Designing the Application of AI in Organizations

By Hubert Saint-Onge

AI is proving to be a transformative force, reconfiguring the very fabric of our work, our companies, and our human potential. As this technology continues to evolve at breakneck speed, it holds the potential for a fundamental shift in the capability of organizations. AI has already proven to be a new factor of production, like steam or electricity. It is being used to help cure diseases, design previously unimagined machines, teach children and adults, and avoid weather disasters.

AI changes the rules of the game for everyone. The rapid emergence of AI represents a fundamental shift in the capability of organizations to increase both productivity and efficiency. The contributions of AI will ultimately result in a significant competitive advantage for enterprises and have a profound impact on their customers and society.

However, we have yet to determine how to integrate AI at the intersection of this technology and how people work to enhance their contributions.

While the great majority of people have by now experienced or used AI in their lives, this article focuses on its impact in organizations where its integration has been more challenging.

Early signs of difficulty applying AI in organizations

It has been widely assumed that AI will yield productivity and efficiency gains for businesses, with the ripple effects of increased productivity and efficiency across the economy. But signs are emerging that it has proven challenging to harvest meaningful results with early initiatives to leverage AI in enterprises.  After companies have spent enormous sums of money on early applications of AI,  MIT has recently reported that some 95 percent of AI investments yield no net return. This finding is consistent with other recent studies on agentic AI, which have shown that IBM has found that only one in four AI projects demonstrates a net benefit, and just 16 percent merit full deployment across the enterprise.

These findings are emerging at a time when billions are being spent building data centers, most of it fuelled by debt.  

How can we best explain this setback?

In some ways, AI follows in the footsteps of other technology-driven transformations, such as railroads, electrification, and the internet, when we initially lacked the means to harness their formidable power.  When these three technology-driven waves successively emerged over a couple of centuries, there was initial excitement and a rush to build the infrastructure to support them; however, the initial inability to leverage the potential value they offered led to a vast economic bubble in each case.

As economist William H. Janeway has pointed out, in each of these three cases, we had to discover how to effectively tap into the potential of these powerful new developments. As Janeway puts it: “… for the AI bubble to deliver long-term, economically stable, financially rewarding results without disruptive discontinuity would be unprecedented in the history of capitalism.” It is somewhat ironic that Sam Altman, the CEO of OpenAI, who has invested hundreds of billions, used the word “bubble” in his recent public pronouncements. Meta, which has recently been offering $100 million signing bonuses for AI talent, is now reportedly considering downsizing its AI division. Neither of these companies has articulated a path to profitability.  They will unlikely be able to do so until organizations have found ways to increase their productivity, value creation and revenue with AI.

The equivalent of railroad tracks with the advent of AI is the rush to build data centers, complete with power plants, distribution grids, and fibre optic cables.  Given the nature of AI, we also face the challenge of dealing with a technology that is continually evolving and increasing its power at a rapid pace.  Another significant challenge is to deal with the nebulous nature of the competitive advantage provided by AI.  Because it is available to all and not contained behind a competitive wall, if it is available to one organization, it is also available to others. The one advantage an organization can derive comes from how it applies AI in its operations.

The MIT study, which reports a high percentage of AI project failures and abandonment, raises questions about the future of AI. The study has heightened the level of uncertainty about the viability of significant infrastructure investments if organizations are unable to leverage AI to improve their performance.  The only way to correct this situation is to show actual returns, which few companies have.  The reported failure may not be solely attributed to the inadequacy of AI technology and models.  As MIT researcher Aditya Challapally notes, failures were "less about the quality of the underlying models and more about how organizations attempt to use them".

Learning to apply AI in organizations

We are currently in this ‘early discovery’ period, where we are seeking the most effective way to harness the power of AI for organizations. Despite all the fanfare about AI, its reliability is a mortal weakness, particularly in corporate settings. Based on the collection of vast amounts of data from the internet and other sources, AI uses pattern recognition to develop algorithms. The resulting ability to marshal massive amounts of data can help individuals be more accurate, faster, and have a broader understanding of issues and opportunities. However, it has no built-in governor to discern the truth.

Beyond not knowing the truth, AI has no conscience: it has neither morals nor ethics. We tend to anthropomorphize AI when we experience AI’s sometimes astonishing leaps of reasoning. We must keep in mind that AI is a machine driven by the computational logic of 0s and 1s.  Technologists claim that guardrails have been embedded into AI to prevent it from veering into what has been labelled as not only factual but also moral and ethical hallucinations. Technical guardrails will never be sufficient to control AI’s propensity to err onto questionable ground. Leadership wisdom and a strong commitment to organizational values represent the best defense available.

