{"id":1491584,"date":"2026-06-08T13:08:46","date_gmt":"2026-06-08T11:08:46","guid":{"rendered":"https:\/\/www.ie.edu\/insights\/?post_type=articles&p=1491584"},"modified":"2026-06-08T13:08:46","modified_gmt":"2026-06-08T11:08:46","slug":"your-team-has-ai-tools-but-is-it-ai-ready","status":"publish","type":"articles","link":"https:\/\/www.ie.edu\/insights\/articles\/your-team-has-ai-tools-but-is-it-ai-ready\/","title":{"rendered":"Your Team Has AI Tools. But Is It AI-Ready?"},"featured_media":1491737,"template":"","meta":{"_has_post_settings":{"highlight_sharing":"default","image_sharing":"default","headline_sharing":"default"}},"schools":[],"areas":[508,481],"subjects":[422,420],"class_list":["post-1491584","articles","type-articles","status-publish","has-post-thumbnail","hentry","areas-artificial-intelligence","areas-leadership","subjects-innovation-and-technology","subjects-managing-people"],"custom-fields":{"wpcf-article-leadin":["AI access is not AI readiness. Ted Yang examines what serious leaders do differently in AI adoption."],"wpcf-article-body":["Most executives are confusing AI access with AI readiness. They buy licenses, launch pilots, run a training session, and then wait for productivity to follow. The assumption is that capable people plus powerful tools equals transformation. But without careful design, even the best people will produce only average results.\r\n\r\nConsider what is happening at Microsoft. The company recently ended most internal licenses for Claude Code<\/a> after distributing them widely just six months ago to much fanfare. Their teams already had access to GitHub Co-Pilot, so offering Claude Code was all about increasing productivity with the best tool available. Part of the reversal may reflect end-of-quarter pressures or the inefficiency of overlapping internal tools. But still, if there had truly been value created, Microsoft would not have been so quick to undo it.\r\n\r\nThe lesson for every leadership team is the same: tools are not a strategy. You need an operating system for AI that builds in what your best people do, optimizes workflows around them, and creates mechanisms for improving both the capabilities of your people and the quality of what you produce. That is a leadership exercise, not a procurement one.\r\n\r\nUsers Systemically Overestimate AI\u2019s Impact<\/strong>\r\n\r\nA recent randomized trial by METR of experienced open-source developers found that participants expected AI to make them faster and afterward believed it had. The measured outcome was the opposite: AI slowed them down. The lesson is not that AI is ineffective. Instead, it is that even experienced technologists can systematically misjudge when and how AI helps<\/a>. If your most technical people cannot reliably self-calibrate on AI effectiveness, your broader workforce certainly cannot, and they will adopt AI in ways that only feel productive while delivering average or worse results.\r\n\r\nThis has a direct implication for how leaders should think about their most experienced people. Research from Gilles Gignac of University of Western Australia and Marcin Zajenkowski of University of Warsaw on cognitive and personality traits suggests that overall functioning can peak in late midlife, even as raw speed declines<\/a>. Judgment, pattern recognition, contextual understanding, and relationship intelligence keep compounding. These are exactly the capabilities AI should amplify first, not replace, or else you risk losing quality. An organization that routes AI toward cost reduction rather than its best people is optimizing for the short-term outcome.\r\n\r\nThe \"Average Work\" Problem Is Structural, Not Incidental<\/strong>\r\n\r\nTo understand why most AI implementations underdeliver, it\u2019s important to understand what large language models actually produce. Due to their probabilistic nature, LLMs are primarily equipped to generate work at the level of an average person or a lower-level expert, not at the level of the company\u2019s best talent. Creativity research from Antoine Bellemare-Pepin of Bard College and Fran\u00e7ois Lespinasse of Concordia University confirms this: large language models can exceed average human performance on certain measures, while still trailing top performers<\/a>.\r\n\r\nThe broader productivity data supports this. An NBER 2026 survey of about 6,000 CEOs found that despite widespread AI adoption, promised productivity gains are not showing up in measurable outcomes<\/a>. Executives expected AI to raise productivity by only about 1.4% over the next three years. As economist Torsten Slok notes, \u201cAI is everywhere except in the incoming macroeconomic data<\/a>.\u201d\r\n\r\nThis is not a temporary limitation that the next model release will solve. It is structural. And it has knock-on effects that compound through an organization. If your strategy documents are average, your results can never be better than average, even if your execution is exceptional. Think of a marketing campaign with a flawed strategy from the outset. Even clever visuals, strong copy, and polished execution cannot compensate for targeting the wrong objective. Average in, average out, regardless of effort downstream.\r\n\r\nThe result is what I call artificial productivity: more artifacts that indicate output while at the same time fewer results that create value. Organizations mistake volume for progress but the actual outcomes that matter: revenue, quality, and client satisfaction either stay flat or decline.