Although artificial intelligence (AI) continues to dominate business conversations, widespread adoption is still some way off. In fact, just one-third of professionals said that AI programmes are being scaled across their organisations.
Many companies remain hesitant to move beyond pilot projects, with a lack of trust a major barrier. Concerns over reliability, data protection, IT security, impartiality and the potential for misuse also continue to slow adoption.
To close that gap, businesses need to understand what is behind it, how to close it and how to ensure these processes scale across the whole enterprise.
What’s behind the trust gap?
Lack of trust in AI systems continues to hold organisations back. A commissioned study conducted by Forrester Consulting on behalf of Cognizant found that only 57% of AI and data teams fully trust the outputs of AI systems. Among product managers and software developers, this figure drops to just one-third, pointing to a wider lack of confidence as AI becomes more embedded in day-to-day operations.
Developers’ fears are based on a variety of factors. Hallucinations, when AI models generate information that appears credible but is in fact false, mean workers are concerned about its reliability in high-stakes or sensitive use cases.
Likewise, cybersecurity, including the security of data entered into AI systems, the risk of leakage to third parties, the potential for attackers to compromise AI environments, compliance with ethical standards, and whether bias in systems could shift over time as they continue learning during live operation are all contributing to workers distrust of AI models.
Building trust in AI
As organisations look to build confidence in AI, quality management is becoming a bigger priority. In fact, 79% of workplace decision-makers see a direct link between trust in AI and active quality assurance measures such as regular testing, monitoring and oversight.
Trust can only be built when both AI systems and the data behind them are subject to comprehensive quality controls that are applied continuously, not just as a one-off check before deployment.
A strong foundation starts with data quality. Organisations need confidence that the information feeding AI systems is complete, up to date, and accurate. AI outputs are only as reliable as the inputs they depend on, so flawed or incomplete information will inevitably lead to poor results. Robust data governance is therefore essential to ensure AI-driven decisions to hold up in real-world use.
Quality management must also cover security, compliance and access controls. Organisations need safeguards to ensure that data does not leak from internal systems to third parties and that AI complies with regulatory requirements such as the EU AI Act and GDPR. Clear accountability for protecting sensitive business is equally important. For example, organisations must ensure that HR records can only be accessed by authorised HR staff, with controls in place to monitor and manage how that data is used within AI environments.
At the same time, trust is not solely about reducing risk. Organisations want to ensure AI investments deliver measurable value, with more than half of companies citing improving return on investment (ROI) from AI deployments as a key priority. This demonstrates that reliability and performance are just as important as security and governance.
Why quality management still struggles to scale
For many businesses, the challenge is no longer deciding whether to use AI, but how to manage it reliably at scale. That is proving difficult because quality management processes are still heavily reliant on manual oversight.
Quality management needs to extend across the entire AI lifecycle, from model design and training through to testing and ongoing monitoring in production. Yet among surveyed companies that conduct regular testing, just 15% to 29% have implemented automated quality assurance processes at any AI development stage. This means that quality controls are often conducted entirely or partially manually, making them time-consuming, costly and prone to human errors.
Ultimately, this limits widespread AI adoption. Large-scale deployment is difficult to achieve when quality assurance relies heavily on manual effort, as scaling would require a significant increase in personnel.
Many organisations are also trying to manage AI adoption without clear governance, with half still lacking an enterprise-wide AI governance strategy. This is often compounded by a lack of expertise and limited understanding of how AI systems arrive at their results. At the same time, pressure to deploy AI quickly can divert resources away from testing, monitoring and oversight.
The consequences of inadequate quality management are quickly felt, particularly when AI systems interact directly with customers. Beyond the risk of regulatory fines, organisations often face dissatisfied customers, missed business opportunities, and reduced productivity when AI-generated results create more work than efficiency gains.
A stronger focus on quality pays off
As long as concerns around reliability, security and governance remain unresolved, businesses will struggle with the successful, large-scale deployment of AI. In line with this, three-quarters of decision-makers believe that more effective AI quality assurance would see their organisations experience a significant or transformative impact on user trust.
Companies must elevate AI to a new level of trust through a combination of suitable toolsets, robust governance frameworks, process expertise and technical know-how. Only by embedding these foundations can organisations deploy AI profitably and sustainably over the long term, while building the confidence necessary to scale its adoption responsibly.
