
Sales roles: test the quota as well as the pay
For a sales hire, the quota decides whether the pay is real. The Bridge Group's 2026 study of account executives at 158 B2B companies reports:
| Measure | 2026 median |
|---|---|
| On-target earnings | $200,000 |
| Annual quota | $960,000 (SaaS: $875,000) |
| Quota as a multiple of on-target earnings | 4.6x, up from 4.2x in 2024 |
| Reps reaching quota | 48% |
| Ramp to full productivity | 6.2 months |
| Prior experience required at hire | 3.7 years |
These companies have established sales teams and repeatable motions. A founding account executive works without either: the first year includes building the process, the pipeline and the messaging the next hire will use. A quota copied from an established team assumes a motion that does not exist yet. Set a ramp period, a first-year target that reflects building work, and a clear split between base and variable pay. Candidates for founding sales roles ask exactly these questions: the quota-to-earnings ratio, the base and variable split, and what is protected during ramp.
Readiness checks by specialty
A role can pay well and still come too early. One question per specialty tests readiness:
Engineering & AI: Can you show a candidate the codebase, the stack and the core problem in the first conversation, and name the technical decisions they will own?
Product: Will you let this person override your feature ideas when customer evidence points elsewhere? If you plan to keep the final say on daily product decisions, the role you need is product operations, design or delivery, and it should be scoped and priced that way.
Sales & GTM: Have you closed repeatable deals outside your own network, with a value proposition someone else can deliver? Repeatability means customers buying for the same reason; customers who buy for different reasons are hard to hand to a new seller.
Leadership: Is there a team, a budget and a set of decisions ready for this person to own in the next 12 to 18 months?
Test the role on real people before outreach
The fastest check is a short list of people already doing the job.
- Find 10 to 15 people in the role at companies one stage ahead of yours.
- Compare their actual scope with your 90, 180 and 365-day outcomes.
- Estimate whether your pay and equity would move them, using the data above.
- Adjust scope, level or pay where the comparison breaks.
Three results tell you the role needs work: nobody holds the combination of skills you wrote; the people who fit earn well above your range; or the people within your range lack the experience your outcomes need.
To put your range, must-haves and what you would trade on one page before outreach, explore the Recruiter Brief.
In a [solved] search
Before anyone is contacted, one of our senior talent partners shows you 3 to 5 representative profiles of people doing this job today, compared with your outcomes and pay. This is the calibration slate: it tests the role, and you confirm the direction before outreach begins. Your shortlist comes later, from people who have been contacted, assessed and verified.
Sources (4)
Sources for this chapter
- Vendor research: Carta (Ashley Neville, Kevin Dowd, Josh Steinfeld). State of Startup Compensation H2 2025. Published May 4, 2026. Over 900,000 salary and equity data points.
- Industry research or guidance, reported by TechCrunch (Julie Bort). $132K to $149K, here's what seed-stage founders pay early employees. Originally published December 2, 2024. Reports Kruze Consulting payroll data from 450+ seed-stage startups and Carta data from 8,000+ initial equity grants.
- Industry research or guidance: The Bridge Group (Matt Bertuzzi). AE Models, Motions & Metrics: 2026 Research Report. Published June 22, 2026. Survey of 158 B2B companies, Q1 to Q2 2026.
- Industry research or guidance: a16z. The First Principles of Executive Hiring. Updated February 12, 2025.
Chapter 3: Reach specialists who are not applying

How do I find senior people who aren't applying?
Key points
- The people you want are working elsewhere. Map where the work is done today.
- Write to each person about their own work, with your outcomes, stage and pay range.
- Compare reply rates only when they are counted the same way.
Why job posts alone fall short for specialist roles
A post reaches people who are looking. Pin, citing Gallup, reports that 11% of U.S. employees are actively job hunting and another 40% are watching for openings (Pin, 2026). Posting has also become noisier: founders describe AI job-hunting tools scraping their posts and hundreds of generated applications landing within days. Gem's 2026 benchmarks, as reported by Pin, show direct sourcing producing 11% of hires from 2.6% of applications, roughly four times its share of applications.
Job posts still have a place. For a senior specialist or a first leader, you usually have to find the person yourself.
Map where the work is done today
Start from your outcomes, then ask who already produces them.
Engineering & AI: engineers who built the system you need at companies one stage ahead of you, maintainers of open-source projects your stack depends on, and authors of technical writing on the exact problem. Split AI roles before you map: model evaluation and production pipelines draw from different pools.
Product: product managers and designers who owned a workflow similar to yours at a B2B or product-led company, with shipped changes and measured results they can point to.
Sales & GTM: account executives who sold a comparable product to your buyer at an early company, ideally as one of the first sellers, because building pipeline from nothing is a different skill from working an established territory.
Leadership: leaders who have already run the next stage of your company once: a team of the size you will have in 12 to 18 months, at a company with your business model. Investors and board members are a strong route here; several founders in our review found their first sales or product leader through an investor introduction.
