In 2022, Levels.fyi published its annual compensation report showing that the median total compensation for a senior software engineer at a large U.S. tech company was $310,000 — base, bonus, and equity combined. Most engineers accepting offers that year had no idea the figure existed. They negotiated against their own current salary, not against the market. The result: offers accepted at $240,000 for roles where the employer's own internal band topped $350,000.
The gap between the informed candidate and the uninformed one was not talent. It was data.
Salary negotiation is an anchoring contest. Whoever states a number first — and backs it with credible evidence — controls the range that follows. Research by Adam Galinsky and Thomas Mussweiler (published in the Journal of Personality and Social Psychology, 2001) demonstrated that even arbitrary anchors shift final outcomes. Credible, market-referenced anchors shift them further and with less resistance from the other side.
The problem is that most candidates anchor to the wrong thing: their last salary. That number reflects their old employer's pay band, their old city's cost of living, and negotiations they lost years ago. It has nothing to do with what the current market will bear.
AI tools change this entirely. You can now synthesize data from Levels.fyi, Bureau of Labor Statistics Occupational Employment and Wage Statistics (OEWS), LinkedIn Salary, Glassdoor, Payscale, and H-1B disclosure records — all within a single research session — and arrive at a defensible market number before you ever receive an offer letter.
In 2021, the Department of Labor's H-1B disclosure database revealed that Google paid its L5 Software Engineers a median base of $197,000 in the Bay Area. That data was public. Candidates who cited it during negotiation had a specific, government-sourced floor. Candidates who didn't cited nothing at all.
Not all salary data carries equal weight in a negotiation. Employers know which sources are rigorous. Citing the right ones signals that you are a serious, prepared candidate — which itself increases leverage.
| Source | Strength | Best For |
|---|---|---|
| BLS OEWS | Government-collected, no self-report bias | Broad occupation baselines, government/nonprofit roles |
| H-1B Disclosure Data | Legally required filings, employer-specific | Tech companies sponsoring visas; reveals internal bands |
| Levels.fyi | Verified TC data for tech roles | Total compensation at named tech employers |
| LinkedIn Salary | Large sample, role-filtered by geography | Cross-industry ranges, location adjustments |
AI does not invent salary data. Its job is to help you triangulate, contextualize, and phrase the data you already have. A well-structured prompt produces a side-by-side synthesis of multiple sources, identifies the 75th-percentile figure you should be targeting, and drafts a negotiation email that cites the data without sounding robotic.
The workflow used by well-prepared candidates in 2023–2024 typically looked like this:
A number without a source is just a wish. A number with a government filing, a peer-reviewed wage survey, or a verified compensation database is a market reference — and employers treat it as one. Your job in this module is to learn how to build that case using AI as your research accelerator.
Tell the AI your target role, target employer (or employer type), and city. Then ask it to help you build a market data brief — synthesizing at least two salary sources into a target number, a floor, and a one-sentence rationale you could say aloud in a negotiation call.
In 2023, a study by researchers at the Wharton School found that professionals using AI for compensation research often received wide salary ranges — sometimes spanning $40,000 or more — because they asked broad questions. "What do data scientists make?" returns a band so wide it is useless. "What does a senior data scientist with five years of experience earn at a Series B fintech company in New York, based on LinkedIn Salary and Levels.fyi data from 2023–2024?" returns something you can actually use.
Every salary range in every database has filters. Most people ignore them when prompting. The five variables that collapse a $40,000 band into a $12,000 band are:
A prompt that produces negotiation-grade intelligence has four parts: role definition, context, data sources cited, and output format requested. Here is a before-and-after comparison based on real candidate research sessions documented in a 2024 hiring manager survey by Greenhouse Software:
"What salary should I ask for as a product manager?"
