Why Traditional Transformation Models Are Not Working For AI
Dr. Brian Massey
Managing Partner at Anordea | Strategy Advisor, Professor & Keynote Speaker | Helping CXOs Drive Faster AI Transformations across Banking and Financial Services.
Hello and welcome to the latest edition of AI Strategy Brief. In this edition we’ll review some recent industry developments in AI, and look at why traditional transformation models are failing when it comes to AI.
AI Industry Updates:
Consulting – BCG On The Rise Of AI Advisory
The CEO of BCG, Christoph Schweizer, revealed that the consulting giant expects to generate a fifth of its 2024 revenues from AI integration services, a figure projected to rise to 40% by 2026 due to a surge in demand for artificial intelligence and generative AI. This growth is driven by companies transitioning from experimenting with AI technologies to deploying them at scale, involving collaborations with tech firms like Microsoft, Google, OpenAI, and Anthropic. Schweizer noted the unprecedented speed at which Gen AI has become a central business priority, with BCG leveraging AI for tasks such as minute-taking, email drafting, and document summarisation within its own operations. Despite a general slowdown affecting other parts of the business, BCG's focus on AI has helped maintain growth, with total revenues increasing by 5% in 2023 to $12.3 billion.
?
BCG's aggressive integration of AI into its consulting services signals a significant shift in how consulting firms will operate in the future. This strategy is likely to catalyse a wave of innovation across the consulting sector, pushing competitors to similarly embrace AI to remain competitive. The pervasive use of AI in consulting will drive advancements in AI technology itself, as demand for more sophisticated, industry-specific solutions increases. For clients across sectors, the strategic implications of these developments are worthy of focus. Clients stand to benefit from accelerated decision-making, enhanced operational efficiency, and more robust problem-solving capabilities, as they work with leading AI consultants. As firms like the MBB competitors demonstrate successful use cases of AI themselves, they provide a roadmap for clients to follow, reducing the perceived risks associated with AI adoption and encouraging more widespread implementation. This evolution could lead to significant shifts in employment patterns, competitive strategies, and even business models across industries, as companies leverage AI not just for automation, but also for more fundamentally transformative strategic initiatives.
Consumer Goods – Coca-Cola Partners with Microsoft
Microsoft has secured a $1.1 billion five-year contract with Coca-Cola to provide cloud computing and artificial intelligence services, expanding upon a previous $250 million agreement made in 2020. Under this new deal, Coca-Cola will utilize Microsoft's Azure OpenAI to develop AI-driven chatbots and other services within Microsoft's Azure cloud platform. Additionally, Coca-Cola plans to test Microsoft's AI assistant, Copilot, which aids in tasks like summarising email discussions and creating business presentations. The agreement also includes an expansion of Coca-Cola's use of Microsoft's traditional software offerings, such as Dynamics 365. However, Microsoft did not disclose how the financial aspects of the deal are divided between AI services and other cloud software.
?
The significant investment by Coca-Cola in Microsoft's AI and cloud services not only highlights the growing importance of AI in corporate strategy but also sets a benchmark for AI adoption across industries. This collaboration between two industry giants demonstrates a commitment to exploring and integrating advanced AI capabilities, such as Azure OpenAI and Copilot, into everyday business processes. It signals to other companies the potential gains in productivity and efficiency that AI can bring. As major players continue to invest heavily in AI, it will likely accelerate a trend where businesses across various sectors seek to harness similar technologies to stay competitive. This move could spur further innovation and development within the AI field, prompting AI service providers to enhance their offerings for more tailored, industry-specific solutions. As a result, we can expect AI deployment and sophistication to expand, influencing broader industry dynamics, driving technological advancements, and potentially reshaping competitive landscapes.
Regulation – AI Firms Opt Out Of UK Pretesting
Despite the UK Prime Minister Rishi Sunak's declaration of a "landmark" agreement at Bletchley Park to conduct pre-release safety testing on major AI models, the AI Safety Institute (AISI) has struggled to access and test the most advanced AI technologies from key industry players like OpenAI and Meta. Six months post-agreement, the AISI has only managed to test models post-release, with Google DeepMind being a partial exception. The difficulty highlights the limitations of voluntary commitments in the absence of binding legislation and the complexities of international cooperation on AI regulation. Jurisdictional issues and the fear of setting precedents for global regulatory requirements have made major AI firms hesitant to share sensitive technological details. This situation underscores the necessity for more formal agreements and possibly legal mandates to ensure the pre-release testing of AI technologies to mitigate potential threats and ensure public safety.
