Most AI projects undertaken by organisations fail. The reason may lie in the lack of a solid strategy. Alessandro Lanteri, Luis-David Benyayer and Lorena Blasco-Arcas created a roadmap to guide organisations in the design, development and implementation of an AI strategy.
Amid the buzz about the transformative potential of artificial intelligence, it’s easy to either worry too much about its disruptive effects on industry or become too enthusiastic about its ability to foster rapid growth. Allow us to ground expectations with some sobering numbers: among 100 organisations that undertake AI-driven digital transformations, only about 20 reach a satisfactory level of success. To make it into the elite group of success stories, an organisation needs a solid AI strategy.
An AI strategy defines how data and intelligent systems will align with the organisation’s broader strategy. However, poor strategic alignment is not the reason why AI fails. Companies need to consider other reasons to fully understand how to better implement strategies. For example, the World Economic Forum (WEF) identifies six common reasons why AI initiatives fail to deliver business value: capability gaps, unavailable or poor-quality data, technology foundations not in place, poor governance structure, lack of understanding about project finance and lack of understanding about project management in artificial intelligence.
Given the diversity of factors that may influence the failure of an AI strategy, companies must anticipate how to avoid these pitfalls by creating a prioritised plan for implementing applications. We developed an AI roadmap with seven stages, grouped into three macro-phases corresponding to the design, development and implementation of a strategy.
Stage one: goal setting
AI-enabled strategic goals are the cornerstone of our roadmap. Initiatives should directly support the organisation’s overall strategic objectives, enhancing customer service, streamlining operations or mitigating risks. The next step is translating those goals into well-defined objectives that leverage AI capabilities. Intelligent technologies excel at predicting and optimising tasks that your organisation performs frequently and for which abundant data are available. A critical aspect of integrating is discerning which tasks are best suited for automation. Identifying the right balance between AI and human input becomes crucial for maximising efficiency and effectiveness.
Stage two: capabilities assessment
To obtain and maintain a competitive edge, organisations need unique resources. Tangible resources such as hardware; intangible ones such as data; and organisational ones such as employee skills. Capabilities in data management or AI project management are also needed. Creating an IT and skills inventory are the first steps to assess your current availability and, if necessary, plan to fill any gaps. A more holistic AI maturity assessment is also in order, to gauge the overall organisational readiness, and plan accordingly.
That is not all. Identifying the non-AI key resources and capabilities (brand, network, domain expertise, etc.) is necessary to identify which ones could be replaced by AI and how to combine AI with non-AI resources to create an even bigger advantage.
Stage three: data strategy
The effectiveness of AI depends on the quality and volume of accessible data. A robust data strategy is necessary to outline how the organisation collects, manages and uses data, keeping it accurate, accessible and safe throughout its life cycle. Recent findings suggest that using AI-generated content in model training can lead to significant biases in the models, which emphasises the importance of data quality. The ability to ensure that humans perform critical tasks unaffected by AI biases is emerging as a competitive differentiator. The challenge lies in effectively implementing strategies that uphold data integrity and quality.
Stage four: dual track implementation
Although this stage sits squarely in the middle of our AI roadmap, this position may be somewhat inaccurate. Running pilots is often one of the earliest steps organisations undertake on their AI journey – and rightly so. There are many lessons to be learnt from running pilots, which come in very handy at later stages of the AI roadmap, and they help build momentum for more substantial AI projects. On the other hand, creating the foundations for such more substantial AI projects is the outcome of the entire AI roadmap. This dual effort features in the middle of the process because it becomes a priority after the organisation has taken the first few steps and because it guides the following steps.
Stage five: budgeting
AI is evolving at such a rapid pace that budgeting even a year ahead is very challenging. Cost curves can suddenly drop. Other costs are hard to quantify and are typically not even included in return on investment (ROI) calculations. That includes costs with user feedback loops and subsequent reworks or funds for the cleaning, normalising, and cataloguing of data.
New applications may become available that create unprecedented savings or revenues. A more incremental and flexible approach to budgeting minimises the risk of major budget overruns and lets you collect valuable data to develop stronger business cases for more substantial investments.
Stage six: ethics and compliance
Technologies are not intrinsically good or bad. It’s how we use them that defines their value and impact. Electricity can power infant incubators and electric chairs used for capital punishment. Mobile phones enable financial inclusion and financial scams. But AI is different. It stands apart due to its profound capabilities and potential risks. At ESCP, we recommend that responsible AI be built around five pillars: fairness, safety, privacy, transparency and accountability.
Stage seven: change
Some people seem to think that as soon as they have access to the latest technology, they have a digital organisation. Technology is only part of the story. As a final stage to successfully deploy AI, you should align with many stakeholders, both inside and outside the organisation, nurture a truly digital culture, and continually upgrade the digital capabilities of your employees. The synergy between technical and human assets is key to realising the full potential of AI. Success in this domain not only depends on redefining processes but also on fostering an organisational culture that understands and trusts AI. Open dialogues about the technology’s capabilities, limitations and optimal use cases are essential to develop a shared vision for AI integration.
Why an AI roadmap
AI projects require significant time and money. When they fail, as they mostly do, they leave behind horror stories that will scare decision-makers away from future attempts.
Our AI roadmap has six ambitions (but we will be pleased if even only one is relevant to your organisation).
Guidance
A map removes some of the guesswork. It saves time spent in meetings to decide what to do before you start doing anything.
Alignment
It ensures that the deployment of AI is directly tied to strategic business goals, helping to maximize ROI and impact.
Efficiency
By planning out when and how resources are needed, organisations can better manage budgets, staffing, and other critical resources.
Clarity
It serves as a communication tool that keeps all stakeholders informed about the goals, processes, and progress, facilitating alignment and support.
Risk mitigation
Our AI roadmap addresses each of the most common AI pitfalls we mentioned above.
Responsible implementation
Every AI project needs to carefully consider and assess the impact of its implementation at the individual, organisational and societal levels. A high-value AI strategy brings a positive impact on societal values. Careful monitoring of biases and ethic-related aspects in the design and development phases seem crucial to avoid further risks in the implementation phase.
Sign up for our weekly newsletter here.
- This blog post represents the views of its author(s), not the position of LSE Business Review or the London School of Economics and Political Science.
- Featured image by Steve Johnson on Unsplash .
- When you leave a comment, you’re agreeing to our Comment Policy.
