Here is the part most growth-stage founders have not fully sat with. If an AI tool in your hiring
process gets something wrong, the exposure usually sits with you, not with the vendor who sold
the software.
That is why this is worth a look now, not after a candidate challenge, a fundraise, or a board
question. This is not an argument against using the tools. Most companies already use them in
some form. It is about not handing a system a decision you cannot explain.
The Rules Are Moving Faster Than Most Teams Are
States, cities, and regulators have been writing rules around automated hiring tools for a few
years. Some are already in force in specific places. New York City requires bias audits for
certain automated employment tools. Colorado has put duties on both the people who build the
tools and the employers who use them. California has taken a hard line that a bad AI outcome is
not a defense. The EU treats many recruiting tools as high-risk when they touch EU-resident
candidates.
You may not be in all of those jurisdictions. If you hire remotely, take money from elsewhere, or
run a process that touches candidates in those markets, you can still inherit the problem.
Most growth-stage companies are not tracking this closely. The ones that find out after a
complaint are in a worse spot than the ones that put basic hygiene in place first.
The Bias Problem Is Not a Glitch
A large share of companies already admit their hiring tools could introduce bias. Age shows up
often. So do socioeconomic and gender patterns. Controlled tests of some LLM resume
screeners have also shown clear name-based skew.
The uncomfortable version of this: the model learns from past decisions and then repeats them
faster. If the historical pattern was uneven, the tool does not correct it. It scales it.
Vendor language about “the platform decided” is not much of a shield. In active litigation around
large HR systems, courts have been willing to treat the tool as part of the employer’s process.
That is the point founders need to hear. Buying the software does not move the responsibility.
Where Growth-Stage Companies Actually Get Exposed
Rejections with no human review. If a candidate is dropped and nobody can say why, you do
not have a process. You have a black box.
Tools nobody vetted. If the vendor cannot tell you what the model was trained on, whether it
has been tested for bias, or how demographic data is handled, it does not belong in the funnel.
No record of what the tool touched. If you cannot reconstruct which system influenced who
advanced, you will not like that conversation later.
No way for a candidate to get a human look. More rules are pushing in that direction. Most
vendor setups do not give you that path unless you build it.
What Basic Hygiene Looks Like
You do not need a 40-page AI policy. You do need a few things:
• Map every point where a tool influences who moves forward.
• Ask vendors the ugly questions before you turn the tool on.
• Put a named person behind every consequential decision. The tool informs it. A person
owns it.
• Keep a simple record: what was used, at what stage, and who reviewed it.
• Periodically look at who the tool advances versus the pool it started with. If the slate keeps
getting narrower in the same ways, stop and look at the model.
Why This Hits Harder at the Executive Level
When a VP or C-suite candidate gets screened out by a tool and later challenges the process,
the question becomes simple. Did a person review this, or did the system just drop them?
A documented human review does not make you bulletproof. It does put you in a different
position than “the algorithm filtered them.”
That is also why we still run executive search as a named-person process. Not because it
satisfies every statute. Because someone has to be accountable for the call.
The Bottom Line
If AI touches who you hire, the responsibility stays with the employer. Building a little
governance now is cheaper than explaining the absence of it later. It is also how candidates
decide whether they trust the process.
This is not legal advice. If your hiring footprint crosses multiple states or countries, get counsel
who lives in this. The point here is narrower: do not let a tool make a decision you cannot stand
behind.
Andcor: Human Accountability at Every Step
At Andcor, the assessment of a candidate, the recommendation to a client, and the
accountability for the outcome sit with us, not with a system. We use tools where they help. We
do not hand them the decision.
If you want a search process built that way, we are ready to talk.