Numerous surveys show that businesses are failing to deliver a measurable return on their artificial intelligence (AI) investment. A big part of the problem, at least for Bernhard Schaffrik, principal analyst at Forrester, is that AI providers are failing to take into account the human impact of their technology.
“They completely ignored the human factor, and also the enterprise factor,” he told Computer Weekly during the CamundaCon 2026 conference in Amsterdam.
There is certainly a fear among employees that AI will take away their jobs. In its Building a pro-worker AI innovation strategy paper, published in 2025, the Trade Union Congress recommended that employers build a meaningful worker participation at every stage of the deployment of new AI technology to drive effectiveness of technology – from strategy development to problem definition, through to tender, application design and deployment.
While change management has always existed, Schaffrik points out that a big difference with AI – and specifically with generative AI – is that, because it’s so accessible, it usually starts as a boardroom discussion. “CEOs immediately understand the potential, so they have been pushing it even harder into their organisations,” he said.
In his experience, CEOs assume that people down the line who are in charge of change management, such as the human resources team, or the people in the business who are driving these transformation programmes, will handle change management. However, he said: “Since there is a direct implication on jobs, with job roles changing and people being displaced, the fear and concerns among employees increases exponentially.”
According to Schaffrik, not only are employees afraid and confused, those people who are supposed to implement the AI are also being impacted.
Along with a lack of addressing work relations effectively, he said that AI providers usually do not really consider the rigidity of enterprise technology frameworks and business processes. “Businesses don’t want to break the payroll process,” he said as an example, which means business leaders need to balance risk. “This is why AI providers are surprised when a deployment of their technology doesn’t work as intended. It’s a mix of human psychology and company inertia, as well as regulations and legal stuff.”
None of these things are new, but Schaffrik believes CEOs and other business decision-makers need to assess what AI is being used for in their organisation.
“If I were a CEO, I’d aspire to automate as much repetitive work as possible and I would be happy to deploy any enterprise-grade technology that allows this, such as workflow engines, robotic process automation, document processing – whatever technologies are available – and that also includes AI agents,” he said.
At the same time, he said business leaders should also strive to automate less repeatable processes that human workers find particularly challenging, such as making mistakes when comparing multi-page documents.
As Schaffrik points out, this can sometimes occur when someone in the legal team is asked to compare three versions of a large contract. “People make mistakes, but if you put hallucinations to one side, then AI is much better at doing these things,” he said.
As for handling AI hallucinations, this is where Schaffrik sees a need for having the human-in-the-loop. But to be a good checker of AI outputs and command a good salary for doing this job, he said that employees need to excel at the work the AI is taking over. In other words, a legal expert needs to be extremely proficient at analysing different versions of multi-page contracts.