AI-adoption maturity

Let’s study the feasibility, desirability and business viability of an AI-based recruiting solutions: are people ready for it?

Perception of AI

Are recruiters and candidates ready and willing to adopt an AI-powered solution marketed as such? Our hypotheses are:

• ‘People perceive AI positively.’ According to PandoLogic, 80% of executives believe that AI can help make recruiting more efficient (PandoLogic, 2018).
• We found no data about a positive or negative perception of AI by candidates. Chances are that the perception is negative. Mine is. We’ll study if this opinion is widely spread or not.

Business Viability

How much would recruiters be ready to spend in our service? Our hypotheses are:

• ‘Companies spend a lot on their sourcing tools, and the amount of spending would be proportionable to the perceived added value brought by the new solution envisioned.’ According to a LinkedIn study, the average company spends 30% of their recruitment budget on advertising on job sites (Semiocast, 2017).
• ‘The cost of a new hire/ bad hire/ turnover justifies the price of the solution.’ According to Appii, the cost of a new hire is between £50-£2,500 (Appii). An average of $4,000 (Yobs, 2018). Cost per hire average is $4,129 for PandoLogic (PandoLogic, 2018). 30% of the employee’s first-year earnings (US Department of Labor). For Retalent, new hire fails after 18 months (Retalent, 2018). Rebric suggests “people stay in a job an average of only two years” (Rebric, 2018). Yobs reports “On average there is an annual turnover rate of 15–19% within the US.” (Yobs, 2018) 1 in 3 quit their jobs after only six months (Elevated).

The price should then be calculated based on these budgets.

Once a good price is set, in relation with the value the AI can bring, what are the obstacles we might face? What could be the motives for mistrusting AI? Our doubts are:

• ‘Machine-learning algorithms are good to track and reproduce patterns (similar profiles), but could also be used to track complementary traits. Which one is the most needed?’
• ‘Machine-learning algorithms are good to reproduce patterns; how can we avoid to replay the current discriminations?’ As a reminder: the French Labor Code (Article L.1132–1) prohibits any distinction between employees based in particular on : origin ; sex; morals; orientation; sexual identity; age; marital status; Pregnancy; genetic characteristics; membership or non-membership, real or perceived, of an ethnic group, nation or race; political opinion; trade union or mutualist activities; religious beliefs; physical appearance; family name; place of residence; state of health; disability.
• If a discrimination occurs, what forms could legal remedies take? It’s easier to incriminate the responsibility of a human being than a deaf and mute algorithm. We need to design a process for this.’

Test image: would you trust ‘Seqy’ for your recruiting process?