A map of 50 to 150 named people is a workable start for most specialist roles. It also tests the role (Chapter 2): if the map is thin, the role combines skills the market rarely holds together.
Write to each person about their work
Strong outreach has four parts, in this order:
- Why this person. One thing their work shows and why it matters for this role: a habit, a system they built, a market they sold into.
- What the role must deliver. The 180-day outcome from Chapter 1, in one or two sentences.
- The facts. Stage, funding, pay range and equity range. Specialists judge a role on these first, and stating them saves both sides a call.
- One easy next step. A short conversation with the founder or the person they would report to, framed around what the candidate wants next.
An example for a founding AI engineer:
Subject: Evals before features | Founding AI Engineer | $[base] + [x]% equity
Hi [name],
Your write-up on reg testing for retrieval quality treats evaluation as the product: you measured what broke before anyone shipped a fix.
Deciding what "better" means before changing anything is the most useful trait for the job I'm hiring for, and it rarely shows on a résumé.
I'm the CTO of a seed-stage team of six building [product] for [buyer], backed by [investor]. The model works in production, but we can't yet prove when a change makes it better or worse, and inference cost grows with every customer.
Our first engineering hire will own model evaluation and bring cost per request down at equal quality within six months. Your work on [project] is that build.
The range is $[base] plus [x]% equity. You may find value in a short conversation where I walk you through the architecture, and I'd love to hear what you want your next move to look like. Are you open to 20 minutes on Google Meet this week or next?
[first name]
Pin's data on 4 million recruiting messages supports the personal approach and the follow-up: three messages captured 93.2% of all replies a sequence earned, and recruiting emails of 150 to 199 words drew the highest reply rate (Pin, 2026). The example above falls in that range.
What reply rates look like
Reply rates depend on how they are counted, so compare like with like.
What reply rates look like, counted two ways
Replies per delivered message
Automated recruiting email
Recruiter-written email
LinkedIn message
Share of people contacted who replied
Email plus LinkedIn, two-step
[solved] direct outreach
Show the data
| Measure | Rate | What it counts |
|---|---|---|
| Automated recruiting email | 4.96% | Replies per delivered message (Pin, 2026) |
| Recruiter-written email | 6.31% | Replies per delivered message (Pin, 2026) |
| LinkedIn message | 17.08% | Replies per delivered message (Pin, 2026) |
| Email plus LinkedIn, two-step sequence | 45.76% | Share of candidates who replied (Pin, 2026) |
| [solved] direct outreach | 67% | Share of people contacted who gave a human reply, including declines and referrals, April 2023 to March 2026 ([solved] record, defined on How we measure) |
Reply rates also vary by discipline. In Pin's data, product managers replied to recruiting email at 9.94%, executives at 7.33%, salespeople at 4.69% and engineers at 4.64%. Engineers and salespeople receive the most outreach, so personal, specific messages matter most for them. Pin notes that multi-channel sequences are partly a selection effect: recruiters add LinkedIn messages for the candidates they want most.
Estimate the time it will take
Run the arithmetic with your own numbers before you decide who does this work:
- Mapping: people on your map × minutes to research each profile.
- Outreach: people contacted × minutes to write a personal first message, plus follow-ups.
- Conversations: people contacted × your reply rate × the share worth a call × call length.
- Interviews: candidates in process × interview stages × hours per stage, plus debriefs.
Add the four lines, then multiply by the number of roles you are hiring this quarter. Compare the total with the hours the rest of your work needs over the same period.
In a [solved] search
One of our senior talent partners maps the market for your role and writes to each candidate personally with your outcomes, stage and pay range, so the people you meet already know the role and want to discuss it.

Sources (3)
Sources for this chapter
- Vendor research: Pin (Erica Stacy). Recruiting Outreach Benchmarks 2026: The Full Data Report. Published July 8, 2026, updated September 24, 2026. 4,000,000+ messages from 1,500+ organizations, June 2025 to May 2026. Vendor data; includes Pin's citations of Gallup (2026) and Gem (2026).
- [solved] record: direct outreach reply rate, April 2023 to March 2026, defined on our How we measure page.
- Practitioner review: our review of founder and hiring-manager posts on LinkedIn and Hacker News, October 2025 to October 2026.
Chapter 4: Read evidence when AI writes the application

How do I assess candidates who use AI in interviews?
Key points
- AI-written CVs and take-homes show presentation, not ownership.
- Credit live evidence: a walk-through of past work, one changed condition, and answers under follow-up.
- Set the AI policy by the job, and tell candidates before the interview.
What changed
Post a role today and you can receive hundreds of polished applications in a day, most of them reading like a close match. That polish now costs candidates almost nothing. 97% of developers use AI tools at work, and 76% say AI makes hiring assessments easier to game (HackerRank, 2025; 13,732 respondents in 102 countries). Among U.S. hiring managers, 65% have caught applicants using AI deceptively, including reading AI-generated scripts (32%) and appearing as deepfakes (18%) (Greenhouse, 2025).
