How AI Bias is Impacting Nearly Every Sector of Industry

While there are no 2025 or 2026 reports that estimate the average costs that businesses pay due to AI bias, numerous studies show that companies' bias is affecting sectors significantly across the board.

Bias in HR systems, as automated hiring tools become the standard, is the most visible one. Stanford University recently followed 3.4 million people who submitted 4 million job applications to 1,700 job postings across 150 employers and 11 industry sectors and found that AI HR hiring tools can yield racial bias and systemic rejection.

AI bias in HR is also spilling over to courts, putting business reputations at risk, as in the case of the HR tool firm Workday, which is facing a trial in California over claims of allegedly having discriminated against applicants based on factors such as age, disability, and race.

Beyond HR and business sectors, AI bias is also a top concern in communities. For example, the AI chatbot Aisha, built by Chromatics, was created to center Black voices, history, and lived experiences while prioritizing safety and privacy, and reducing risk and bias. AI startups, too, are beginning to hire philosophy majors to help tackle AI bias.

With AI bias affecting communities, factory floors, and businesses from different sectors, decision-makers are scrambling to navigate the AI bias fix to reduce harm, avoid damages, and mitigate the risks and costs, while creating better services and products.

The Impacts of AI in Different Industries and Business Sectors

Recent reports have put AI bias back in the spotlight, even as the wider debate shifts from bias to AI governance.

On June 22, 2026, the United Nations warned that AI is getting women wrong, after a study of 133 AI systems found that 44% of these systems demonstrated gender bias, while more than a quarter showed both gender and racial bias. UN Women urged governments, companies and developers to ensure gender equality

In insurance, Stanford University researchers recently found that limited transparency and review in AI-based insurance decisions could lead to wrongful care denials. The study concluded that AI could worsen existing flaws in prior authorization processes and called for ethical governance and oversight in health care AI use.

Meanwhile, in the U.K., the British supermarket giant Sainsbury paused its live facial recognition system in south-east London after a false shoplifting accusation sparked a public outcry.

Other areas, from lending to finance, to education and health, and industrial operations also report problems with AI bias and AI drift.

Rohit Shinde, Senior Process Engineer at Atlas Prediction Control, a U.S.-based industrial automation company, told the International Business Times that industrial AI bias is not demographic, "so nobody goes looking for it".

"In a plant, the upside of an AI deployment is a few percent of optimization, and the downside is a catastrophic event," Shinde said.

In lending, Saugat Nayak, from the Dallas, Texas SMU Cox School of Business, who builds AI credit risk models for a U.S. Fintech that lends money to small and minority owned businesses, told the International Business Times that AI bias in the industry is not an academic problem.

"It is the difference between a real business getting capital or being denied," Nayak said.

Traditional credit models are trained on data from borrowers with long banking histories, established collateral and traditional revenue patterns. But minority-owned and small businesses often don't fit that profile.

"Not because they're riskier, but because their financial histories are different," said Nayak.

"So when it (AI) looks at a minority-owned business with a thin credit file, it doesn't see creditworthiness; it sees strangeness, and risk is priced as unfamiliarity."

Dr. Adnan Masood, former Microsoft Regional Director and current Chief AI Architect at UST, the global digital technology and transformation services company, headquartered in Aliso Viejo, California, told the International Business Times also spoke to us about AI bias in different sectors, including high-risk sectors like healthcare.

"In healthcare, biased inputs become clinical decisions," said Dr. Masood.

A widely cited care-management algorithm used past healthcare spending as a proxy for illness and concluded Black patients were healthier simply because the system had historically spent less on them, said Dr. Masood.

Black Box AI: What leaders and decision-makers need to know and how to build an AI vetting layer

AI black box refers to an AI system that keeps its internal process hidden from users, offering no visibility on how it reaches its conclusions or generates responses.

IBT asked Larry Adams, CEO and founder of Chromatics, the firm behind Aisha AI, how they balance technical complexity with the need for a transparent user experience through Aisha's vetting layer.

"The 'black box' concern is fair, but the technology built for Aisha is done in that it constructs a response to a user's question through a defined pipeline, and every stage of that pipeline is traceable," Adams said.