"I am a senior product manager with 7 years of experience, focused on growth and monetization at B2C subscription companies. I have an offer from a Series C SaaS company in Austin, TX. Using LinkedIn Salary, Glassdoor, and any BLS OEWS data you can reference for this MSA, give me: (1) the median base salary, (2) the 75th percentile base, (3) typical equity and bonus structures at this stage, and (4) one sentence I could say on the phone to justify asking for the 75th percentile figure."
Base salary is the easiest number to research. Total compensation — including RSUs, options, and bonuses — is where most candidates leave money. A documented example: in 2022, Carta published data showing that the median equity grant for a VP of Engineering at a Series B company was 0.25% of the company's outstanding shares. Most candidates had no idea what percentage to request or how to value what they were offered.
Prompts that help AI help you on equity look like this:
"The company offered me [X] shares / [Y]% equity with a 4-year vest and 1-year cliff. They have raised $[Z] in Series B. Using Carta's 2022–2023 equity benchmarks and standard dilution assumptions for a Series B company, help me understand: (1) is this grant competitive for my level, (2) what is a rough current value assuming a 3–5x exit multiple, and (3) what counter-offer percentage would be defensible?"
A single prompt rarely produces the final answer. Treat the AI conversation as a research interview. Follow up with narrowing questions: "You gave me a range of $140K–$180K. What moves someone from the bottom of that range to the top?" Then: "What evidence from my background would I cite to justify the top?" This iterative approach — documented as best practice in the 2024 LinkedIn Learning report on AI-assisted career development — turns a generic range into a personalized, defensible case.
Practice writing a strong compensation research prompt using all five variables: seniority, industry vertical, company stage, MSA-level geography, and recency. The AI will evaluate your prompt, identify what's missing, and help you refine it until it would return genuinely useful data.
In 2023, Jobvite's annual Recruiter Nation report found that 84% of recruiters said they had room to negotiate on at least one component of the offer they extended — but only 37% of candidates actually asked. Of those who did ask, the ones who provided a specific market reference received a counter more than twice as often as those who simply said they hoped for more.
The message is not that data alone wins. It is that data paired with professional language is the combination that gets a counter on the table.
A negotiation email that works has five components, each doing a specific job. AI can draft all five — but you must provide the data and the tone. The five components are:
A software engineer who accepted a Stripe offer in 2022 reported on Blind that their initial offer was $195,000 base. They responded in writing, citing Levels.fyi data showing the median base for the equivalent level at comparable companies was $215,000, and requested $220,000. Stripe countered at $210,000. The email took eleven minutes to write; the outcome was $15,000 per year in additional base salary.
The prompt structure that produces a usable first draft in one pass requires you to pre-load four things: the offer details, the market data you found, your value rationale, and the tone you want. Leaving any of these out forces the AI to guess — and it will fill gaps with generic language that sounds exactly like every other negotiation email the recruiter has read this month.
"Draft a salary negotiation email for me. Details: I received an offer for [Role] at [Company] with a base of $[X], [bonus structure], and [equity]. My market research using [Source 1] and [Source 2] shows the 75th percentile for this role in [MSA] is $[Y]. My specific value rationale is [2–3 sentences about your background]. I want to ask for $[Z] base. Tone: warm, direct, professional — not apologetic, not aggressive. Length: under 200 words."
Most offers are extended by phone. Most negotiation happens by phone or video. The email sets the context; the verbal conversation closes it. AI can script the verbal version too — and it should, because candidates under pressure revert to apologetic language ("I was just hoping…") that signals low confidence and invites a smaller counter.
The three lines that do the work in any verbal negotiation:
Two phrases reliably reduce outcomes. The first is "I need" — this frames the conversation around your personal circumstances rather than market value, which weakens your position. The second is "Is there any flexibility?" — this opens with a yes/no question that a recruiter can answer with "no" and end the negotiation. Always state a number and ask for a response to it, not permission to discuss.
Provide the AI with your offer details (or a realistic hypothetical), the market data you've found, your value rationale, and your target ask. The AI will draft a negotiation email under 200 words with the correct five-part structure, then help you refine it.