\r\n\r\nAvoiding this requires a deliberate answer to one question at every stage of implementation: where in this workflow does the quality ceiling matter, and are we protecting it?\r\n\r\nWhat Happened When One Company Got This Wrong<\/strong>\r\n\r\nThe failure rates in enterprise AI are not abstract. According to RAND Corporation, over 80% of AI projects fail to deliver measurable results<\/a>. A separate S&P Global survey found that 42% of companies abandoned most of their AI initiatives in 2025<\/a>, up from just 17% the prior year.\r\n\r\nA learning software company I am familiar with illustrates exactly how this plays out. They initially avoided AI entirely. The nature of their work made its use feel like cheating. But as adjacent competitors adopted AI tools to update and produce materials faster, the pressure became unavoidable. Leadership made a decision: they purchased multiple licenses of powerful AI tools and distributed them to their teams with limited instruction. Executives viewed AI as technology that, like other technologies before it, their teams would be able to figure out. They expected their employees to fill in the blanks.\r\n\r\nThe results were highly uneven. Teams that had sufficient internal expertise and momentum saw real gains: content quality improved, output volume increased, and materials were updated faster. But other teams suffered. Inertia, lack of understanding, or outright counterproductive use led to worse outcomes. In one recurring pattern, teams produced content that was later rejected in quality control, resulting in less net output than before AI was introduced.\r\n\r\nEventually, the company brought in outside expertise to diagnose the problem. But the deeper cost was not the lost productivity. It was the damage to morale among high performers who had begun to fear for their jobs. That fear was entirely avoidable. It emerged directly from a rollout that prioritized tool distribution over implementation design.\r\n\r\nThe analogy is from 2020. Companies gave Zoom to employees and expected them to figure out remote work. Some did. Most did not, until someone redesigned workflows for virtual work.\r\n\r\nWorkflow Redesign Is the Actual Lever<\/strong>\r\n\r\nAchieving above-average results requires more than adding new tools to an existing process. It means redesigning workflows deliberately. Many organizational workflows are inherited and were never explicitly designed, or at least not designed recently, for their current purpose. In many cases, even the people executing them no longer fully understand why they operate the way they do. This is the same for high performance itself. Organizations may not know exactly how their best people consistently produce exceptional results, because much of that is tacit, relational, and part of a day-to-day practice rather than documented. All of this must be surfaced before AI is introduced into the system. If not, AI enhancements will simply enhance the wrong things, and it will scale limitations rather than strengths.\r\n\r\nThree questions should drive this work.\r\n\r\nFirst, where does human judgment need to remain? Not everywhere, but always at the highest-stakes decision points: the critical intersection between competing choices, or in direct engagement with stakeholders where the relationship itself is the asset. This is where experienced people add value that no LLM can replicate. Protect it explicitly.\r\n\r\nSecond, what work can be eliminated entirely? Look for steps that move data around, bridge communication without adding analysis, or have accumulated over time without ever being deliberately designed. Find the make-work. Find the artificial steps that do not lead to better results. Removing them creates space for quality work to get the attention it deserves.\r\n\r\nThird, where are the human gates in the new workflow? Because AI increases speed and volume, it\u2019s important to have explicit checkpoints so that an internal decision does not quietly become a high-stakes mistake without sign-off. Speed without verification means you reach the wrong answer faster. This must be built into the system, not left to individual judgment in the moment.\r\n\r\nThe Cognitive Overload Risk<\/strong>\r\n\r\nThere is a failure mode that looks like success for longer than it should. Being able to do more with AI, especially simultaneously and in parallel, creates a different level of management overhead. Without systems for quality control, human managers become overwhelmed. The bottleneck shifts from production to review, and if leaders are not watching for it, the review step degrades quietly under volume pressure.\r\n\r\nThe research on this is now specific and sobering. A March 2026 Boston Consulting Group study of 1,488 full-time workers found that 14% of AI users experience what researchers now call AI brain fry<\/a>: cognitive overload caused not by using AI, but by overseeing it. They also experienced 12% more mental fatigue, and major error rates increased 39%. Productivity peaked when using three simultaneous AI tools and declined measurably beyond that. Perhaps worst of all, among workers experiencing brain fry, 34% showed active intention to leave their company.\r\n\r\nSeparate research at UC Berkeley studied a 200-person tech firm over eight months and found that AI accelerated individual tasks but raised organizational expectations for speed, creating what researchers termed workload creep. The time AI saved was immediately refilled with more work<\/a>, not reclaimed for deeper thinking.