As companies accelerate the deployment of artificial intelligence tools across their workforces, employees are increasingly divided between those integrating AI into nearly every task and those who remain reluctant to use the technology. For business leaders, the challenge is no longer introducing AI into the workplace, but determining how to use it productively without surrendering human judgment.
Chris Williams, a former Microsoft vice president of human resources and leadership adviser with more than 50 years of business experience, says AI can deliver meaningful value when employees treat it as a tool for critical thinking, information management and routine administrative work rather than as a substitute for their own expertise.
The adviser, who has experimented with a broad range of AI systems and encourages executive clients to do the same, acknowledges that the technology remains controversial. He does not expect to persuade people who are fundamentally opposed to using AI. However, for employees and executives willing to experiment with it, he identifies three practical applications that can improve workplace efficiency and decision-making.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
His approach is notably different from the widespread expectation that AI’s greatest value lies in generating polished reports, presentations, images, or other creative material. Instead, he explains that its usefulness often emerges when it helps people evaluate existing ideas, extract essential information from lengthy explanations, and organize accumulated digital clutter.
AI As a Second Opinion, Not a Substitute for Judgment
The first recommendation is to use AI to challenge assumptions and identify weaknesses in a person’s reasoning.
Rather than asking a chatbot to produce an argument or make a decision, employees can present it with their own thinking and ask a straightforward question: “What am I missing?”
The adviser acknowledges that AI-generated creative work can be generic, mediocre, or embarrassing, frequently attracting criticism as “AI slop.” However, he notes that dismissing the technology because of its limitations in creative tasks overlooks its potential to provide an alternative perspective.
AI’s tendency to produce conventional or obvious observations can be useful when the objective is to identify considerations that may have escaped an individual. By asking a system to review an argument and explain where it might be wrong, users can uncover overlooked risks, competing interpretations, and weaknesses in their reasoning.
He also cautions that AI systems can be overly agreeable, reinforcing a user’s existing assumptions instead of challenging them. Explicitly requesting criticism can help address this tendency, although it does not guarantee an independent or accurate assessment.
Even when AI fails to uncover a genuinely overlooked issue, examining a counterargument can strengthen the original position. Employees can determine whether an objection exposes a real weakness or whether they have sound reasons for rejecting it.
This approach is more relevant in workplaces where colleagues increasingly use AI to prepare reports, proposals and arguments. Understanding the likely counterarguments to a position can help executives prepare for meetings, negotiations and internal discussions.
Williams’ focal principle is that AI should support the evaluation of an argument rather than make the argument on someone’s behalf. Its output remains a starting point for human analysis, not evidence that a position is correct.
Summarizing Information Without Losing The Underlying Detail
Williams’ second recommendation is to use AI to make dense information easier to understand.
As employees adopt generative AI, the volume of written material circulating through organizations can become overwhelming. A simple question may generate a lengthy response filled with explanations, qualifications, and tangential observations that obscure the main point.
The adviser compares this problem to dealing with colleagues who provide substantially more information than a situation requires. AI can reproduce the same tendency, producing pages of text when a few sentences would suffice.
One response is to ask the author to explain the central message more concisely. Williams says he sometimes requests a version without the “AI slop” to strip away unnecessary wording and expose the essential argument.
Another option is to use an AI system to summarize the original material, regardless of whether it was written by a colleague or generated by another AI tool.
He stresses, however, that summaries should not be treated as complete substitutes for the underlying information. A summary can help readers identify the main arguments and decide which sections of the original document deserve closer examination, but it may omit qualifications, contextual details or important exceptions.
This distinction matters in executive decision-making, where a concise explanation can improve efficiency but an omitted detail may materially change the interpretation of a proposal, financial report or policy document.
The adviser acknowledges the apparent circularity of using AI to summarize material that may itself have been produced by AI. Yet he considers the arrangement useful because the objective is not to eliminate every stage of processing, but to reduce the effort required to locate the information that matters.
The broader implication is that AI can serve as an information-filtering tool in organizations struggling with excessive documentation and communication. Its effectiveness, however, depends on users checking the summary against the source when accuracy and completeness are important.
Using AI to Organize Work While Retaining Control
The third recommendation concerns routine administrative tasks, including sorting, filtering and organizing files, emails and other digital records. These activities are well suited to AI systems because they involve identifying patterns, grouping information and helping users distinguish useful material from unnecessary accumulation.
Williams nevertheless draws a clear boundary between assistance and unsupervised automation. He is reluctant to allow AI systems to filter or file information without oversight and would not permit them to send emails, text messages, or documents without careful human review.
His caution reflects a broader concern about delegating actions that could have professional, financial, or reputational consequences. A system may correctly organize information in many cases but still misclassify a document, misunderstand its significance, or take an action that its user did not intend.
He illustrates the potential benefits with an experience involving a move to a new computer system. Before transferring his data, he asked AI to review what he planned to retain. The exercise exposed decades of accumulated digital clutter, including gigabytes of programmes, files and settings. After reviewing the inventory, he discarded approximately 90% of the material he had initially considered moving.
The example demonstrates how AI can help people make decisions about information they have accumulated over time but rarely examine systematically.
Employees facing thousands of emails or sprawling document folders could apply the same approach. Rather than instructing AI to clean everything up automatically, they can ask it to produce an inventory, identify duplicates, flag obsolete material, and recommend what might be retained or deleted.
The user then reviews those recommendations before taking action.
AI’s ability to identify patterns does not mean it understands the full organizational or personal significance of every file. A document that appears outdated may still be needed for legal, compliance, historical, or operational reasons. Human review remains necessary, particularly when deletion is irreversible or sensitive information is involved.
The adviser’s conclusion is that AI may be more useful for eliminating unnecessary material than for creating new content. In an environment increasingly saturated with AI-generated text and digital records, reducing clutter can be as valuable as producing more information.
Taken together, the three recommendations offer a measured approach to workplace AI adoption. Rather than treating the technology as an autonomous decision-maker, employees can use it to challenge their thinking, extract key information, and manage routine organizational tasks.
The approach also highlights a distinction that is becoming increasingly important as companies expand their use of AI: generating more output is not necessarily the same as improving productivity. Additional reports, emails and presentations can increase the volume of work without improving the quality of decisions or business outcomes.
AI’s practical value therefore depends partly on whether it helps employees reduce unnecessary effort, identify important information and make better-informed choices. That requires clear instructions, appropriate access controls and human oversight proportionate to the consequences of an error.
For executives, the implication is that successful AI adoption does not have to begin with complex automation or wholesale changes to business processes. It can start with relatively simple tasks that help employees work more critically and efficiently.
The adviser encourages employees to experiment with AI while remaining careful about how they use it. His message is not that the technology should be trusted indiscriminately, but that its limitations do not erase its usefulness.
Used appropriately, AI can provide another perspective, cut through excessive information and help organise work. The responsibility for judging its output, verifying important details and deciding what action to take remains with the human user.