In 2023, a survey by Robert Half found that 58% of workers who received a counter-offer from their current employer after accepting a new job accepted the counter — and of those, 50% were no longer at that employer within 18 months. The counter-offer, while flattering, is often a retention tactic, not a genuine recalibration of how the employer values you. Knowing this shapes how you respond.
On the other side of the table, candidates with competing offers in 2023 negotiated 18.6% higher total compensation on average, according to data published by Glassdoor's economic research team. The competing offer is the single most powerful piece of leverage in any negotiation — but only if used correctly.
When the employer comes back below your ask, you have three options: accept, make a second counter, or shift to non-salary components. Most candidates accept on the first counter because they assume it is the final answer. It often isn't. The 2022 LinkedIn Salary Negotiation Survey found that candidates who made a second, narrowed counter — moving to a number between their original ask and the employer's counter — succeeded 41% of the time.
The second counter must be smaller in absolute terms (you are moving toward them) and must cite a new reason, not just repeat the first ask. AI can help you identify what new element to introduce — a specific project outcome, a certification, or a signing bonus ask that bridges the gap.
"The employer countered at $[X], below my ask of $[Y]. I want to make a second counter at $[Z]. Help me write a response that: (1) acknowledges their move positively, (2) introduces one new data point or value rationale I haven't cited yet, (3) proposes $[Z] or asks for a signing bonus of $[W] to bridge the gap, (4) signals this is my final ask without ultimatum language."
A competing offer is only leverage if it is real and if you would genuinely consider it. Fabricating one is fraud and has ended careers when discovered — there are documented cases of offer rescissions after candidates were caught misrepresenting competing offers at Amazon and Meta in 2022 and 2023.
When you have a real competing offer, the correct approach is transparency without ultimatum. You are informing them of a market signal, not threatening them. The phrase "I want to be transparent that I have another offer at $[X] that I'm considering, and I'd much rather be here — is there anything you can do to help me make this decision easier?" is both honest and effective.
AI's role here is drafting the language of this disclosure — striking the tone between informative and pressuring, which is harder to calibrate than it sounds in the moment.
In 2022, a product designer reported on the Work Chronicles newsletter that disclosing a competing offer from Figma while negotiating with Shopify resulted in Shopify increasing their offer by $22,000 in total comp — not because of the competing number itself, but because the disclosure signaled that the candidate had genuine market demand and would not wait indefinitely.
"This is our best and final offer" is a negotiation tactic in approximately 60% of cases where it is used, according to an analysis by career coach Ramit Sethi published in 2023 based on 1,200 documented negotiation outcomes. The remaining 40% are genuine. You cannot know which you are facing.
The correct response to a "best and final" is not to immediately accept or reject. It is to pause, thank them genuinely, and ask whether any non-salary components (signing bonus, additional PTO, earlier review date, professional development budget) have flexibility. This shifts the conversation away from the declared-final base salary to components that often do have remaining room.
Signing bonus (often has more room than base) · Additional PTO (1–5 days frequently granted) · Earlier performance review date (90-day instead of 12-month, allowing faster raise eligibility) · Remote work arrangement · Professional development / conference budget · Equity refresh schedule · Relocation assistance
Closing is its own skill. Once you have reached an agreement, confirm every component in writing — email back a summary of what was verbally agreed before you sign anything formal. Errors in offer letters are common and almost never caught by candidates who don't do this step. AI can draft the confirmation email in under two minutes from your notes of the call.
The confirmation email is not a negotiation email. Its tone is warm, grateful, and precise. It lists every agreed component: base, bonus target, equity grant, start date, signing bonus, and any non-standard arrangements discussed. If any element is missing from the formal offer letter that was verbally agreed, you want to know before you sign, not after.
Practice the closing phase of a negotiation. Tell the AI what counter the employer came back with, and work through the second-counter strategy, non-salary component pivots, or the final confirmation email — whichever stage you want to practice.