The direct, personal involvement of a company\u2019s CEO is essential if any significant transformation is to succeed.\u00a0 Yet for AI, most have stayed on the sidelines, unclear what exactly they must do, or assuming AI belongs to IT.\u00a0This will not stand. This article proposes five non-delegatable tasks belonging on their To-Do lists.\u00a0These include:<\/p>\n
All things AI press on at an almost mind-boggling pace. Tech companies \u2013 and the Trump administration, Europe, China, and many countries \u2013 are making enormous infrastructure and other investments.<\/p>\n
Stories featuring game-changing potential (e.g., AI co-scientists or co-workers) appear daily; AI providers announce breakthroughs with increasing frequency; companies (e.g., Duolingo or Shopify) announce stunning new AI initiatives; and consultancies project multi-trillion-dollar benefits in just a few short years.<\/p>\n
The story would not be complete without geopolitical implications, as concerns about Nvidia chips, Taiwan Semiconductor factories, the availability of rare earth minerals, and progress in China (e.g., DeepSeek) illustrate. It is hard not to get caught up on the excitement!<\/p>\n
In stark contrast, AI is yet to deliver on its promises: Too many companies don\u2019t know how to assess implications or where to start, many others can\u2019t get a simple pilot into production, data continues to be a second-class citizen, and resistance comes from all quarters. On top of this, many companies mistakenly believe they already have an AI strategy simply because they have activated tools like Microsoft Copilot.<\/p>\n
When they fail to see the productivity gains they were expecting, they conclude that AI has little to offer them. There is a lot to learn and do in a very short time. After all, a technology can run on hype for only so long.<\/p>\n
When it is nearly impossible to separate the signal from the noise, what should a company do?<\/p>\n
We\u2019ve contributed to, led, and advised on plenty of AI and data efforts \u2014 some proved transformational, some moderately successful, and some outright failures. In our view, progress has been shackled by a lack of both understanding and imagination, paralysis in the face of uncertainty, a failure of courage, bad data (both structured and unstructured), half-heated measures, a \u201cwe build it and the come\u201d misconception, and organization charts best described as \u201cunfit for data, people and AI.\u201d It is time to take the shackles off!<\/p>\n
We\u2019ve also learned that:<\/p>\n
Combining these perspectives leads us to\u00a0five non-delegatable responsibilities for CEOs\u00a0<\/b>hoping to garner their company\u2019s share of the benefits cited above:<\/b><\/p>\n Let\u2019s explore each in turn.<\/p>\n Considering the constant flow of AI \u201cnews,\u201d we understand why CEOs might prefer to make smaller bets on easier projects. That can be fine for a first step, but don\u2019t limit your ambition of where you want to eventually reach. Consider this sequence of increasingly bold possibilities.<\/p>\n What if?<\/i><\/b>\u00a0Indeed, AI enables organizations to rethink entire business models.\u00a0The best current example is the impact AI is having on warfare. As\u00a0pointed out by the CNAS<\/a>, \u201cAI is emerging as a significant asset in the ongoing Russian-Ukrainian conflict\u201d and \u201ctranslates into more precise and capable responses to adversary forces, movements, and actions.\u201d<\/p>\n Another example involves the\u00a0Singapore National AI Strategy<\/a>\u00a0that focuses on \u201cfundamentally rethinking business models and making deep changes to reap productivity gains and create new areas of growth.\u201d\u00a0Every company and government agency should ask itself extensive \u201cwhat-ifs\u201d involving AI and competitors, new entries, long-standing business problems, and relationships with customers.<\/p>\n Of course, at any point in time, CEOs must decide what projects to actually fund, or to fund first, in the face of Board pressure, competition among subordinates, and employee concerns. Don\u2019t let politics hamper your ambitions.<\/p>\n The cold, brutal reality is that everything about AI is uncertain: Estimated benefits and feasibility vary greatly, most \u201cprojects\u201d so far have either failed entirely or failed to get out of pilot stage, trust in AI is often low, with many fearing it, and national governments are unsure what and how to regulate.<\/p>\n And who knows what impact the new U.S.-China \u201cspace race\u201d will have? It has even gotten to the point that insurance companies such as\u00a0Munich Re offer insurance products for new AI ventures<\/a>.<\/p>\n Not a comfortable environment for managers who\u2019ve spent their entire careers trying to reduce uncertainty. Making matters worse, we don\u2019t see any relief in sight. Our best historical analogy involves the\u00a0printing press<\/a>, where it took two full generations for some semblance of order to emerge.<\/p>\n Senior leaders, especially CEOs, are well-advised to embrace uncertainty, though \u201clearn to live with it\u201d is probably more apt. Indeed, increasing uncertainty, especially rapid-fire \u201cunknown-unknowns,\u201d creates winners and losers. To do so, CEOs and companies must learn to experiment, in everything from technologies to funding mechanisms, to people and organizational structures.<\/p>\n All experiments take time and resources, so it is best to plan them carefully. Some experiments, such as those that involve moving from model to adoption will be large. \u00a0 And plenty fail. Importantly, a part of experimentation involves reaping the benefits of successful experiments and learning from failed ones.