Engineering leaders see the same change from the hiring side. 71% say AI makes technical skills harder to assess, and they report that take-home projects and automated code tests lose their value fastest (Karat, 2026; 400 engineering leaders in the U.S., India and China).
The change reaches past engineering. In a vendor study of AI-led first-round interviews, 48% of candidates for technical roles were flagged for AI assistance, compared with 12% for sales roles, and 61% of flagged candidates scored above the pass mark (Fabric, 2026; 19,368 interviews, July 2025 to January 2026, platform data weighted to India). The study links the lower sales rate to open-ended questions. The principle holds in every specialty: questions with one textbook answer are the easiest to assist.
Which evidence still carries weight
Which evidence still carries weight
Produced alone, before the interview
- Weak
CV and cover letter
AI tailors both to the job post in seconds.
- Weak
Unsupervised take-home or automated test
Ranked first among methods losing value (Karat, 2026).
Produced live, or confirmed by others
- Strong
Live walk-through of past work
Follow-up reaches detail only the person who did the work holds.
- Strong
Live exercise plus one changed condition
Shows reasoning in real time.
- Strong
Reference from someone who saw the work
When that person worked alongside the candidate.
- Strong
Short paid project
At a cost in candidate time.
Show the data
| Evidence | What it shows | How it holds up when candidates use AI |
|---|---|---|
| CV and cover letter | How the candidate presents their history | Weak. AI tailors both to the job post in seconds. |
| Unsupervised take-home or automated test | Whether a correct answer was produced | Weak. Engineering leaders rank these first among methods losing value (Karat, 2026). |
| Live walk-through of past work | What the candidate personally built, decided and changed | Strong. Follow-up reaches detail only the person who did the work holds. |
| Live exercise with the tools the job uses, plus one changed condition | How they direct, check and correct AI output | Strong. The changed condition shows reasoning in real time. |
| Reference from someone who saw the work | Whether the walk-through matches what colleagues observed | Strong when the person giving the reference worked alongside the candidate. |
| Short paid project | Working rhythm, communication and decisions on a real problem | Strong, at a cost in candidate time (see below). |
Set the AI policy by the job
62% of organizations prohibit AI in technical interviews, while the same leaders estimate that over half of candidates use it anyway (Karat, 2026). A ban that candidates ignore measures willingness to follow a rule. A stated policy matched to the job measures the work.
- If the person will use AI daily in the role, allow it in the exercise. Assess how they direct it, check its output and catch its mistakes.
- If part of the role happens live, with no time to consult a tool (a customer call, a board discussion, a production incident), assess that part live and in conversation.
- Write the policy into the interview invitation. Candidates then show their real working method, and every interviewer assesses the same thing.
Three probes that work in every specialty
- Explain. Ask the candidate to walk through one piece of their own work: the goal, the options considered, the choice made and what happened. Ask for one specific detail at each step.
- Change. Alter one condition (budget, deadline, scale, buyer, team size) and ask what changes in their approach.
- Defend. Ask what failed, what they would do differently and who disagreed with them at the time.
Fabric recommends drilling into a specific failure example, and reports that assisted answers invent detail where genuine candidates say plainly what they have and have not done (Fabric, 2026). Real experience produces specific constraints, decisions, trade-offs and regrets.
Evidence by specialty
- Exercise: the candidate walks through a system they built, then works live on a real piece of your code with their usual AI tools. Midway, change one constraint: ten times the load, half the memory, a new compliance rule.
- Strong evidence: they name the trade-off they made and why, check AI output before using it, and find the flaw the change exposes.
- Thin evidence: a correct-looking solution with no account of why it works or where it would break.
- Exercise: the candidate presents a product decision they owned, with the data they had at the time. Then cut engineering capacity in half and ask what they would ship.
- Strong evidence: they separate what users said from what users did, cut scope with a stated reason, and name the metric they would watch.
- Thin evidence: a framework recited in general terms, with every feature kept.
- Exercise: the candidate walks through a deal they lost: the buyer, the champion, the moment it turned. Then run a live role-play of a hard objection from your own pipeline.
- Strong evidence: they name the gap in the deal precisely and show how they changed their approach afterward. In the role-play, they ask questions before they answer.
- Thin evidence: a loss explained entirely by price or product, or a polished pitch that skips discovery.
- Exercise: the candidate describes a hard call from their last role: a reorganization, a missed target, a hire that failed. Then move a deadline forward by a month and ask what they would cut and how they would tell the board.
- Strong evidence: specific people, dates and consequences, a decision they would now make differently, and references who confirm the account.
- Thin evidence: management principles in the abstract, with every outcome described as a success.
When a paid project fits
A short paid project on a bounded, real problem shows working rhythm better than any interview. It also asks the most of the candidate, and the specialists most worth hiring are usually employed. Keep it to one working day or less, pay for it, define the deliverable in writing, and offer a live walk-through of past work as an equal alternative. Present both options so neither reads as the lesser one.