To harness the power of AI, human beings need to complement the technology's capabilities with knowledge, judgment, and ethics. This powerful technology must be implemented based on a socio-technical framework, where humans and technology contribute what each is best suited to provide, thereby adding optimal value.  Accordingly, the technology must be shaped to allow active, ongoing collaboration with human actors. On the human side of this equation, organizations must shape the intersection of humans and technology to learn and adapt to these powerful new tools.  Organizational learning will play a key role in the success of the socio-technical approach.  

If human validation is essential, we must reconsider the “automate everything” approach and look to AI to augment human work, rather than aiming to replace it. In addition to having experience-based expertise, human decision-makers operate within a larger context when making decisions or choices. AI systems have access to a vast amount of information, but they lack access to the broader context that experience brings to humans. The organizations that will thrive with AI must experiment thoughtfully, adapt quickly and scale what works. For instance, this may involve creating feedback loops to help AI adapt to organizational workflows.  The term “Hybrid Intelligence”, being adopted in some circles, refers to the optimal combination of human and machine contributions. We have yet to demonstrate how this can be achieved most effectively.

Managing AI in the socio-technical space.

If we aim for collaboration between humans and AI, we need to design the interactions between the two with complementarity in mind. AI strengths include quickly gathering a massive amount of relevant information and the ability to draw inferences, find connections and identify steps to taking action. Humans bring experience and a broader perspective on what needs to be done, as well as critical thinking, creativity, moral reasoning, and an understanding of the big picture.  Through this collaboration, AI will learn from the choices made and refine its reasoning. This applies when we use AI not to replace but to augment people.

A case study on enhancing collaboration in teams with AI.

Procter & Gamble involved 776 employees from their Product Development group to determine the impact of AI on team performance. This was a randomized controlled trial in which half the teams received the support of AI, while the other half did not. The teams with AI incorporated were given what they called a “cybernetic teammate.”  Participants worked on real product innovation challenges.

The findings showed that AI contributed to enhancing performance and connectivity. It also showed that AI has the potential to break down silos by taking a broader view of the questions discussed. AI has been shown to enhance teamwork. Unlike other digital technologies, AI can come across as more human-like than a machine. It was found that AI helped team members dig deeper and enabled them to broaden their perspectives in dealing with obstacles or challenges outside their core expertise.

The teams with AI collaboration produced better solutions that integrated technical and commercial perspectives. AI-enabled teams performed best overall, achieving their tasks 12 to 26% faster and creating more detailed outputs. When using AI, less experienced employees performed at levels comparable to their experienced teammates. And were able to bridge knowledge gaps. AI users reported higher levels of excitement, energy and enthusiasm.  AI became viewed as a cybernetic teammate that contributed to enhancing collaboration rather than merely being a productivity tool.   The researchers found that AI can augment human capabilities and flatten traditional hierarchies.   The study concluded that AI can reshape teamwork by enhancing performance, breaking down silos and improving emotional experiences. (Reference: Dell’Acqua, et al. (2025) The Cybernetic Teammate: A field Experiment on Generative AI Reshaping Teamwork and Expertise, Harvard Business School, Working Paper, No 25-043).  

Conclusion

The integration of AI in organizations has proven to be a challenging task.  Many companies thought that they could integrate AI in the same way as other software systems. However, when AI is introduced into an organization, it’s not simply a matter of adopting a new tool. The adoption of AI involves redesigning the way companies work. The key is to redefine how human and machine intelligence interact with each other. Even though AI has been widely available since 2022, we are still in the early, emerging phase of this technology.  It remains challenging to determine the impact of AI on organizations precisely.

AI is now helping people feel competent, supported, and connected - roles that we once looked to managers, mentors, and coworkers to fill. That doesn’t mean we’re heading for a cold, robotic future. But it does mean we need to think more deeply about what work is when the office becomes optional and people get their directions from algorithms. If we’re not careful, AI won’t just shape how we work; it will reshape who we rely on, how we grow, and what we believe constitutes good work.

As we increasingly integrate AI into organizational operations, it is essential to remember that AI will acquire significant influence without inherent morals or ethics. Embedding such a powerful tool into an organization could lead to disastrous decisions if we don’t instill a strong sense of ethics within organizations as a protection against potentially harmful deviations in the future.   

We are entering an exciting but potentially perilous period. If we’re not careful, we will face an extended and messy adjustment period in organizations. To forestall these risks, organizations will need to:

• Take a more systematic approach to the change required to make AI work effectively for them.

• address concerns about privacy, data security, and the potential for bias in AI algorithms.

• discuss the importance of articulating and reinforcing strong organizational values.

• have leaders with the wisdom to detect instances where advice is provided that is contrary to the values of the organization.

• develop clear moral and ethical guidelines to ensure AI is used responsibly and fairly.