<\/p>\n Said differently, rather than investing in a long-term strategy \u201cget out there,\u201d sort out your company\u2019s capabilities, learn, grow, and adapt.<\/p>\n One final point:<\/i><\/b>\u00a0Adopting a \u201cwait and see\u201d attitude qualifies as an experiment. An even worse one involves \u201cdabbling,\u201d perhaps motivated by FOMO-fear of missing out. Such halfway measures are not big enough to succeed, but may well convince you, \u201cAI is not for us,\u201d or \u201cthe technology is not ready,\u201d and lead to missing out.<\/p>\n As James Betker of OpenAI observes, \u201cThe \u2018it\u2019 in AI models is the dataset<\/a>.\u201d And today, data availability and quality are generally regarded as the\u00a0number one limiting factor for AI<\/a>,\u00a0as highlighted by the recent failed launch of\u00a0France\u2019s AI chatbot, Lucie<\/a>.<\/p>\n Of course, very few companies will actually develop a foundation LLM such as DeepSeek, Chat, or Gemini, nor can they do much about data scraped from the Internet used to train LLMs. Instead, companies must focus on the data they use to augment LLMs and to train and operate predictive AI models.<\/p>\n The issues are broad and deep:\u00a0Much data is simply wrong<\/a>, poorly defined, or even hard to find; a host of \u201cright data\u201d issues,\u00a0such as relevancy and bias<\/a>, bedevil individual projects; and models themselves provide bad answers (inferences), even hallucinate.<\/p>\n Exacerbating this, most data is \u201cunstructured\u201d (i.e., emails, contracts, forms, recordings of meetings and so forth) and subject to less scrutiny than data in corporate systems. \u00a0 Finally, many individuals have been well aware of the problems for some time (e.g., Garbage in, garbage out) and recognize that their companies\u2019 data is not ready. But companies have misdiagnosed data quality as a technical problem, not the management problem it really is.\u00a0No surprise, data has not received the attention it deserves!<\/p>\n AI brings data and data quality to the fore<\/a>.\u00a0We strongly suspect that data quality will go a long way in separating winners and losers.<\/p>\n The implications for CEOs are clear:<\/i><\/b>\u00a0They must launch and fully support aggressive data quality programs, likely akin to those of the quality revolution in manufacturing, and\u00a0featuring end-to-end focus on processes that cross multiple departments to supply AI applications<\/a>. Fortunately, companies such as Aera Energy, AT&T, Chevron, Gulf Bank, Hello\u00a0Fresh, and Shell have\u00a0shown that\u00a0data quality improvement is a big winner<\/a>!<\/p>\n As Arthur Jones observes, \u201corganizations are\u00a0perfectly designed to<\/a>\u00a0achieve the results they achieve.\u201d AI presents some tough challenges, such as sorting the various roles for IT and \u201cthe business.\u201d But such issues are just the tip of the organizational iceberg.<\/p>\n The litany of issues includes too many people not understanding their responsibilities, issues that fall \u201cin the white space,\u201d too much \u201cup and down\u201d management, and not enough \u201cleft-to-right,\u201d and skepticism of analytics generally, and AI specifically.<\/p>\n In some respects, such issues should surprise no one. After all, today\u2019s organizations were\u00a0designed for industrialization, not data<\/a>\u00a0and AI.<\/p>\n Here is where the CEO\u2019s role is especially crucial:<\/i><\/b>\u00a0To meet these challenges, CEOs must actively take ownership of reshaping the organization so it is ready for a world where data and AI are central. This means driving a transformation that promotes the right organizational qualities: clear accountability, cross-functional collaboration, adaptability, and a commitment to continuous learning.<\/p>\n The CEO must ensure that the organization becomes capable of systematically addressing data and AI challenges, breaking down (or bridging across) silos, encouraging experimentation, and scaling successful innovations. Importantly, they must foster a culture where learning from both success and failure is valued, where teams are empowered to test and improve, and where the organization as a whole becomes more agile, resilient, and ready to evolve.<\/p>\n In short, the CEO\u2019s mandate is to redesign the organization not by replicating a particular structure or best practice, but by building the capacities and virtues needed to embed data and AI into the heart of the company\u2019s operational and strategic work.<\/p>\n This organizational work is going to require a lot of effort, considerable experimentation, and some painful choices.\u00a0 But there is no getting around it.<\/p>\n We\u2019ve pointed out that the stakes are high, though uncertain; that there is a lot of new, unfamiliar, and likely unpleasant work to do; competing priorities; and plenty of excuses to \u201cwait.\u201d\u00a0For example, the tech community proudly proclaims, \u201cToday\u2019s AI is the worst AI you will ever use.\u201d\u00a0Why not wait until it is ready for prime time?<\/p>\n\n
1. Don\u2019t limit your ambitions<\/b><\/h5>\n
\n
2. The uncertainty is enormous \u2014 Learn how to live with it<\/b><\/h5>\n
3. AI is data: You must make it a first-class citizen<\/b><\/h5>\n
4. \u201cBuild AI into\u201d your organization rather than \u201cbolting it on\u201d<\/b><\/h5>\n
5. Act boldly and soon, but thoughtfully \u2013 The race to create value using AI will be a long one!<\/b><\/h5>\n