By Chris Morrow.
The short answer
The question is not how large the search firm is. It is whether anyone there can assess the work.
A generalist executive search firm will run a competent process and reach senior people. What it usually cannot do is tell you whether the VP of AI in front of you has shipped anything that stayed up, whether the research engineer’s interesting papers are the ones they led, or whether an agent platform on a CV was a demo or a production system. That judgement is the whole difficulty of hiring in this field, at every level from a first research engineer to a Chief AI Officer.
So the split that matters is AI depth against general leadership reach, not seniority. If the technical bar is the hard part of the hire, and in AI it almost always is, use a firm that works only in this field. If you are hiring general engineering leadership that happens to sit near an AI team, that is a different market and the generalists own it.
The three cities are different markets
Roughly 300 of our 500+ AI/ML placements since January 2021 have been in the US. These are the differences that actually change how a search runs.
New York
New York searches skew towards research and towards companies still in stealth. (We have written up how we work in New York separately, including the two searches described below.) The pattern we see repeatedly is a Series A lab, often a16z or tier-one backed, hiring its first engineer before it has a brand to open a conversation with. That inverts the usual problem: the constraint is not finding the person, it is that the company cannot yet explain who it is.
The pool for a specific research direction is global rather than local, and frequently a few hundred people worldwide. Most are mid-PhD or settled in research posts at major industrial labs. Almost none are looking. A New York address does not shrink that problem, it just means the person will have to move.
San Francisco Bay Area
The Bay Area is where you are most directly competing with the frontier labs, and where a candidate’s alternative offer is most likely to be one. That changes what a search has to be: less about reach and more about whether the work is genuinely interesting, because you will not win on money.
It also has the deepest pool by a distance, which cuts both ways. Everyone worth hiring is already being contacted, several times a week.
Los Angeles
The thinnest of the three for AI specifically, and the most likely to involve relocation or a hybrid arrangement. It is also where we have seen the clearest example of how these searches actually resolve: on one recent engagement several agencies were briefed on the same forward deployed engineering roles, and one candidate cleared every stage. That was the only placement any agency made.
That is characteristic rather than unusual. Where the overlap of skills is genuinely thin, briefing more firms does not produce more candidates.
What you are competing with
At seed to Series C you are rarely competing with another startup for the person you want. You are competing with the fact that they are comfortable somewhere already, usually a frontier lab or a large industrial research group, and are not looking.
This is the single most useful thing to understand before choosing a search partner, because it rules out the ones whose method is reach. A firm that solves your problem by contacting more people is solving a different problem. What matters is the quality of the approach, and whether the person making it can hold a technical conversation well enough to be worth replying to.
Which firm for which hire
The firms below all work in or adjacent to this market in the US. They are grouped by what they are best known for rather than ranked, because the ranking depends entirely on your hire.
| Firm | Focus | Strongest when |
|---|---|---|
| Digitalent | AI and machine learning only, across the US, UK and Australia | The technical bar is the hard part, at any level from a founding research engineer to Chief AI Officer |
| Riviera Partners | Engineering, product and design leadership, VC-backed | The role is broad engineering leadership rather than AI specifically |
| Daversa Partners | Executive search across venture-backed leadership teams | The hire is a general C-level appointment |
| True | Global executive search with a technology practice | The search spans several markets and functions at once |
| Harnham | Data and analytics recruitment at scale | The role is analytics or data engineering rather than applied AI |
| Acceler8 Talent | AI and machine learning contingent recruitment in the US | You are filling several mid-level engineering seats quickly |
Two things worth knowing that a comparison table cannot show.
Check who actually covers the US, rather than who says they do. Several firms with strong US positioning run those searches from elsewhere. That is not disqualifying, but it changes your interview scheduling and your offer conversations, and you should know before you engage rather than after.
Ask how many searches the person you meet is running. At the specialist end the person who takes your brief usually runs it. At the larger firms the person who wins the work and the person who does it are frequently different people, which is fine when the pool is deep and a real problem when it is not.
Where we are the wrong choice
We work the learning side of AI and nothing else, so we are the wrong choice for hardware robotics, for the mechanical and electrical engineering that sits next to an AI team, and for general software hiring with no AI dimension. We will say so at kickoff rather than take the brief and find out a month in.
We are the wrong choice outside the United States, the United Kingdom and Australia. We have made placements in three markets and we know those three. A firm with a real presence where you are hiring will beat us on a market we would be learning.
And we are the wrong choice if what you want is a name on the invoice. Some boards want a recognised search brand attached to an appointment, and that is a legitimate reason to choose one. It is not a thing we can offer, and we would rather say so than pretend the consideration does not exist.
What to ask before you engage anyone
Five questions that separate firms quickly, and none of them is about fees.
- Who runs the search day to day? Not who is in the pitch. Ask to speak to them before you sign anything.
- How will you assess the technical bar? A firm that cannot answer this for an AI role will send you people who interview well and cannot do the job.
- What does your shortlist look like? Ask for the shape, not names. If the answer is a number of CVs rather than a written case for each person, you are buying volume.
- What have you placed that is genuinely like this? Specifics, with the difficulty attached. Anyone can list logos.
- When would you tell me not to hire? The answer tells you whether you are buying judgement or effort.
Fees in the US
For a permanent (full-time) hire the fee is a percentage of the candidate’s starting base salary. What varies between firms is the structure rather than the principle: retained search is usually split across engagement, shortlist and placement, while contingent search is payable only on a hire.
What to establish before you engage, rather than the percentage itself:
- What triggers payment, signature or start date, and what happens if the person leaves inside the guarantee period
- Whether exclusivity is required, for how long, and what you get for it
- Whether the fee is on base salary alone or on total compensation including equity and bonus, which on a US senior hire is a material difference
- What happens if you hire two people from one search
A firm that answers all four in plain language on a first call is telling you something useful about how the rest of it will go.
Visa and relocation
Whether a firm can help with candidates who need sponsorship depends entirely on the client, not the recruiter. Establish it at kickoff rather than at offer stage, because discovering it late costs you the candidate and several weeks.
Relocation is a live question in all three cities, and more so than most clients expect. If the pool for your research direction is a few hundred people worldwide, most of them are not where you are.
Working hours
If your search partner is not in the US, ask directly whether calls, interviews and offer conversations run to your time zone or theirs. It sounds trivial and it is the most common practical failure in a cross-border search.