My MBA Journey

Record of my personal journey completing an MBA

Week 5 – A foundation for AI – the AI ladder approach

A foundation for Artificial Intelligence - The AI L


This module examines the practical, technical, and structural challenges organisations face when implementing AI. IBM has developed a guide known as the AI Ladder to assess an organisation’s readiness for AI transformation.

Utilising this ladder can aid in evaluating an organisation’s capacity and preparedness for introducing and integrating AI within its structure.

The primary lesson from this week’s study is the transformative potential AI affords organisations, a process that requires meticulous planning and execution due to its unfamiliarity. It necessitates gathering pertinent information within the organisation about where AI can be most beneficial, supported by suitable tools and frameworks.

Interestingly, particularly in SMEs, many business individuals still fail to comprehend the revolutionary power of AI or understand its data and infrastructure needs (Ransbotham et al. 2017) [^1]. Consequently, when an attempt at implementing AI fails, it is typically due to a deficiency of strategic knowledge needed for technology implementation.

Therefore, hiring consultants may be beneficial if affordable; otherwise, significant investment in gaining technological knowledge should be considered essential for smaller businesses.

Common obstacles to adopting AI include “developing an AI strategy with clearly defined benefits, finding talent with the appropriate skill sets, overcoming functional silos that constrain end-to-end deployment, and lacking ownership and commitment to AI on the part of leaders” (Manyika & Bughin 2018) [^2].

Another crucial aspect of implementing AI involves managing change within the organisation. The introduction of new technology can incite fear amongst employees who must be involved in this journey.

Assessing internal skills related to dealing with AI is also vital; recruitment might become necessary if there are not enough qualified people on board already. Organisations that disregard considering AI now risk falling behind early adopters on the innovation curve who could monopolise available talent (Manyika & Bughin 2018) [^2].

Figure 1: AI adoption indicators.

Figure 1: AI adoption indicators
Source: Manyika & Bughin 2018 ^2.

The AI Ladder

The AI Ladder, a framework conceived by IBM and Rob Thomas, enables organisations to chart their readiness for the implementation of Artificial Intelligence (AI) solutions (Thomas 2019) [^3]. This model primarily revolves around four key areas as shown in the image below.

Figure 2: The AI Ladder

Figure 2: The AI Ladder
Source: The AI Ladder Framework from Thomas 2019 p.2 ^3.

Strategic management remains crucial when introducing new technology, such as AI, because of its novelty. Many people involved in planning may not fully comprehend potential consequences that could emerge from its application. By consequences, it means unforeseen outcomes that could impact an organisation’s profit-and-loss or balance sheet adversely.

Preparing data for transition into an AI environment is paramount. This often necessitates migrating substantial amounts of data to the cloud where computational capacity is superior. Eliminating data silos, isolated pools of stored departmental data, also becomes essential since fragmentation can significantly diminish the value of data and impede an organisation’s ability to extract meaningful information.

Another concern arises if departments utilise separate cloud providers for their respective datasets using proprietary software locks. This compounds complexity further. Therefore, it becomes crucial for an entire organisation to operate under the optimal solution for cloud utilisation by consolidating all their datasets into a single cloud application (Thomas 2019) ^3.


The initial activity of an organisation is to Collect. Thomas (2019, p. 8) [^3] describes this as follows:

For too long, data has been held captive within systems of record, and isolated by rigid platforms, segregated business functions, and data types. This results in siloed data, which is difficult to access, making it impossible to gain true analytical insight.

The process of data collection involves capturing and consolidating all relevant information in a single location, an activity intrinsically linked to modernisation. The decisions made during this stage significantly influence future organisational strategies regarding data storage and access.

Given the dynamic nature of data sources within organisations, maintaining flexibility is vital when deciding on storage locations for collected data. When collecting, it’s not just about the immediate organisationally generated data, but also potential external datasets that might prove beneficial for analytical purposes and competitive advantage enhancement.

While primary sources of data are often recognised by marketers and business executives, secondary sources such as social media trends or news items should not be overlooked. These can potentially impact business direction and offer advantages through AI interpretation.


Moving on to the Organise step, Thomas (2019, p. 10) ^3 describes it as follows:

In Forrester’s Predictions 2019: Artificial Intelligence report, 60% of the decision-makers at firms adopting AI cite data quality as the number one challenge when trying to deliver AI capabilities. Today, organisations spend 80% of their time preparing data for productive use, creating a bottleneck for business agility, competitiveness, and profitability.

The proliferation of cloud and mobile technology has resulted in an exponential increase in both structured and unstructured data. However, many organisations are unaware of the nature, location, processing requirements, and compliance status of their own data.

In this context, three critical considerations arise for any organisation embarking on the AI journey. Firstly, they must assess the quality of their data to determine its cleanliness or completeness. They need to ensure it complies with relevant laws and regulations, and also ascertain that it is ready for AI model development. Without these prerequisites being met, using such data will prove challenging for anyone involved , whether it be management, analysts or end-users interacting with your business.

Properly designing a data lake can resolve inconsistencies within the data and create a reliable single version. The result will enhance the overall trustworthiness of the data to engender the confidence of users.

