Making offers AI candidates accept
Speed, transparency and the parts of an offer beyond salary decide whether you close the AI candidates you want.
Know the market before you make an offer
Pay for AI roles moves quickly and varies widely by specialty and level, so a salary band set two years ago is likely out of date. Check current figures before any offer conversation: the TensorHunt AI & ML salary guide, public company data such as Levels.fyi, and what candidates tell you about their other offers.
If your bands are below the market, fix them before the search starts, not halfway through it. An offer you already know is too low wastes everyone’s time, and the AI community is small enough that candidates compare notes.
Location still matters for remote roles, but experienced ML candidates often hold offers from companies in the highest-paying markets. Know what your candidate’s alternatives look like, not just your local market.
The whole offer, not just salary
Experienced candidates compare whole packages. Put every part in writing: base, equity (type, total grant, cliff, vesting schedule, latest valuation or strike price), target bonus, benefits and any signing bonus.
For AI roles, add the parts that decide whether the person can do good work: how much compute they will have and who approves it, whether they can publish or contribute to open source, and the budget for conferences. For researchers these can matter as much as salary.
Offer breakdown checklist
- ✓ Base salary (annual)
- ✓ Equity: type, total grant, cliff and vesting schedule
- ✓ Latest valuation or strike price (for options)
- ✓ Target bonus and payout history
- ✓ Compute budget or GPU access
- ✓ Publishing, open-source and conference policy
- ✓ Health benefits and retirement match
- ✓ Signing bonus, if any
Speed is a competitive advantage
Candidates at the offer stage are usually talking to other teams. Every day you take to prepare an offer is a day another company can make one. Calling the candidate with a verbal offer soon after the final decision shows them they are wanted.
The usual bottleneck is internal approval. Agree the pay range, including equity and any compute budget, before the search starts, and let the hiring manager make an offer within it without another round of sign-off.
When they negotiate
Expect negotiation, and decide your limit before you send the first offer. Know which parts you can move (base, equity, signing bonus, start date, remote work, title, compute, research time) and which you cannot. A candidate who negotiates is interested enough to try to make the offer work.
Consider closing a gap with something other than base salary. A signing bonus or a larger equity grant does not raise base pay permanently, and a compute budget or publishing rights can matter more to a researcher than another increase in salary.
Negotiation levers (easiest to hardest)
- ✓ Signing bonus: one-time, no long-term cost
- ✓ Compute budget or research time
- ✓ Equity grant, especially before an IPO
- ✓ Start date and remote flexibility
- ✓ Base salary: move last, it affects every future raise
The offer letter
Write the offer letter in plain language that a candidate can read in five minutes without a lawyer. Explain any restrictive terms, such as non-competes or IP assignment, in plain words next to the legal text. Candidates who find a hidden clause stop trusting you.
IP terms matter especially to ML people who publish or maintain open-source projects. If personal projects and papers are allowed, say so clearly. Allow some flexibility on the start date, since most candidates must give notice, and call within a day of sending the letter to answer questions.