metamorworks -

How an AI startup is trying to fix gender bias in the workplace

Startup Pipeline Equity provides enterprises with an AI platform to help make decisions about hiring, pay, performance, potential and promotion -- free of bias.

When Katica Roy returned to work after the birth of her daughter, her supervisor asked her to take on two new teams, tripling her workload in a matter of two weeks without additional pay or a promotion. Meanwhile, management asked a male colleague to take on one extra team. With his new responsibilities came a new promotion and additional compensation.

In order to receive the pay equity due to her, Roy notified her human resources team about the Lilly Ledbetter Fair Pay Act, a federal law that helps ensure pay practices are nondiscriminatory and fair, without gender or other bias, by making it easier to file equal pay lawsuits.

While she ended up succeeding in her gender bias protest, the process led Roy to found and become CEO of Pipeline Equity, a SaaS vendor that uses cloud-based AI, machine learning and natural language processing (NLP) technology to improve the financial performance of its users by trying to close the gender equity gap.

Trying to fix the gender equity gap

Pipeline, based in Denver, examined gender equity disparities among women, men and nonbinary people in the workplace when it developed its platform.

In its research, Pipeline found that there currently isn't a country where women are equal in terms of pay, and that things have worsened due to the pandemic.

"In the last year, we've added 11 years to the time to gender equity in the workplace," Roy said during a keynote presentation at the 2021 AI Summit New York on Dec. 8. "We are now 268 years from gender equity in the workplace."

Screenshot of Roy presenting with a slide about gender equity worldwide
Katica Roy of Pipeline gives a keynote on how AI can be used to reduce gender equity bias, at the 2021 AI Summit New York.

In a study across 4,000 companies in 29 countries, Pipeline found that for every 10% increase in intersectional gender equity (gender, race, ethnicity and age), there's a 1% to 2% increase in revenue. Based on that research, Pipeline created an augmented decision-making model that helps organizations make decisions about internal hiring, pay, performance, potential and promotion.

Pipeline provides APIs into the HR systems of its users so that when enterprises make decisions about those five categories, those decisions run through the platform's algorithms, and the Pipeline system recommends an action.

An example is when a manager writes a draft performance review about an employee; Pipeline's algorithms and NLP tools read through the performance review and flag any bias.

In the last year, we've added 11 years to the time to gender equity in the workplace ... We are now 268 years from gender equity in the workplace.
Katica RoyCEO, Founder, Pipeline Equity

While Pipeline doesn't release data on how often its customers follow the platform's recommendations, Roy said organizations that use the platform increase equity in their organization in the first three months of adopting the platform. "When we provide a recommendation, your choice now if you choose to reject that recommendation ... is actually you're choosing potentially to be inequitable," she said in an interview.

Considerations in letting AI address the gender bias gap

Gender inequalities are rooted deep in biases, said Kathleen Walch, an analyst at Cognilytica. She said for organizations that are looking to use AI to eliminate or reduce those biases and close the gap, it's important to keep a human in the loop.

"Never let the AI be the sole decision-maker," Walch said. "Monitor, double-check things, [quality control] results so that you are checking what's going on. If you don't do that, things can go awry really fast."

Walch said that without humans involved, AI systems often can continue to perpetuate some of the gender inequalities. An example of this is when Amazon used an AI hiring system that discriminated against women, though the tech giant stopped using it when it realized it couldn't fix the system.

Walch added that another way to keep from furthering the gap is for organizations to continue to ask questions about the data brought in to train the system and which employees are on the team doing hiring and promotions. The team, she said, must be representative to avoid more error and biases.

This means organizations, especially large ones, need to upskill and re-scale some of their workforce to achieve representation across different teams. They also need to educate their current workforce to look at data sets with a focus on equitable gender and other representation.

Pipeline's traction

For a vendor that is nearly five years old, Walch said she would be interested in seeing the traction Pipeline has achieved in the past few years. While the vendor is bringing awareness to a topic that needs to be addressed, Pipeline needs to show that it's different -- that it's gained customers and generated funding, as well as proving it's helped reduce the gender bias gap in organizations.

"Unless you have critical mass, then it doesn't necessarily make an impact," Walch said. "You need to have widespread impact at very large organizations so that it really starts changing the game."

Earlier this year, Accenture Ventures invested in Pipeline, but did not disclose the amount of funding. Through July 2021, Pipeline had secured a total of $2 million in seed funding in two rounds, from 12 investors, according to Crunchbase.

List pricing for the Pipeline platform is $60 per employee per year for all modules, Roy said.

Dig Deeper on AI technologies

Business Analytics
Data Management