Get legal team buy-in, as early on as possible

Legal will be a big part of the process, because they are responsible for minimizing risk, analyzing and proving validity, lodging and debating ethical considerations, and pondering worst-case scenarios. That is to say nothing of the usual commercial and contractual duties they will have to perform.

It can be an arduous process, make no mistake. Legal diligence can extend a project’s implementation timeline by three to six months (and sometimes even longer). Here are some tips to help speed things up.

• Prepare the boilerplate documents early. Legal will want to see documents on GDPR, general data security, and privacy policies. Get these from your technology provider as early on in the process as possible. If your chosen provider is on the ball, they’ll have these documents packaged and ready to hand over.
• Ask your provider for an independent validity study. Many technology providers do not offer this as a matter of course, but it will help give your risk-minded lawyers peace of mind. Even knowing, at the start of the process, that a validity study will be done, and by an independent third-party, is enough to secure buy-in.
• Prepare answers to the following questions. Again, this is something you’ll need your provider to help with. If they don’t have ready-made answers to these questions, that’s a big red flag. Ask, and answer:
  • What training data do you use?
  • Can you explain why person Y was recommended by the Ai, and not person Z?
  • What assumptions and scientific methods back the product? Have they been validated?
  • What are the bias tests you use, and how often do you test for bias?
  • How can you actually remove bias from an algorithm?
  • On which minority groups have you tested your products?
  • What kinds of success have you had creating hiring equity?
  • What is the composition of the team building this technology?

Remember that Ai is a novel and complicated technology. At Sapia, we know that our solution and our processes can stand up to the most rigorous legal and ethical scrutiny, and so we help you with the necessary diligence. However, not all Ai systems are made like ours – and very few, if none, have their own self-regulatory fairness framework.