How to transform a company into an AI-Powered organization

Transforming into a data-driven, AI-first organization, will always be an ongoing journey since it is part of the overall Digital Transformation Initiative which calls for continuous improvement and adaptation (being agile as an organization). In a typical transformation scenario, which is well documented in the book ‘A Data-Driven Company’ by Richard Benjamins, we can consider 4 general states in this journey:

1.    Exploration

2.    Standardization & Transformation

3.    Data democratization

4.    AI-first

What happens in each state, depends on each organization and there is no golden rule for all. This means, that each organization will need to have the tools and metrics to measure its fluency in becoming a data-driven and AI-first company. Measuring its fluency helps understanding where the organization is and what needs to happen next.

Transforming into a data-driven, AI-first organization, is an ongoing journey.

In the ‘Exploration’ state the organization usually attempts to run ‘Proof of Concept’ (PoC) for specific business (use) cases. Usual candidates for such PoCs are marketing campaigns with the goal to improve their effectiveness or use cases like predicting and reducing churn (where applicable) or increasing average order value by cross-selling related/recommended products. From an organizational perspective, an owner is assigned, sometimes also called a champion. In general, this is a bottom-up approach, which means that those involved could be some managers that favor experimentation and data enthusiasts – typically data engineers and/or scientists. There is no formal organizational structure in this state – as this is seen as a specific initiative. This process can repeat over and over with the same or different teams; however, the organization is still considered to be in the “Explorer” state as this happens.

An organization has reached the ‘Standardization & Transformation’ state when there has been an executive decision that data is to be treated as a strategic asset, and that customer value creation must use data in a systematic way. Having reached this point, the organization already has a backlog of use cases to be addressed, and this backlog has been prioritized from the top and communicated throughout the entire organization. Two major initiatives take place during this phase: (a) Data Standardization and (b) Organizational Transformation. To deliver on these initiatives, the following must be there:

-       Data Sourcing Strategy. This strategy is about what data to use, and where to find it but it also aims to address organizational difficulties where certain functions consider data “their property”. Thus, the Data Sourcing Strategy also includes how data is to be shared across the organization.

-       Discrete Budget. In this state, each department needs to explicitly state their data, analytical or BI requirements, and this is to be approved in the annual budgeting process.

-       Formation of a ‘Data Team’. In this state, the organization has realized that a special Data Team is required, with a Chief Data Officer heading it. The positioning of this function within the organization is a whole other discussion, but in general, if we look at successful initiatives, we will see that the CDO is usually placed in organizational structures that are horizontal and apply to the entire business, such as IT or the Digital Transformation Team or under the COO. In some cases, the CDO can also be under the CEO but again, this is a topic for another discussion. The Data Team will consist of data engineers and data scientists and will have worked with IT on the technological choices required to set up the tools to perform the required work.

The next state is ‘Data democratization’. An organization has reached this state, when the use of data is included in the normal decision-making process, in addition to intuition, experience, or expertise. In this state, the Data Team has matured and has delivered the ‘Data Architecture’ for the organization. This includes:

-       The inventory of data sources which is always kept up to date.

-       A data dictionary to ensure a common ‘language’ and understanding across the organization. In addition, the data dictionary ensures data is traceable to its source, is granular enough and there is only one version of it.

-       Processes for data ownership and stewardship to ensure data is of the required quality, always up to date, and available to all those required.

In this state, the organization has already acknowledged that a lot of processes need to be adapted. Hiring needs to change to include skills such as ‘use of insights for decision-making’. Training programs need to be introduced to allow existing staff to acquire the new skills required. Training is very important because, in this state, employees are empowered through ‘self-service’ (processes, tools, knowledge of data, etc) to use data and insights in their daily tasks. Finally, for the Data Team, this state flags the transition from the focus on data engineers to data scientists, meaning that the number of data scientists at the end of this state, must surpass that of the data engineers.

‘Data democratization’ has been achieved, when data is used in regular decision-making.

The final state is ‘AI first’. In this state, the organization uses Machine Learning (ML) and other AI technologies to create value. This is the state where the organization will also reflect back and decide if adjustments are to be made to the data strategy, organizational structure, and so on. Having reached a level of data maturity, the organization will not hesitate to use ML for direct interactions with customers, for example for personalization, product recommendations, or chatbots (NLP); such decisions will be Business As Usual (BAU).

The Data Team will be enhanced by AI engineers (e.g., ML engineers, ML researchers) and transformed into a Data & AI Team. Organizational structure is critical in this state. AI talent and resources should not be working in a silo but should be part of agile end-to-end teams, besides the Product Owner and the other roles required to deliver a product or service. This will enable them to contribute their expertise in real-world situations and not operate in a vacuum. This agile structure is what makes an organization an AI-first organization.

So how does an organization measure its data fluency in order to assess where it is positioned in the data transformation journey and what needs to be done next? There are four dimensions required to be measured, to assess the organization's fluency:

1.    The Technology dimension. What technology choices are made for tools and platforms, how they are used, and by whom. Budgets will need to exist for all these tools, platforms, and so on.

2.    The Data Management & Governance dimension. This includes data protection, legal compliance, data security (encryption, anonymization, access control, etc). Data management also includes the functions of Data Architecture as previously explained.

3.    The Organizational Dimension. Changes the organization is making to adopt the use of data and AI.

4.    The ‘Business Dimension’. This measures the adoption of using data and AI throughout the organization to help in decision-making and optimization of processes but also includes efforts (such as R&D efforts) to use insights for new business opportunities.

Measuring each of the above dimensions will help the organization understand where it needs to adapt.[1][2][3]

Harry Mamangakis is a Technology Executive for over two decades, balancing between fluency in technological breakthroughs and having a business mindset. He has led and participated in several transformation engagements for leading brands in industries such as Telcos and Retail.

[1] The ins and outs of becoming a data-driven organization - Telefónica (telefonica.com)

[2] The AI-first Company, Ash Fontana, ISBN: 0593423089

[3] A Data-Driven Company, Richard Benjamins, ISBN: 1912555883