Next comes organising the information. Drawing parallels with a library without a catalogue would be apt here. Books scattered everywhere but impossible to locate due to lack of organisation. The same principle applies to business related data. It needs structure for easy access.

Lastly, establishing protocols and policies governing ethical usage is paramount. Compliance with legislation should be ensured along with granting access only to authorised personnel, thereby reinforcing proper governance over any organisation’s valuable resource, which is its data.

A significant piece of information in the video was the statement that the Organise stage can “take up to 80% of data science project time” (AltexSoft 2021) ^4.


The progression to the Analyse step is the transition from the data to where the building of the AI begins.

According to Thomas (2019, p. 13) ^3

In today’s world of regulations, GDPR, and data privacy laws, the way organisations engage with AI is under intense scrutiny. Organisations need to manage their AI across the entire AI life cycle in order to explain either to a consumer or another business how their systems came to a decision and why.

To reiterate, the importance of understanding decision-making processes in systems is crucial for fostering trust. It’s imperative that organisations can identify and communicate these reasons to users.

Now, let’s proceed with Thomas’ (2019) three-step process for AI model construction:

  1. Build – The first step involves building the AI models, akin to working on a bench. Organisations create and train models they can use for predictions. At this stage, identifying proper algorithms for model creation is vital to ensure reliable predictions.
  2. Run – The second step entails running the constructed model within an application or process to test its decision-making capacity concerning selected processes, pricing models and other roles it may need to perform. This phase also accommodates any necessary retraining of the model to adjust bias or other identified issues.
  3. Manage – In the third step, we address model management post-construction and implementation. How will it be scaled? Trustworthy and transparent behaviour of the model must be ensured by organisations. A black box scenario where decisions by AI cannot be explained to authorities or users is undesirable (Schatsky 2019) [^5] . Lastly, having procedures in place allowing management to track data access history, training undertaken and alterations made to algorithms is essential. This safeguards responsibility, accountability, and transparency regarding all actions executed within the organisation.

A positive aspect to be considered is that all parties in the AI space are aware of the difficulties surrounding bias, transparency and other issues. There are deliberate attempts at all levels to develop appropriate frameworks to deal with the matters around governance. It is conceivable however, that given the speed with which governments operate, they will always be behind with such fast moving technology.

The video raises an interesting point and suggests that our current approach to knowledge encourages depth over breadth. It prompts us to question whether this strategy is effective in the realm of AI, where models, programs and algorithms are predominantly viewed through a computer science lens without interdisciplinary integration. The consideration of an interdisciplinary method, where social sciences such as philosophy, sociology and law are linked with computer science, may be a more valuable model. This comprehensive approach could promise improved outcomes.


The final step in the AI Ladder is Infuse where the completed and tested model is ready to be rolled out. Referring to Thomas (2019 p. 17) ^3 again, who states, “It’s about automating and optimising processes in whole new ways and applying AI to unlock new value for its business, regardless of size or industry.”

Undeniably, the application of Artificial Intelligence (AI) is universal across all sectors, irrespective of their size or type.

While small to medium-sized enterprises and non-profit organisations may opt for readily available solutions, others might choose to develop bespoke systems. Regardless of the approach taken, it’s essential to reiterate that AI holds tremendous transformational potential for any organisation.

Revisiting Thomas (2019) [^3], he provides several instances of AI’s potential implementation in organisations, notably in customer care. Additionally, it could be employed by knowledge workers in various domains such as marketing, supply chain management, human resources, insurance sectors and financial planning. Reference to the AI ladder document is highly recommended for a deeper understanding of AI’s capabilities within organisations. Consideration should also be given to how it could benefit your own organisation.


In conclusion, Thomas (2019) [^3] refers to AI as the “new electricity” because of its transformational power in the ways we do business, not to mention the broader aspects across society itself. Perhaps it is valuable to finish with a quote from Thomas (2019 p. 17) ^3] about AI when he says:

Finally, AI is not magic. It’s hard work. It requires the proper tools, methodologies, and mindset, to overcome the gaps that companies are facing (data, skills, and trust) to truly embrace an AI practice and put it to work across your organization.


[^1]: Ransbotham, S, Kiron, D, Gerbert, P, & Reeves, M 2017, ‘Reshaping Business With Artificial Intelligence: Closing the Gap Between Ambition and Action’, MIT Sloan Management Review, Vol. 59, No. 1, p. n/a-0, Massachusetts Institute of Technology, Cambridge, MA, Cambridge, United States.
[^2]: Manyika, J & Bughin, J 2018, AI problems and promises | McKinsey, McKinsey & Co, viewed 23 July 2023,
[^3]: Thomas, R 2019, The AI Ladder: Demystifying AI Challenges, ,
[^4]: AltexSoft 2021, ‘How is data prepared for machine learning?’, , YouTube video, August 31, viewed 23 July 2023,
[^5]: Schatsky, D 2019, Can AI be ethical?, Deloitte Insights, viewed 24 July 2023,

Leave a Reply

Your email address will not be published. Required fields are marked *

Picture of Ric Raftis

Ric Raftis

Find out more about me on my About Me page.

Share this post:

Read & Learn More

More From The Blog

Inspirational content to help you shift your life into the path of success