Signs of assisted answers in a live interview
Fabric's interview data points to three patterns (Fabric, 2026; vendor study):
- A steady delay of a few seconds before every answer, whatever the difficulty.
- Eye movement that tracks across a line and snaps back, as in reading.
- Polished answers that turn vague when asked for a specific example.
One sign is a reason to ask more follow-up questions. The decision rests on the evidence the candidate produces.
In a [solved] search
One of our senior talent partners assesses every candidate through live walk-throughs and changed-condition probes, and records what each person built, decided and can defend. Each shortlisted candidate arrives with that record written up: what the evidence shows, how it relates to your role and what your team should still test. We call it the Candidate Evidence Brief.
Sources (4)
Sources for this chapter
- Vendor research: Fabric. State of AI Interview Cheating in 2026: Insights from 19,368 Interviews. January 14, 2026. Vendor study of AI-led interviews on Fabric's platform; cheating flag set at a probability above 40%.
- Vendor research: Greenhouse. An AI Trust Crisis: 2025 AI in Hiring Report. November 19, 2025. 4,136 respondents in the U.S., U.K., Ireland and Germany; figures cited here are from U.S. respondents.
- Vendor research: HackerRank. 2025 Developer Skills Report. Survey of 13,732 developers, managers, recruiters and executives in 102 countries, Q4 2024 to early 2025.
- Vendor research: Karat. Engineering Interviews in 2026: 3 Trends Hiring Leaders Must Prepare For. Published January 7, 2026, updated March 26, 2026. Survey of 400 engineering leaders in the U.S., India and China.
Chapter 5: Confirm identity in remote hiring

How do I verify a remote candidate is who they say they are?
Key points
- Verify every remote candidate after the first real conversation and before your team interviews.
- Match a government ID to the live person, then confirm their history through references you find yourself.
- Tell candidates at first contact, get consent, and keep the result, not the biometric data.
What the data shows
- Interview fraud is measurable. In a Gartner survey of 3,000 job candidates in Q2 2025, 6% admitted to interview fraud: posing as someone else, or having someone else pose as them. Gartner predicts that by 2028, one in four candidate profiles worldwide will be fake (Gartner, July 2025).
- Hiring managers are seeing it. Among U.S. hiring managers, 18% have caught applicants appearing as deepfakes, and 39% now run more in-person interviews to confirm authenticity. 36% of U.S. job seekers say they have used AI to alter their appearance, voice or background on video interviews, most to look more professional (Greenhouse, 2025).
- Some fraud is organized. The FBI reports that North Korean IT workers use AI and face-swapping during video interviews to hide their identities, and once hired have copied company code and extorted employers. Its guidance to employers: verify identity during interviewing, onboarding and throughout the employment of any remote worker (FBI, January 2025).
Founders describe a related loss of trust: they doubt applications that read as AI-written, and some have moved interviews back in person to filter them.
A verification sequence
A verification sequence
- 01
First contact
Tell candidates
When the check happens, what it collects, how long the result is kept.
- 02
After first conversation
Verify before your team interviews
Interviewers spend time on verified people only.
- 03
Verification
Match document to person
Government photo ID plus a live check.
- 04
Verification
Match person to history
References you locate yourself, not ones the candidate supplies.
- 05
Onboarding
Confirm again
The person who starts is the person you verified.
Show the data
- Tell candidates at first contact. State that identity verification is a standard step, when it happens, what it collects and how long the result is kept. Stated at the start, it reads as a sign of a careful process.
- Verify after the first real conversation, before your team interviews. Your interviewers then spend their time on verified people, and candidates who leave early are never asked for documents.
- Match document to person. The candidate submits a government photo ID and completes a live check: a selfie or short video matched to the ID photo, with a test that a live person is present. Specialist verification providers run this and return a result.
- Match person to history. Confirm employers, dates and roles through references you locate independently: a company's main line, a mutual contact, a former manager you find yourself. Numbers and emails the candidate supplies confirm only what the candidate arranged.
- Confirm again at onboarding. The person who starts is the person you verified. The FBI advises watching for changes to address or payment details during onboarding.
Consent and data rules
Face matching uses biometric data, which carries specific legal duties in several places hiring founders recruit from:
- EU and UK (GDPR, Article 9): biometric data used to uniquely identify a person is a special category. Processing it is prohibited unless an exception applies, such as the person's explicit consent for a specified purpose.
- Illinois (Biometric Information Privacy Act, Section 15): before collecting biometric data, a company must inform the person in writing and receive a written release. It must also publish a retention schedule and destroy the data once the purpose is met, or within three years of the person's last interaction, whichever comes first.
In practice: choose a provider that returns a pass or fail result and deletes the images, keep the result in your records, explain all of this to candidates in plain language, and have counsel confirm the rules for each location you hire in.
Signals worth a closer look, by specialty
Each of these is a reason for more checking, and none is a verdict on its own.