?
As AI companies gear up to launch advanced models like OpenAI’s GPT-5 and Meta’s Llama 3, the challenges encountered by the UK's AI Safety Institute (AISI) in securing pre-release testing access could prompt a re-evaluation of regulatory relationships between tech giants and governmental bodies. This situation may also accelerate the push for more robust, legally binding international frameworks for AI governance. Governments may see a need to impose stricter regulatory measures to ensure that AI systems are safe before deployment, particularly for AI applications that could pose significant security risks or ethical concerns. This could lead to heightened scrutiny and potentially stricter compliance requirements for AI developers, influencing how companies approach the development and rollout of AI technologies. Furthermore, the reluctance of AI firms to engage voluntarily with regulatory bodies might encourage a more unified global approach to AI safety standards, potentially streamlining the regulatory process but also imposing more rigorous accountability measures on these tech companies.
AI Transformation Blueprint
Today’s newsletter is sponsored by Anordea’s AI Transformation Blueprint. At Anordea, we advise CEOs and senior leaders across Financial Services, helping them tackle their toughest problems. With the rapid pace of development and change in the AI space, many leaders are faced with a paradox of how to build on and govern the technology that is as complex as it is powerful.
?
Our AI Transformation Blueprint solves this paradox quickly and effectively by giving leaders immediate insight into the state of AI across their organisation and industry, highlighting the opportunities and challenges which need to be addressed to maximise their AI opportunity in a safe and well-governed manner. Based on our leading expertise and ongoing research in the AI Strategy and Governance field, the AI Transformation Blueprint offers a fast and high impact solution to one of the biggest challenges facing FS leaders today.
?
Working with a dedicated member of our senior leadership team, the benefits of this process include ROIs in the strategy, governance, operational, financial and personal arenas. For those interested in exploring what our AI Transformation Blueprint involves, and how it can help your organisation gain a leading advantage over competitors, reach out to one of our senior partners here or contact us at [email protected].
Why Traditional Transformation Models Are Not Working For AI
Over the last two decades, several transformation approaches have been used extensively in digital transformations across various industries. These methodologies vary widely but often include elements such as strategy alignment, technological integration, and organisational change management. The major approaches are not mutually exclusive by any means and are often blended to fit the unique needs and contexts of specific organisations or industries. Some of them fit better with certain industry cultures as well, while others are more universal in deployment. For transformation leaders, the choice of approach often depends on various factors including the organisation’s culture, existing systems, business goals, and the specific challenges and opportunities of the digital landscape they operate within. However, when it comes to AI Transformations in particular, we are seeing striking problems across all these methodologies, as none of them appear to be wholly appropriate for AI Transformations.
?
The approaches mentioned below, while effective for general digital transformation, encounter specific challenges when deployed directly to AI Transformations, that cannot be easily solved. This is due to the unique nature of the AI technologies we see emerging, but also the manner in which they graft on to organisational structures, and the diversity of form they take across these organisational structures. Simply put, it appears traditional transformation methods are not up to the job. In this article, I want to review the major transformations approaches that have been deployed on digital transformations in the past and explain why they are proving inappropriate for the todays AI Transformation needs.
领英推荐
?
Agile Transformation
Agile is probably a phrase most people are familiar with by now, having come from its origins in project management and spreading through transformation and management thinking in a much broader sense. Agile transformations focus on rapid iterative development cycles to quickly deliver features that meet user needs. This encourages collaboration across functional teams with the aim of enhancing flexibility and responsiveness. Agile transformations typically include scrum methodologies, sprints, and frequent reassessment of project goals and achievements. This makes agile methods particularly useful in traditional transformation initiatives. However, when it comes to AI Transformations there are some issues that Agile approaches encounter. Agile focuses on rapid iterations, which can sometimes mean a trade-off with the thoroughness required in training and refining AI models or tools. The speed of Agile approaches also leaves little room for the essential data governance and ethical considerations that are crucial in AI deployments. The iterative nature may not adequately address the stability and reliability needed for AI applications when they are plugged into wider organisational systems, where changes can significantly affect outcomes and require extensive validation. Overall, while Agile is a good approach to small scale technology transformation, it appears unsuited to AI Transformations, at least on its own.
?
