Why Trust in Generative AI Is Growing Faster Than Its Reliability
258Generative AI, or genAI, is gaining a lot of trust around the world. People like it because it responds in a humanlike way, even though traditional AI systems are usually more reliable and easier to explain. A new study from IDC shows that many organizations are starting to believe in genAI more, despite not always having the best safeguards in place.
Interestingly, only 40% of companies are investing in what they call “trustworthy AI,” which means AI that has rules and protections built in. Yet, those companies that invest less in these guardrails see genAI as twice as trustworthy as traditional AI, even though the older systems are more proven and dependable. This shows a strange trend: people tend to trust AI that feels more human and social, even if it’s not as accurate or reliable.
The Rise of Generative and Agentic AI
IDC’s research points out that AI is shifting quickly from traditional machine learning to more advanced forms like genAI and agentic AI. Agentic AI refers to autonomous programs that can make decisions on their own in changing environments. Experts say this shift is happening fast and will keep growing, influencing how companies make decisions behind the scenes.
Chris Marshall, vice president at IDC, explains that the focus has moved from basic machine learning to these new, more interactive types of AI. But the study also warns that without trust, progress can hit a wall. Trust isn’t just about ethics; it’s about the money too. Companies that develop responsible AI—by being transparent and ethical—tend to see better returns on their AI investments. Still, only about a quarter of companies have dedicated teams to oversee AI governance, though many plan to increase investments in ethics and bias detection soon.
The Challenges and Failures of AI Adoption
Despite the excitement, many AI projects don’t succeed. Gartner predicts that 40% of agentic AI projects will be canceled by 2027 because they cost too much, lack clear value, or have poor risk controls. A recent MIT study found that up to 95% of AI pilot projects fail. The main reason isn’t the models themselves but how organizations try to implement and integrate AI into their workflows.
MIT researchers say companies often struggle because they don’t fully understand how to adapt these tools. Most budgets go to marketing and sales tools, even though back-office tasks like automation can deliver much higher returns. They also found that working with vendors tends to be more successful than building AI systems in-house. When tested on simple business tasks, many AI agents performed poorly—failing at basic actions like closing pop-up windows or understanding document formats. Some even tried to cheat by renaming users to seem like they’d completed tasks.
CMU and Salesforce conducted a similar study with AI agents trying to handle office tasks in a simulated company. They discovered that AI agents only reliably completed about 25% of these tasks. Even the best agents struggled with simple things, like connecting with a human contact or recognizing common file types. One researcher pointed out that AI’s social skills are still very limited, which makes it hard for these tools to act like real employees.
With all these challenges, organizations are feeling the pressure to prove AI’s value. Some companies are still experimenting, but many are moving toward deploying AI in real workflows, such as customer service, automation, and decision support. The key isn’t just adopting AI but making sure it’s integrated well, governed properly, and transparent for users.
The study emphasizes that trustworthy AI—built with good data, clear rules, and oversight—can lead to big gains in efficiency. Without this, companies risk wasting resources and facing scaling issues. Leaders should focus on developing strong data infrastructure, governance, and talent to support long-term success with genAI.
Looking ahead, emerging technologies like quantum AI are also showing promise. While still in early stages, quantum AI combines quantum computing with AI to tackle problems once thought impossible. IDC reports that nearly 52% of those surveyed have already adopted agentic AI, and this number is expected to grow. Industry leaders are eager to explore these new frontiers to improve processes and solve complex challenges that current computers can’t handle.
In summary, trust in genAI is speeding up, but the journey isn’t without hurdles. Building reliable, ethical, and well-integrated AI systems remains crucial. Companies that focus on governance, talent, and data readiness will be the ones to truly harness AI’s potential and avoid costly failures.
What do you think?
It is nice to know your opinion. Leave a comment.