Not long ago, I watched an experienced executive use generative AI (GenAI) to prepare for an important customer conversation. Within minutes, the system had gathered industry information and suggested questions while challenging parts of the account strategy. The preparation was faster than it would have been a few years ago, but the useful part came when the executive started questioning what had been produced.
Some recommendations were technically sound but did not fit the customer. Others missed the history behind earlier decisions. The technology had assembled the information. The executive still had to determine what was relevant and how the relationship should be handled.
That kind of judgement is becoming more important as AI moves further into everyday work. South African findings from the latest SAS and IDC research show that inadequate context is the leading local reason for overriding AI recommendations, cited by 35.1% of respondents. People still need enough experience and understanding to judge whether an answer fits the situation they are facing.
AI needs people who know the work
AI is already taking time out of routine work. What interests me is how the quality of the work changes once people have that extra capacity. In a customer role, faster research is useful because it leaves more room to think about the account itself and the judgement needed before a conversation.
An account executive can walk into a customer meeting better prepared. A manager can spend less time gathering information and more time working out where an employee needs support. In both cases, AI gives the person more to work with, but experience still influences the decision.
SAS research on agentic AI makes a similar point. As AI systems take on more action across workflows, lasting value depends on organisations scaling automation without sidelining human judgement. People still need to set goals and define guardrails around the system, particularly where the consequences of a poor decision are high.
This is what I mean by a bionic enterprise. Human expertise and AI capabilities work together, and the organisation gets better at scaling what works from one person or team across the wider business.
When one good practice spreads
One employee becoming faster with AI is useful, but the gain can remain isolated if nobody understands why the approach worked.
Consider an account manager who develops a strong AI-assisted way of preparing for customer conversations. If that practice stays with one person, the benefit stays local. If the organisation understands what made it useful and incorporates that into how other teams work with accounts, it becomes an organisational capability.
Managers are important here because they sit close to the work. AI can give them more context about where someone may need help or where a pattern is emerging across a team. The manager still has to interpret that information in light of what they know about the person and the environment in which they are working.
Skills development has to reflect that reality. Employees need enough domain knowledge to question an output with confidence. Training people to operate an AI tool may improve adoption, but it does not automatically improve the quality of the decisions made with it.
Trust is tested in everyday use
People learn pretty quickly where an AI system helps and where it can get things wrong. That experience is important. It gives employees a better sense of when an output is useful and when they need to question it.
The commercial case is becoming clearer too. The latest SAS and IDC study found that organisations applying trustworthy AI practices were 15 times more likely to report strong or high ROI from their AI projects. Organisations with the strongest governance, data quality, and auditability practices also reported at least double the ROI from AI deployments.
For employees, those foundations show up in practical ways. They need enough visibility into the basis for a recommendation to decide how much weight to give it. They also need to know who remains accountable when the system gets something wrong.
Leadership has to reach beyond the technology
Leaders need to look beyond how much AI the organisation has bought or how many employees have access to it. The harder work sits in understanding where AI can genuinely improve the job and what has to change around the technology so people can use it well.
Sometimes that will mean redesigning part of a process instead of adding an AI tool to the process that already exists. It may also change the manager’s role in adoption, because employees still need help building judgement around the technology and recognising where its limits show up in real work.
The executive I watched did not get value from AI because the technology made the decision. The value came from being able to work faster without giving up the experience needed to question what came back.
That is the larger challenge for organisations. AI can extend what people are able to see and do. The enterprise gains more when human judgement is built into the way the technology is used and when a good practice can move beyond the person who discovered it. That is what turns an AI investment into a bionic enterprise.