Lean Methodology
Another driving force in transformation methodologies over the past two decades has been the Lean approach. Derived from manufacturing knowledge, this approach focuses on value creation while minimising waste. This means that Lean thinking can be applied to a variety of organisational processes common not just the production line. In digital transformation, it emphasises understanding customer value, the flow of value creation, and continuous improvement. However, when it comes to AI Transformation, Lean's emphasis on reducing waste does not align well with the exploratory and experimental nature of AI, where initial phases often require substantial resource investment without immediate, tangible returns. AI projects can involve significant complexity and unpredictability, which is at odds with Lean’s focus on streamlining and efficiency. Lean thinking is essentially based on a philosophy of optimisation at the process level. However, AI technologies are not yet far enough along their development curve, to benefit from this sort of normative approach.
DevOps
If you have a background in software engineering or computer science, you would be familiar with the approaches taken that are colloquially known as DevOps. This approach to change and transformation blends software development (Dev) and information-technology operations (Ops) aiming for shorter development cycles, increased deployment frequency, and more dependable releases. Behind DevOps is a desire to align development and operations teams to improve code quality and operational reliability. When engaging in change in the software space DevOps is a fantastic approach to take and can build a strong culture across an innovation team. However, while excellent for operational integration between development and operations, DevOps does not sufficiently address the broader strategic and organisational changes than an AI Transformation demands. Factors such as cross-departmental collaboration beyond IT and operations, require a more strategic view of change if a transformation is to be successful. Additionally, DevOps practices may need to be extensively modified to incorporate AI-specific considerations like model versioning, data lineage, and continuous training cycles, not to mention the governance or organisational model aspect of the change.
?
Change Management
Moving out of the technological realm a little, management science has traditionally put forward strong change management methodologies focused on organisational and institutional change. These approaches are often seen as being more comprehensive when looking at transitions at the organisational level, from current states towards desired future states that maximize employee adoption and usage of new technologies. There is a variety of tools and models here such as Prosci’s ADKAR Model or Kotter’s 8-Step Change Model. However, traditional change management approaches are too rigid and slow to accommodate the adaptations needed in rapid technology deployments such as AI Transformations. This problem is exacerbated by the pace of advancement in the AI space and the speed at which disruptions are emerging and evolving. As a simple example look at the speed at which me moved from impressive chatbot based AIs, into multimodal forms. Change Management also suffers from an emphasis on top-down approaches that can stifle innovation and experimentation essential for AI Transformations, not to mention the cultural challenges presented by the new AI organisational contexts.
?
Design Thinking
A darling of the innovation coaching space over the last decade, Design Thinking has been the star behind a lot of digital transformation work across industries. And rightly so, given the novel thinking and refocusing on customers and users it has enshrined in its approach. It benefits strongly by?utilising the user-centred techniques to solve problems in a creative and iterative manner, involving stages like empathising with users, defining pain points, ideating solutions, prototyping, and testing. This was huge help in a clunky Web 2 world. However, AI Transformation and innovation requires quite a different approach that is somewhat at odds with the strengths of Design Thinking. For instance, where Design Thinking focuses on user interfaces and journeys, Generative AI provides a simple one-stop chat interface, that is quickly becoming highly adaptive and tailored to diverse user interaction needs. Also, Design Thinking’s focus on the user, does not help transformation leaders needing to address the backend complexities and the technical scalability challenges of AI systems, all while saying little on the governance, ethics, and risk mitigation aspects of AI Transformation.
Digital Adoption Platforms (DAPs)
While not strictly a transformation methodology, Digital Adoption Platforms are often used as a leading technique to improve user adoption in technology initiatives across organisations. There are times when the technology itself can become the methodology, and this is likely a result of the success that DAPs have had in aiding technology transformations over the last decade. DAPs utilise software layers integrated with existing systems to guide users through new processes and provide on-the-go training. They aim to accelerate user adoption and enhance proficiency with new digital tools, by speeding up the connection of user to information as needed. While these platforms can assist in technology adoption in general, they are highly prescriptive and do not offer much in the way of flexible support, needed when dealing with highly customised AI experiences. This is due to the highly customised nature of AI applications where user needs and system capabilities can vary greatly. AI requires continuous learning and adaptation, which DAPs may not support dynamically, as they often focus on static workflows and processes. In fact, DAPs are one of the likely victims of Generative AI success, as users turn more to programs like GPT-4, and even custom GPTs, to meet their guidance and tutoring needs.
?
