AI can make individual employees faster without helping a company make better decisions. Son Nguyen, an Orient Software CTO writing in Forbes, argues that organisations need to turn what people learn using AI into shared, traceable knowledge, not just hand out more licences.
Watch Desk analysis
What happened
Nguyen says useful information often stays trapped in individual chats, while inconsistent definitions and answers without evidence make it harder for colleagues to trust or reuse what AI produces. He cites a McKinsey report in which 80% of respondents said AI improved individual productivity, but only 37% attributed any earnings impact to it. Those are figures he cites, not proof that the gap has one cause.
His proposed test for an organisation’s AI setup has five parts: answers should show their source, owner and last update; teams should share consistent meanings for key information; people and AI agents should have appropriate permissions; corrections should improve future answers; and an executive should own the outcome.
Nguyen’s suggested starting point is refreshingly unglamorous: choose recurring questions from teams such as finance or operations, then measure how long it takes to get a trustworthy answer with a verifiable source, owner and date. He calls that measure “time to trusted answers”.
Why it matters
A faster answer is not automatically a better business decision. If an AI system cannot show where its information came from, or a correction stays in one person’s chat, the next colleague may have to start the detective work all over again.
The proposed test gives organisations something more useful to discuss than licence counts: whether AI knowledge can be checked, shared and safely acted on. It is a framework from an industry executive, not a validated scorecard, but the questions are practical enough to try without first commissioning a grand platform project.
Our read
Nguyen’s strongest point is that buying access is the easy bit. Start with one recurring question, record how long a reliable answer takes, and check whether the answer’s source and owner are clear. If those basics are missing, a more capable model may simply produce confusion at greater speed.
What to watch
- Whether organisations measure trusted answers, not just AI usage or time saved.
- Whether corrections made by one employee reach other people and systems.
- Who owns the quality, permissions and ongoing cost of AI-generated answers.
Discussion spark: Should companies judge AI maturity by how much faster employees work, or by whether the organisation can turn that work into reliable shared knowledge?
Sources and evidence
- Why Individual AI Gains Don’t Add Up To Company Value – Forbes (4 November 2026, 18:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.