Engineering & AI: video that drops at key moments, or lip movement out of sync with speech, that continues once connection problems are ruled out; a public code history that conflicts with the work described, after private employment or a career change has been considered; requests to change shipping or payment details after an offer.
Product: portfolio work the candidate cannot walk through in detail; case studies that closely match work published by others.
Sales & GTM: deal sizes, logos and quota attainment that references cannot confirm; reluctance to connect you with a former manager.
Leadership: claimed team sizes or results that do not fit the company's size at the time; references limited to peers the candidate selected.
Founders rarely write about faked track records in sales and leadership hiring, so the last two lists come from standard reference practice and the checks above, without founder testimony behind them.
When a check fails
- Pause the process.
- Offer one chance to fix technical problems: an expired document, a poor image, a failed upload.
- If the mismatch remains, close the process courteously and state the reason as an incomplete identity verification.
- Keep the result and the dates. Delete the images on your provider's schedule.
- If you see signs of organized fraud, such as invented employers or redirected payments, follow the reporting steps in the FBI's notice.
In a [solved] search
Every candidate on a [solved] shortlist has been interviewed and then identity-verified before you meet them, with consent and data handling explained to the candidate at the start.

Sources (6)
Sources for this chapter
- Industry research or guidance: Gartner. Gartner Survey Shows Just 26% of Job Applicants Trust AI Will Fairly Evaluate Them. Press release, July 31, 2025. Includes the Q2 2025 survey of 3,000 job candidates.
- Vendor research: Greenhouse. 2025 AI in Hiring Report. November 19, 2025. Figures cited are from U.S. respondents.
- Government and law: Federal Bureau of Investigation. North Korean IT Workers Conducting Data Extortion, Alert I-012325-PSA. January 23, 2025.
- Government and law: GDPR Article 9: Processing of special categories of personal data.
- Government and law: Illinois General Assembly. Biometric Information Privacy Act, 740 ILCS 14/15.
- Practitioner review: our review of founder and hiring-manager posts on LinkedIn and Hacker News, October 2025 to October 2026.
Chapter 6: Give every interviewer one responsibility

How should a startup run its interviews?
Key points
- Give each interviewer one outcome, set questions and a written scale.
- Every interviewer scores alone before the debrief.
- The debrief discusses recorded evidence only.
Why structure decides the result
A 2022 re-analysis of a century of selection research ranked structured interviews first among selection methods, with a mean validity of .42 (Sackett and colleagues, 2022). The same research shows a wide spread around that mean, from .18 to .66 across studies. The interview format alone settles little. The practical value depends on how the interview is built: questions drawn from the actual job, consistent scoring and trained interviewers.
Founders describe what happens when the interviews test the title and the job goes untested. One founder searching for a first product manager found that most candidates interviewed were project managers. Another account describes eleven of twelve seed founders hiring a first designer for execution when the company needed judgment. Every interview went well; the loop simply tested the wrong thing.
Give every outcome one owner
List the outcomes from Chapter 1. Give each outcome to one interviewer. That person assesses it in depth; everyone else leaves it alone.
This does three things. Candidates answer each question once, in depth, and the loop gets shorter. Every outcome receives a full assessment by someone prepared for it. And the debrief has one accountable view per outcome, which makes disagreement visible and specific.
Culture is the exception: every interviewer watches for it, records it apart from their score, and passes forward questions for the next stage to probe, never their score. When stages disagree, weight what the candidate did live over what they described.
One outcome, one owner, one score
- Score to debrief
- Questions passed forward, never scores
- One culture concern, tracked across stages
Sample loops by specialty
Four stages cover most consequential hires. Each stage has one owner and one responsibility.
| Specialty | Stage 1 | Stage 2 | Stage 3 | Stage 4 |
|---|---|---|---|---|
| Engineering & AI | Founder or CTO: role, stage and motivation | Senior engineer: walk-through of a system they built, then a live change of constraint | Second engineer: production judgment (failure, monitoring, review of AI output) | Product lead: working with product and customers |
| Product | Founder: market understanding and product judgment | Product or design lead: a past decision, then a live re-prioritization | Engineering lead: technical trade-offs and working with engineers | Customer-facing lead: how they learn from customers |
| Sales & GTM | Founder: stage, buyer and why now | Sales lead or founder: walk-through of a won and a lost deal | Solutions engineer or founder: live discovery and objection role-play | Customer success or operations: handoffs and process building |
| Leadership | CEO: mandate, the next 12 to 18 months and board context | Peer executive: building and running a team | Functional expert or board member: depth in the function | Two reports or skip-level peers: how they lead day to day |
Keep each stage between 45 and 60 minutes, and the full loop to four stages where you can.
What a scorecard contains
A scorecard turns an interview into evidence the team can compare. It has five parts:
- The outcome being assessed, in the words of Chapter 1.
- The questions or exercise, the same for every candidate.
- A rating scale with a written description of strong, acceptable and thin evidence for this outcome.
- An evidence log: what the candidate said or did, in specifics, recorded during or straight after the interview.