Data-Driven Transformation (DDT)
Data Driven Transformations involve leveraging big data and more powerful analytics to make informed decisions that guide business strategies. In many ways, DDTs have laid the ground work for, and are the parent of, the AI Transformations we now see emerging. Much of the work in Data Driven Transformations focus on integrating data silos, enhancing data quality, and adopting advanced analytics capabilities. This experience is helpful for AI Transformation work, and indeed reconciling data silos, and integrating legacy systems that harbour crucial data with newer AI tools, is an integral part of AI Transformation. However, while DDT emphasises the use of data, and the need for quality data, it is not well equipped to meet the business disruption and governance needs of an AI Transformation at the organisational level. AI Transformations move the point of change from the data reservoir or decision support system, and simultaneously place it at multiple points across the organisational structure. Additionally, many of these points, which are candidates for automation, present AI use cases that are external facing in that they cross the boundary of the organisation. This is a vast departure from the work of DDT in the past, which has always kept the major points of change, and therefore risk, as internal facing use cases. Additionally, AI Transformations necessitate a nuanced understanding of data biases, data security, and the implications of machine learning decisions, which can be overlooked in a straightforward data-driven strategy. For these reasons AI Transformation is at best an evolution from DDT, if nothing something completely new itself.
Scalable Platform Models
Another major player in the digital transformations we have seen over the last two decades, is the trend towards Scalable Platform Models. These approaches involve creating flexible and scalable platforms that support a broad range of services and adapt to changing business needs. Think of the sorts of cloud migration strategies we have seen leading to the adoptions of SaaS, PaaS, or IaaS solutions. Even something as simple as a SharePoint migration serves as a good example here. While software, platform and infrastructure innovations are all inherent in an AI Transformation, the nature of the change we are seeing is fundamentally different from how Scalable Platform Models work. Think of it in terms of directionality. The nature of AI Transformation involves a multitude of diverse automation, use case, centaur usage, and governance decision points across the organisation, each of which needs to be considered in their own right. Platform approaches do not address the need for conceptual scalability in AI, such as the ability to integrate new theories, models, or data sources rapidly. In fact if anything, pervious experience of Scalable Platform Models has trained people across organisations to see technology in a centralised way, which is diametrically opposed to how AI is manifesting across the corporate enterprise. Scalability in the context of AI often requires flexibility in cognitive architectures and learning approaches, which go far beyond traditional platform strategies that offer a single unifying answer and resource.
?
Ecosystem-Driven Approaches
Lastly, it is worth considering the Ecosystem-Driven Approach, which is based on creating networks of partnerships and collaborations to extend capabilities and innovate. This is a well-established strategy in many industries, often utilising the Triple Helix relationships that exist between commercial, research, and public sector institutions. With an Ecosystem approach, the emphasis is on the co-creation of innovation and the leveraging of synergies across different technologies and sectors. While beneficial for innovation and scalability, this approach can dilute focus and spread resources too thinly over diverse partnerships and technologies without ensuring deep integration or alignment specific to AI objectives. Ecosystem approaches can be very slow in terms of seeing gains and ROI, which is at odds with the pace of advancement we see in AI today, and crucially, they do not offer governance certainty, rather increases the risk aspects inherent in AI deployment, ethical AI use, and data security. Any form of AI ecosystem engagement requires careful management and control, to ensure data privacy, security, and ethical use across partners, which is often impossible to guarantee. For this reason, and the slow response speed of such diverse networks, Ecosystem-Driven Approaches alone are not working for organisations engaged in AI transformation.
?
Putting It All Together – Towards Something New
So, what is the answer? I’m aware I’ve gone through each of our established approaches finding faults and pointing out problems. I’m also aware it’s not enough to provide problems without also offering some solutions of value. You may be expecting me to suggest that all of the above methods are needed in order to successfully engage an AI Transformation. However, to say that we can combine all of the above together at once, would be a bit of a cop-out, and in my opinion completely unworkable in practice. That’s not to say some of the elements of the above approaches are needed in AI Transformation work, but even together these ‘best parts’ are not enough to successfully engage an AI Transformation. Due to the nature of the novel transformation problems AI is creating, which have no precedent in previous digital transformation work, I believe a new transformation methodology is needed. In fact, we are already beginning to see the first evidence of this new approach emerging in both our leading client engagements and research work. What do we know so far about this new approach to AI Transformation and what does it entail? These are questions I will attempt to answer and explore in greater detail in future editions.
Leadership Takeaways:
?
That’s it for this edition of AI Strategy Brief. Subscribe now, and if you like what you read today, please like and share it with your network to help me reach a wider audience. Have a great day, and I'll see you next time!