Adams explained that in Aisha, a query moves through classification, routing, grounding, verification, and evaluation before anything reaches the user.

"The vetting layer lies on the grounding stage: rather than pulling from the open web indiscriminately, we built it so that it grounds responses in vetted, culturally relevant sources," said Adams.

After that, the AI model verifies them (the sources) for accuracy and relevance, and then evaluates the output for bias and hallucination risk before delivery, Adams explained.

"When we talk about visibility into the model's reasoning, we're saying that every output is logged, traceable, and reviewable," said Adams. "And when something is flagged, it triggers automated and human review that feeds back into the system."

All of that structure lives in the architecture but is transparent for the user, said Adams.

"People don't want to audit a model to trust it; they want an answer that's accurate and grounded in their reality," said Adams. "The rigor is in how the response is constructed and validated, but the outcome is simple."

"AI bias and hallucinations are applied governance failures," Lawrence Snapp, CEO of TrustScale, an AI Trust and Verification Platform, told the International Business Times. "Businesses are deploying probabilistic systems that will be confidently wrong some percentage of the time, often without any instrument on the dashboard telling them when it happens."

The issue isn't simply whether a model can make a mistake, Snapp said. "It's whether organizations have the verification and accountability in place to catch that mistake before it becomes a decision about someone's livelihood."

How Leaders Navigate the AI Bias Fix

While some experts claim that completely eliminating AI bias is not possible due to the complex issues involved from data to training to output, other experts agree it can be dramatically mitigated.

"On a business level, the most effective way to remove bias is to train your employees to look for it," Erin Bortz, Manager of Recruiting at Huntress, the American cybersecurity company, told the International Business Times.

"When workers know what a biased output might look like, they'll be better prepared to identify and tackle it," said Bortz.

"Especially within AI outputs or generations, make sure you have a human in the loop who looks for bias and verifies accordingly," said Bortz.

Nayak from the SMU Cox School of Business told us that bias auditing should not be an optional step after a problem, but a required step before deployment. "Second, the training data should be carefully diversified," said Nayak. "Third, explainability must be a requirement for deployment, not a nice-to-have," Nayak added.

"Artificial intelligence systems that make consequential decisions about people's access to jobs, capital or services should be required to explain those decisions in human-understandable terms that can be contested," said Nayak.

Dr. Masood, from UST, told us that there is no patch that fixes bias, only architecture.

"Fix the data first," said Dr. Masood. "Trace its lineage, understand whose history it encodes, and hunt for proxy variables before training rather than after deployment".

"Make decisions explainable," Dr. Masood added. Explainable AI (XAI) techniques expose which variables drove a specific outcome, so a bank can show a loan denial came from debt-to-income ratio and not zip code. "If you cannot explain a consequential decision, you should not be automating it," said Dr. Masood.

Finally, Dr. Masood spoke about the importance of continuous auditing. "Models drift as they ingest new data, so bias testing is a monitoring discipline like security, and frameworks such as the NIST AI Risk Management Framework give you a defensible structure," said Dr. Masood, highlighting the importance of keeping humans meaningfully in the loop.

Final thoughts on AI Bias, Inclusivity, Cultural Intelligence, Personalization and Transparency

An argument among AI model developers when discussing inclusivity and cultural intelligence is the concern of making a model too narrow. However, Adams from Chromatics said that cultural nuance makes systems work more broadly, not less.

"Don't treat inclusivity and cultural nuance as a tradeoff to the balance," Adams added.

Adams added that cultural intelligence drives better business outcomes.

"Personalization isn't a nice-to-have, but a revenue driver," said Adams. In 2021, McKinsey found that companies leading on personalization generated up to 40% more revenue from it than their peers, and Deloitte research shows roughly three in four consumers are more likely to buy from brands that personalize. Harvard went as far as calling personalization the "most profitable outcome" of the AI wave back in 2024. "These insights still stand," Adams said, adding that aligning to how people actually communicate drives adoption, conversion, and loyalty.

"But personalization only pays off when data is collected transparently, and trust is built," said Adams.