- A recommendation, submitted before the interviewer reads anyone else's.
Run the debrief on evidence
- Every interviewer submits a scorecard before the debrief starts.
- Each interviewer presents their evidence for their outcome. Opinions come after evidence.
- The founder or hiring manager speaks last, so the most senior view anchors no one.
- A concern counts when it rests on recorded evidence. A concern without evidence becomes a question for a reference or a follow-up conversation.
- Decide by the outcomes: does the evidence meet the bar for each one?
Treat candidates as future colleagues
The process is the first piece of work a candidate sees from your company. HackerRank's 2025 survey of over 13,000 developers lists slow responses, unclear expectations and too many rounds among the main frustrations in hiring, and 40% of developers say they plan to leave their current job within a year, so the people you interview today talk to the people you interview next year. Three practices protect your reputation:
- Publish the process at the start: the stages, who the candidate meets, how long each step takes and when they will hear back.
- Close every loop quickly, including a clear decision for candidates who are declined.
- Pay for any work beyond a normal interview, and keep it bounded (Chapter 4).
To give every interviewer one thing to test before the first interview, explore the Interview Process Planner.
In a [solved] search
One of our senior talent partners designs your written interview plan with you, gives each interviewer a scorecard tied to one outcome you approved, and prepares the debrief around recorded evidence.
Sources (3)
Sources for this chapter
- Peer-reviewed research: Sackett, P. R., Zhang, C., Berry, C. M. and Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. Journal of Applied Psychology. Range of structured interview validity reported in SIOP's review of the paper.
- Vendor research: HackerRank. 2025 Developer Skills Report. Survey of 13,732 respondents in 102 countries.
- Practitioner review: our review of founder and hiring-manager posts on LinkedIn and Hacker News, October 2025 to October 2026.
Chapter 7: Close the hire and protect the first 90 days

How do I close a candidate and handle a counteroffer?
Key points
- Make the offer within a day or two of the final debrief.
- Present equity as numbers, and prepare for the counteroffer before it arrives.
- Check in at 30, 60 and 90 days, then review at one year against the 365-day outcome.
Speed decides more offers than founders expect
Ashby's analysis of 230,000 offers made between January 2021 and March 2024 shows how quickly decisions form. Candidates who accepted took about two to three days. Candidates who declined held the offer for about six. Technical roles were accepted 73% of the time, compared with 84% for business roles, and technical candidates took about a day longer to respond (Ashby, 2024).
Two practical rules follow. Make the offer within a day or two of the final debrief. And when a candidate goes quiet for a week after an offer, treat the offer as at risk and ask what they need to decide.
To see how long your own finalists wait, and what each extra day costs, explore the Candidate Experience Self-Audit.
Speed decides more offers than founders expect
Days to respond to an offer
Accepted
Declined
Acceptance rate
Technical
Business
Show the data
| Measure | Group | Value |
|---|---|---|
| Days to respond to an offer | Accepted | 2 to 3 days |
| Days to respond to an offer | Declined | about 6 days |
| Acceptance rate | Technical | 73% |
| Acceptance rate | Business | 84% |
Surface the counteroffer early
A strong candidate who is employed will often be asked to stay. Plan for it from the first conversation:
- What would have to change at your current company for you to stay?
- Who else are you talking to, and where are you in those processes?
- If your company offered more money tomorrow, what would you do?
- What does the next role need to give you that this one cannot?
If the answers are about money alone, a matching offer from the current employer is likely to work. If they are about scope, ownership or the problem itself, your outcomes from Chapter 1 are your strongest argument. Before the candidate resigns, rehearse with them how they expect the conversation with their manager to go, so the counteroffer arrives as something expected.
Present the offer so it can be valued
A percentage alone is hard to judge. Put every component on one page:
- Cash: base salary, and for sales roles the variable pay, the quota, the ramp period and anything protected during ramp (Chapter 2).
- Equity: number of options or shares, the percentage of fully diluted shares it represents, the exercise price, the price per share at the last round, the vesting schedule and cliff, and the exercise window after leaving.
- Scenarios: what the grant is worth at the last round price and at two higher valuations, with the assumptions written out. Label these as illustrations.
- Scope: the 90, 180 and 365-day outcomes, the decisions the person will own, and who they report to.
For context, the median Series A company reserves 15% of fully diluted equity for its option pool (Carta, 2026). Chapter 2 covers typical grants for early hires.
References: timing and questions
Check references after the final debrief and before the written offer, or alongside it. Speak to people who saw the work: a former manager, a peer, someone who reported to the candidate. Find at least one yourself, through a mutual contact or a company you know.
Ask about the outcomes from Chapter 1 and about any concern the debrief raised. "Tell me about a time they had to rebuild something that was already working" yields more than "Would you hire them again?"
From signature to the first day
The weeks between signature and start are when a counteroffer can still land. Keep contact steady: introduce the people they will work with, share the 90-day plan, and set up accounts and access before day one.
