A common occurrence in talent acquisition referred to as the “search for the purple squirrel” leaves both applicants and recruiters disappointed. This search refers to the costly quest to find the perfect candidate for a role. Like purple squirrels, these candidates do not exist. Because of the evolving nature of an organization’s needs, the perfect candidate will never be found. On top of this frustration, in recent years, organizations see more and more offer rejections come in.
Facing these less than ideal realities has inspired organizations like SLB to take a closer look at their talent acquisition processes and explore ways to improve their talent pipelines. In a recent Dataiku Product Days session, Modhar Khan, People Analytics Manager at SLB, walked us through why and how SLB leverages Dataiku in AI and machine learning workflows to improve (and infuse responsibility into) talent acquisition.
The Talent Acquisition Challenge
With over 100 years of experience and operations spanning across 120 countries, SLB is a trailblazer in energy exploration and production services. Employing over 86,000 employees worldwide, SLB aims to champion the strong culture of diversity and inclusion that they have built over many years as their organization expands. In order to meet these diversity and inclusion targets and approach talent acquisition in a transparent, efficient, and fair manner, the collected data needs to help and not hinder the consideration process for each applicant.
While in many cases talent acquisition volume is a good problem to have, it means that the talent acquisition process is placed under high pressure. The average number of applicants to SLB reaches around 500,000 each year. The data received from these applications supplies a wealth of information for diversity and inclusion insights but, on the flip side, the plethora of profiles also places a heavy responsibility on recruiters. A large number of applicants does not always equate to an excess of talent that will serve the organization well. In fact, having this large quantity of applicants means that the management of talent acquisition workflows is more complex and requires greater attention than ever before.
This is why SLB has turned to Dataiku to use AI to optimize and responsibly manage the data-heavy workflows of talent acquisition.
Introducing AI Into the Process
Something crucial to be aware of is the sensitive nature of talent acquisition data. Introducing AI systems to this data is a high-risk initiative. Ensuring responsible action around the AI applications is of the utmost importance in order to make sure that any biases and errors at scale are controlled. As new data security rules emerge, managing the clarity, consistency, and confidence in AI systems becomes increasingly critical. SLB has clearly defined each component of their Responsible AI project to reinforce this.
SLB's AI Project Outline and Results
Here is a brief breakdown of the components and outcomes of the integration of AI into talent acquisition workflows at SLB:
Objectives
- Expand candidate opportunity profile matching from a single acquisition to multi-requisition based on candidate features.
- Improve the efficiency of recruiters finding the right match for a role.
Methodology
- Evaluate the features and historical data available in the data warehouse for recruiting and HR records.
- Enable project objectives while testing for biases and drifts with machine learning.
Success Metrics
- Blind test with recruiters on candidates acceptance/rejection to establish percentage accuracy.
- Silence test with real data monitoring to compare model generated insights to actual results.
- Confusion matrix evaluation of model performance.
Results of the AI project have shown an 82% agreement in the blind test between model outcome and recruiters’ selection. In the silent test, 95% matching was found for candidates who accepted and 72% for candidates who rejected offers. This demonstrated that the process including model predictions results in a higher rate of offer acceptance.
Dataiku Enablement
Taking off all the way from ideation and inception to application and deployment in under two months, Dataiku has introduced capabilities for the continuous monitoring of model performance and advanced bias testing to talent acquisition workflows at SLB. In Dataiku, the historical data is tested and applied to models to implement with new applicant information. The API endpoint allows the results of the models to be easily shareable across the talent acquisition team.
In addition to the initiatives regarding AI fairness, Khan also highlighted the ease of integration and customer support system that is inherent to working with Dataiku. Dataiku is unique in the way that it allows SLB to introduce data from many different sources into a common platform and also utilize API deployment, as mentioned above, in the infrastructure. Not only is the integration process easy, but the customer success team is always readily available to reduce any barriers to solutions and support the talent acquisition team as they collect insights from the AI and machine learning processes. Seamless integration and accessible support systems have helped SLB avoid unnecessary costs and save time as they scale AI in their talent acquisition processes.
Having the right talent is crucial for overall organizational success, and infusing responsibility into the talent acquisition AI processes at SLB has been a key step to making sure that talent is sourced in an equitable and effective way.