Due diligence on an AI recruiting technology business

Diligence on an AI recruiting technology business needs the standard small-business review — financials, contracts, corporate standing — plus a specific set of checks most buyers have never had to run before, because this product makes decisions that carry legal consequences for the people it screens. Skipping the standard review is a mistake anywhere; skipping the recruiting-technology-specific checks here is the mistake that costs the most.

Verify the bias-testing record is real

Ask for the actual validation studies and adverse-impact testing results on the core screening or ranking model, not a summary or a marketing description of the testing that was done. Confirm when the testing was last run, since a model that’s been retrained or meaningfully updated since its last validation may not be covered by results the seller is pointing to. Where no testing exists at all, that gap doesn’t automatically end the deal, but it needs to be treated as an active, quantifiable risk being inherited on closing, not a formality to note and move past.

Trace candidate data consent

Review how candidate personal information was collected, whether it was ever used to train models shared across more than one customer, and what consent language covered that use at the time. Where personal information from Quebec candidates was involved, check specifically whether automated-decision disclosure and human-review-request obligations under Quebec’s privacy legislation have actually been met, since this is a distinct requirement from general federal privacy compliance and is easy for a fast-growing company to have missed.

Confirm the IP chain, especially contractors

As with most AI products, the single most common gap found in diligence is a contractor who built part of the screening or ranking model without ever signing an agreement assigning IP rights to the company. Review every contractor and vendor agreement that touched the model, the training pipeline or the underlying code, and treat a missing assignment as something to close before signing, not after — while there’s still a business relationship and leverage to ask for it.

Check Ontario disclosure compliance directly

Where the product is used by customers posting jobs in Ontario, confirm directly — don’t take the seller’s word for it — that those postings carry the AI-use disclosure Ontario’s employment-standards legislation requires. This is a customer-facing compliance obligation that sits partly with the recruiting-technology vendor and partly with the employer using it, and a buyer needs a clear picture of which side of that line the target’s contracts and product design actually fall on.

Corporate, contract and complaint history

Beyond the recruiting-technology-specific items, run the standard corporate diligence — confirming good standing, checking for any outstanding CRA debts, reviewing the federal registry of individuals with significant control against who the seller says actually owns the company, and reviewing customer and staffing-partner agreements for assignability on a change of ownership. Also ask directly whether the business has received any human-rights complaint, privacy-commissioner inquiry or customer complaint related to discriminatory or inaccurate screening outcomes, even one resolved informally, since that history is relevant regardless of its outcome.

Check contract assignability and brand ownership

Review the assignment language in the largest customer subscription agreements and staffing-partner contracts directly, since some enterprise and public-sector customers who selected the product through a formal procurement process built in a right to reapprove or terminate on a change of ownership — a provision a buyer needs to know about well before closing, not after. Separately, confirm that the trademarks, product domain and brand the business trades on are actually registered to and owned by the company, rather than a founder personally or an agency that built the brand without a documented assignment. Neither check takes long, and both are common enough gaps in a fast-growing technology company that skipping them is simply a wasted opportunity to catch something cheap to fix before closing and expensive to discover after. It’s also worth confirming that the product domain is registered for a reasonable term and isn’t set to expire shortly after closing, and that no trademark application central to the brand is under an active opposition or has lapsed — either one is a fast, inexpensive check a buyer’s counsel can run directly against the public registries.

What a finding actually means

An undocumented bias-testing gap usually means pricing the acquisition to reflect the cost of remediation and the residual legal exposure until it’s fixed, sometimes backed by a holdback or a specific indemnity. A missing contractor IP assignment means obtaining a retroactive assignment before closing wherever possible, or addressing the risk directly in the purchase agreement where it can’t be. And a provincial disclosure gap usually means fixing the product or contract language promptly after closing — it rarely kills a deal on its own, but it does mean the buyer inherits an active compliance project on day one, not a finished product.