The first 90 days by specialty
Tie each check-in to the outcomes you wrote in Chapter 1.
| Specialty | Day 30 | Day 60 | Day 90 |
|---|---|---|---|
| Engineering & AI | Systems and codebase understood; first change shipped; first bottleneck named | First bounded outcome delivered (for example, an evaluation suite in production) | Owns an area end to end, including releases |
| Product | Customer conversations complete; current metrics and drop-off mapped | First change based on that discovery shipped | Owns discovery and roadmap within the agreed decision rights |
| Sales & GTM | Ideal customer profile and target accounts agreed; first outreach running | First qualified pipeline the founder did not source | First deals advancing; sales process written down |
| Leadership | Team, delivery and budget reviewed; first priorities agreed | First structural or process change in place | Owns the function's plan, hiring and reporting to the board |
Hand over what you agreed to hand over
The founders in our review describe the hardest part of a first hire into a founder-run function plainly: it means delegating decisions that used to be yours. One founder describes hiring a first product manager and product owner, then taking product back when neither had real ownership. Write down what you will stop doing on day one, and review that list at each check-in alongside the hire's outcomes.
Check in at 30, 60 and 90 days, then at one year
Each check-in reviews the written outcomes, the evidence of progress, and what the hire needs from you. Small mismatches surface early when someone asks. Keep a one-year review against the 365-day outcome; it closes the loop on how well the search defined the role.
Check in through the first year
Signature
Day one
30 days
Check-in
60 days
Check-in
90 days
Check-in
1 year
Review against the 365-day outcome
In a [solved] search
One of our senior talent partners manages the offer and counteroffer stage with you, checks in with you and your new hire at 30, 60 and 90 days and again at one year, and every search carries replacement coverage.

Sources (3)
Sources for this chapter
- Vendor research: Ashby (Bryce Codell). Offer Acceptance Rates: Talent Trends Report. June 6, 2024. 230,000 offer-stage applications, January 2021 to March 2024.
- Vendor research: Carta. State of Startup Compensation H2 2025. May 4, 2026.
- Practitioner review: our review of founder and hiring-manager posts on LinkedIn and Hacker News, October 2025 to October 2026.
Chapter 8: Choose the right outside help

Should I use a recruiter, and what do recruiters charge?
Key points
- Choose outside help by the hire, your hiring volume and your own time.
- Compare what the fee covers, when it is due and what the provider earns for.
- Ask every provider the same questions before you sign.
The six models
Published fee terms vary by source, so the table shows each source's own figures.
| Model | How it bills | Suits | Real risk |
|---|---|---|---|
| Founder-led, with an applicant tracking system | No external search fee; founder hours (Chapter 3 shows the arithmetic) and hiring tools | One to three hires a year, network and inbound-heavy roles | Founder time leaves product, customers and fundraising; reach stays inside the network |
| Contingency search | A percentage of first-year salary, due when a hire starts: 15 to 25% (lemonly, 2026), 15 to 30% (Dover, 2026), 20 to 25% (Funded.club, 2026) | Roles with a clear market where you want to owe a fee only on a hire | The fee rises with the salary you agree; firms earn only on placements, so effort follows the roles most likely to fill; several firms on one role can mean duplicate outreach |
| Retained search | 25 to 35% of salary, billed in installments during the search, whatever the outcome (lemonly, 2026) | Executive and confidential searches that need exclusivity and deep research | The highest upfront search-fee commitment of any model, made before results arrive |
| Fixed-fee search | A set price per hire, agreed before the search, usually part at the start and the balance on a hire. One published example: $4,900 to $21,900 by salary band (Funded.club, 2026) | Consequential hires where you want the fee fixed regardless of the salary you negotiate | What the price covers varies widely between providers; some still set price bands by salary |
| Fractional recruiter | Hourly or monthly. Dover reports $2,000 to $7,000 per hire on its marketplace, about $4,000 on average (Dover, 2026) | Five to fifteen hires a year, when you want to own the pipeline and the process | You still run the process and the decisions; results depend on one individual's network and hours |
| In-house recruiter | Salary and overhead: $85,000 to $170,000 a year loaded (Dover, 2026) | Fifteen to twenty or more hires a year (Dover, 2026) | A fixed cost whatever your hiring volume; one person's reach |
The Dover and Funded.club figures are their own published terms and comparisons; each company sells one of these models.
What the fee comes to
Percentage fees scale with salary. Here is the arithmetic at three base salaries, using lemonly's published ranges:
What the fee comes to at three base salaries
- Contingency, 15 to 25%
- Retained, 25 to 35%
- Fixed fee, one published example
$120,000 base salary
$180,000 base salary
$250,000 base salary
Fixed fee
Show the data
| Base salary | Contingency (15 to 25%) | Retained (25 to 35%) |
|---|---|---|
| $120,000 | $18,000 to $30,000 | $30,000 to $42,000 |
| $180,000 | $27,000 to $45,000 | $45,000 to $63,000 |
| $250,000 | $37,500 to $62,500 | $62,500 to $87,500 |
Check what the percentage applies to. Some agreements calculate the fee on base salary only; others include bonus, commission at target or sign-on payments, which matters most for sales roles with variable pay. A fixed fee stays the same however the salary negotiation ends.
