Editorial
The future of digital work will not be defined simply by AI capabilities, but by how well leadership can create a partnership between their workforce and AI systems. Over the past few years, we've seen companies implement technology to automate tasks, streamline workflows and simplify decision-making. From content generation to customer service automation, these advancements are changing how employees work, with an eye towards increased efficiency.
While AI’s ability to boost productivity is widely acknowledged, its risks are often misunderstood or overlooked. Companies may dive headfirst into AI without recognizing the blind spots that come with it, like algorithmic bias, data inaccuracies or the non-deterministic nature of AI models.
A Gartner survey in May found that nearly half of AI leaders find it difficult to prove AI’s return on investment, with other concerns including a lack of skills and confidence in AI technologies. Despite widespread AI adoption (as 95% of IT leaders report AI has been implemented in at least one business process), many organizations are not seeing the expected return. This is likely because these companies don’t have the right human oversight in place to truly optimize and build on AI’s capabilities.
AI Risks: Bias, Hallucination and Transparency
Organizations are eager to leverage GenAI to unlock new value for their business, but these solutions aren’t turnkey. Keeping humans at the center of your AI strategy is key to minimizing the top risks inherent in AI use.
For example, AI bias is a key concern. Models learn from historical data, and if that data contains bias, the AI can carry those biases forward. But, by carefully curating data, selecting representative datasets and continually monitoring AI outputs, organizations can reduce the chances of biased outcomes. In hiring, for example, this helps avoid reinforcing existing inequalities and promotes fairness in decision-making.
Another challenge is AI hallucinations, where models generate seemingly credible but incorrect information. Here, human involvement plays a critical role. By setting guardrails, selecting the appropriate models, and deciding how and when to integrate large language models into workflows, teams can prevent errors that might otherwise lead to poor business decisions, particularly in high-stakes areas like finance or sales.
Finally, while AI models, especially large language models, can sometimes lack transparency, businesses can improve trust by involving humans in reviewing and explaining AI-driven insights. This oversight ensures that AI isn’t used as a black box but as a tool that employees can understand and trust, encouraging broader adoption and reducing resistance across the organization. For example, Anthropic's research with the Claude 3 Sonnet model mapped the model’s inner workings to understand how neuron-like features affect outputs. This transparency is crucial for mitigating risks and ensuring that AI models behave as intended.
Related Article: Reduce Uncertainty to Drive AI Adoption
Create AI Governance Frameworks
Companies must embrace a balanced approach where AI amplifies human capabilities within a well-defined governance framework with human expertise remaining central to critical AI decision-making.
To effectively integrate AI into operations, organizations must identify a reliable "AI Advisor." The advisor role may be filled by the company’s CTO or CIO, or by partnering with a GenAI consultant or expert, who can guide the selection of appropriate technology, ensure successful integration, and oversee upkeep and updates. The last step is where many organizations falter. They tend to expect immediate gains without implementing a clear change management strategy to promote long-term buy in.
The day-to-day users of AI don’t need to be data scientists or AI engineers, but these employees must know how to use AI effectively to drive results in their specific areas and make the AI work for them. A marketing manager overseeing an AI-driven campaign, for instance, needs to recognize flawed insights without implicitly trusting the model’s output.
These interactions then create a powerful feedback loop where AI generates initial outputs, and humans refine and improve them, going far beyond simple fact-checking. The process introduces the crucial context and judgment that AI lacks. While AI models excel at processing historical data, they often struggle to adapt to real-time market shifts, regulatory changes or the nuances of human behavior. By maintaining human oversight, organizations can ensure AI-driven decisions are grounded in reality and aligned with broader business objectives.
Conference
Sep
16
HR Tech Conference Las Vegas 2025
Register
Conference
Oct
27
Gartner HR Symposium/Xpo Orlando 2025
Register
Related Article: Generative AI, the Great Productivity Booster?
Recommendation: Start Small
Are you looking to integrate AI into your business process but don’t know where to start? You’re not alone. Guidehouse found that 76% of respondents say their organization is not fully equipped to harness GenAI.
Let’s first acknowledge that industries and companies have unique pain points, and will need customized software to help push their processes forward. While there’s no “one size fits all” in technology, there are a few tried-and-true methods to jumpstart use. These include:
- Don’t bite off more than you can chew: Start with small, high-impact use cases where AI can quickly demonstrate value. Use these wins to build momentum and encourage adoption.
- Make sure your AI can talk back: Not literally, but ensure transparency in AI systems, where they can provide employees with clear explanations of how decisions are made and enable them to intervene when necessary. This can include asking your technology to generate citations and references that trace the origin of the data used, as well as an audit trail of its decision-making process.
- Intervene early and often: Reliable AI is strong engineering, so integrate testing processes into the software development life cycle with human oversight.
- Create guardrails for security and compliance: Train employees on how to work with AI systems. But don’t forego a stringent policy when it comes to remaining secure, including access controls and authorizations on critical data transmissions.
The real power of AI doesn’t just come from technology. It comes from how we as leaders make sure it complements and amplifies the strengths of our people. Start small, rack up a few wins and build from there. Think of AI as a plant. For it to flourish, you have to water it, nurture it and sometimes trim it back. Do it right, and you’ll create something worthwhile. Do it wrong, and you’ve just got an expensive, dead houseplant.
Learn how you can join our contributor community.