\r\n\r\nBefore AI, the limiting factor was raw output. The new limiting factor is quality output. Ten times more average work only creates ten times more downstream effort to sift through to find the result that actually matters. Humans cannot spot-check everything, especially when the context is complicated. Quality control mechanisms need to be built into the workflow alongside AI, not treated as a separate step that can be skipped when things get busy.\r\n\r\nLeaders need training to understand the difference between more and better. That distinction is not intuitive when dashboards are showing higher numbers across the board.\r\n\r\nWhat Serious Leaders Actually Do<\/strong>\r\n\r\nChange management is necessary. The biggest obstacle is always inertia, compounded by the fact that AI remains widely misunderstood due to hype on all sides. A 2025 survey of 2,500 global employees and IT leaders found that 86% admit they are not using AI tools to their full potential<\/a>. Of those, 82% say they are not familiar with how AI can be applied practically to their day-to-day work.\r\n\r\nIf people sense that the goal is to speed up broken processes, or that the real agenda is headcount reduction, adoption will quietly fail and morale will erode alongside it.\r\n\r\nTherefore, leaders must establish up front that the goal of AI implementation is not to cut jobs, but to empower people to do their best work. This, of course, requires an understanding of how teams and high performers actually work before redesigning anything around them. Getting this right, and avoiding average AI outcomes, requires six specific actions:\r\n
\r\n\r\nMap before you deploy.<\/em> Understand current workflow before trying to AI-power it. Identify where your best people spend their time today and how they produce their strongest results. You cannot amplify what you do not understand, and you cannot avoid average results if you are building on a foundation you have never examined.\r\n\r\nTreat this as change management.<\/em> Assign clear ownership and responsibility. Set explicit expectations about the investment required and the timeline to outcomes. Understand the role of every team member in the transformation. Unmanaged adoption produces uneven results.\r\n\r\nAim for amplification, not replacement.<\/em> Design AI workflows around your best performers and their actual work patterns. Your top people are the ones most likely to stay above the AI quality floor. Make them better, and you widen the gap between your organization and a competitor that is simply producing more average work at a faster rate.\r\n\r\nVerify your aim.<\/em> Critical outputs need an explicit human review step, and it must be ongoing, not a one-time check. The developer research is instructive: people believed AI helped when it measurably had not. Without deliberate verification, your organization will have the same blind spots.\r\n\r\nTrack quality, not just volume.<\/em> If your team is producing three times as many deliverables at half the quality, you have a problem that looks like success. Build the metrics that surface the difference before it compounds into client or market consequences.\r\n\r\nSupport honest feedback.<\/em> AI tends to breed hyperbole internally as well as in how results are reported upward. Without solid feedback systems in place, organizations begin optimizing around perception rather than measured outcomes. The companies that avoid average results are the ones willing to test their assumptions and receive internal feedback, even when the answer is inconvenient.\r\n\r\n<\/div>\r\nThe Real Question<\/strong>\r\n\r\nA strong organization is ready for AI when it is realistic about what the technology can achieve. AI is a tool, and widespread implementation of any tool requires much more than purchasing it.\r\n\r\nHaving a plethora of AI licenses is not the recipe for transformation. If it were, we would not be watching sophisticated organizations pull back and reassess their AI strategy. The competitive advantage in this era will not go to the organizations with the most tools. It will go to the organizations whose leaders understood that AI changes what work looks like, and who have the discipline to redesign around that reality before their competitors do.\r\n\r\nThe real question is not whether your company has AI tools. It is whether your organization has the leadership to use them exceptionally well.\r\n\r\n\u00a9 IE Insights."],"wpcf-article-extract-enable":["1"],"wpcf-article-extract":["AI access is not AI readiness. Ted Yang examines what serious leaders do differently in AI adoption."],"wpcf-audio-article":["https:\/\/www.ie.edu\/insights\/wp-content\/uploads\/2026\/06\/Yang-Audio.mp3"]},"_links":{"self":[{"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/articles\/1491584","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/articles"}],"about":[{"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/types\/articles"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/media\/1491737"}],"wp:attachment":[{"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/media?parent=1491584"}],"wp:term":[{"taxonomy":"schools","embeddable":true,"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/schools?post=1491584"},{"taxonomy":"areas","embeddable":true,"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/areas?post=1491584"},{"taxonomy":"subjects","embeddable":true,"href":"https:\/\/www.ie.edu\/insights\/wp-json\/wp\/v2\/subjects?post=1491584"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}