What each model earns for
Every model rewards something, and the reward steers the work:
- Contingency earns on a hire, so speed to a placement matters most. The fee also grows with the agreed salary, which is often the first thing hiring managers notice when they read the terms.
- Retained earns for running the search, so depth, research and confidentiality are what the client buys.
- Fixed-fee earns a set amount per hire, so the fee stays independent of the salary; the scope of work behind that amount is what you compare.
- Fractional earns for time, so hours and process discipline are what you manage.
No model is right for every hire. One founder may use fractional help for steady engineering hiring and retained search for a first executive in the same year.
Questions to ask any provider before signing
- 01What exactly does the fee cover: sourcing, assessment, identity verification, references, offer and counteroffer support?
- 02How is the fee calculated, on which parts of compensation, and when is each part due?
- 03Who does the work day to day, and how many searches does that person run at once?
- 04What will we see during the search: the market map, the outreach, the pipeline?
- 05What happens if the shortlist misses the brief, if we stop the search, or if we hire someone we found ourselves?
- 06What replacement terms apply, for how long, and what triggers them?
- 07Is the search exclusive, and what does exclusivity change in the terms?
- 08How do you protect our reputation with the candidates you contact and decline?
To take questions like these into your first call with any recruiter, explore the Recruiter Brief: its last step lists what to agree with your recruiter before the search starts.
In a [solved] search
[solved] runs fixed-fee searches, with the fee set by the role's scope and complexity and agreed before the search begins.
Sources (4)
Sources for this chapter
- Vendor research: lemonly. How Much Do Recruiters Charge in 2026? Contingency, Retained and Flat-Fee. July 11, 2026. Listed on the lemonly Guide to Hiring.
- Vendor research: Dover. From Agency Recruiters to In-House Hiring for Startups. Published June 13, 2025, updated September 27, 2026.
- Vendor research: Funded.club. Pricing. Accessed October 7, 2026.
- Practitioner review: our review of founder and hiring-manager posts on LinkedIn and Hacker News, October 2025 to October 2026.
Before you begin the search
- The 90, 180 and 365-day outcomes are written, and every interviewer would describe them the same way.
- You know which decisions the hire will own, and what you will stop doing.
- Pay, equity and, for sales roles, quota and ramp are checked against current data.
- Real profiles of people doing this job confirm the role holds together.
- You know where the people you need are working today.
- Your AI policy for interviews is written into the invitation.
- Identity verification, consent wording and data handling are settled.
- Each interviewer has one outcome, set questions and a written scale.
- You can move from the final debrief to an offer within two days.
- You have decided who runs the search, and what it costs in fees or in your own hours.
To hand this search to one of our senior talent partners: Start a Search. To test the role with us first: Talk Through a Role.
Frequently asked questions
What should my first engineer, product manager or sales hire own in year one?
Write down what the hire must deliver at 90, 180 and 365 days before you write a job description or contact anyone. Add the decisions the person will own and the trade-offs you accept. These outcomes set the level, the pay, the interview plan and the first-year review.
How do I set salary, equity and scope before I start hiring?
Check the role against outside evidence: current pay and equity data for the function and stage, whether your company is ready for the role, and whether real people already do the job you wrote. Adjust scope, level or pay while the role is still private.
How do I find senior people who aren't applying?
Map where people doing this work today sit, then write to each person about their own work, with your specific outcomes, stage and pay range. Follow up across more than one channel. Personal, specific outreach earns replies that job posts and generic sequences rarely see.
How do I assess candidates who use AI in interviews?
Credit the evidence a candidate produces live: a walk-through of work they did, an adaptation when one condition changes, and a defense of their decisions under follow-up. Set an AI policy that matches the job, and tell candidates before the interview.
How do I verify a remote candidate is who they say they are?
Verify the identity of every remote candidate after the first real conversation and before your team interviews, with their consent and in line with the law where you hire. Match a government ID to the live person, then confirm their claimed history through references you find yourself. Tell candidates about the check at first contact, and keep the result, not the biometric data.
How should a startup run its interviews?
Assign each interviewer one outcome to assess, with set questions or an exercise and a written scale. Each person scores independently before the debrief, and the debrief discusses only recorded evidence.
How do I close a candidate and handle a counteroffer?
Learn early what would make the candidate move and what would make them stay, move quickly once you decide, present equity as numbers, and prepare for a counteroffer. After signature, check in at 30, 60 and 90 days against the written outcomes.
Should I use a recruiter, and what do recruiters charge?
Choose by the hire, your hiring volume and your own time. Compare models by what the fee covers, when it is due and what the provider earns for, and ask every provider the same questions about scope, terms, replacement and how they treat your candidates.
