Data-Driven strategies

Principles

When dealing with strategic questions it helps to follow three principles. The first principle is defining purpose. It is essential to begin with your organizational goals in mind. Data and analytics are a means to an end and it pays off to make explicit what impact you want to achieve. Do you want to leverage data and analytics to increase profit, efficiency, customer satisfaction, the number of customers, or do you want to reduce your organization’s carbon footprint? Whatever goals you have as an organization, make sure they are measurable. These measurable goals make explicit which organizational value your data and analytics efforts aim to contribute to. They also make it possible to evaluate the impact of your strategic efforts.

The second principle is evaluating readiness. To generate value from data it is important to align your strategic ambitions with your organization’s current data maturity. The approach to leveraging data and analytics will vastly differ for a company with no prior data and analytics experience compared to one where data is at the core of its operations. This is not to say that data should not play a role in your strategy when you find yourself in the first scenario. Realizing where your organization currently is on a spectrum from data aware to data fluent helps in keeping your data strategy focused and realistic. Our free online data maturity survey can help you with this assessment.

Finally it is crucial to cover all your bases. Keep in mind that your data maturity manifests itself in multiple dimensions. We distinguish five important dimensions: people, organization, process, technology, and data. In the people dimension you cover questions like what skills are necessary to meet your ambitions? What profiles and career paths do you define? What is the right mix of internal and external talent? The organization dimension involves questions like what is the right team composition? How are teams, projects, and data products managed? How are budgets allocated and by whom? The process dimension covers topics like ways of working, best practices, and product life cycle management. Only in the technology dimension do you start dealing with the technological capabilities that are in place or need to be in place. This includes architectural choices as well as build or buy decisions. Finally the data dimension relates to all data governance issues. Among other things, this includes data availability, accessibility, understandability, and quality.

Evaluating your current data maturity as well as your ambitions across all these five dimensions is crucial. As you define your data and analytics strategy, pay special attention to any dimensions that might have been overlooked in the past.

Summarizing, the key is to generate measurable value through a maturity-aligned strategy that covers all five organizational dimensions. Figure 1 ties the above principles together visually. This framework helps in drafting a complete and effective data and analytics strategy. To illustrate the process we now turn to a strategic question that has become extremely relevant recently: what should your organization do with generative AI technologies and tools, such as ChatGPT?

Case study

In this post we take as an example a fictitious for-profit energy provider EverGreen with a well-established data and analytics department. Let’s assume you are responsible for their data and analytics strategy. EverGreen has been at the forefront of using data to optimize its operations and improve their customer’s satisfaction. Recently, you have read many articles about the transformative potential of generative AI technologies like ChatGPT. Naturally you wonder what role this cutting-edge technology should play in your organization’s current data strategy.

Defining Purpose

Evergreen is committed to improving customer satisfaction, reducing cost, and reducing their customers environmental impact. Given these high-level goals Evergreens leadership team has formulated a business strategy to make their chatbot the go-to-place for customers to ask any questions they might have about their energy contract as well as effective energy saving measures for their specific house. Questions that the chatbot cannot answer should be forwarded to human customer service agents. The improved chatbot should be intuitive, effective, but also reduce the overall number of calls to the call center.

Note that the formulation of the above business strategy focuses on organizational purpose and criteria for success, not on technological implementation details. This is where your data and analytics strategy comes in. To meet these ambitious strategic goals, you expect some form of generative AI is needed to get the current chatbot to the next level. Since the use of generative AI for business purposes is relatively new, it is yet unclear if the technology can fully live up to its promise. Also, without proper guidelines generative AI can pose significant risks. Customers or employees interpreting incorrect answers from a chatbot as facts has resulted in incidents with world-wide media coverage. For example, a lawyer in New York has presented ChatGPT output containing fictitious legal cases as actual legal cases. Therefore a deliberate strategy is advisable when starting to use generative AI in an organizational setting.

To leverage generative AIs potential while mitigating the risks you decide to take a step-by-step approach and add two strategic goals to your current data and analytics strategy. The first goal is to allow employees to safely use tools like chatGPT. The second goal is to learn first-hand about the benefits and limits of embedding generative AI in EverGreen’s data products in a clearly scoped strategic pilot. This example illustrates how you can align the goals of your data and analytics strategy to your organization’s business strategy and ultimately EverGreen’s organizational goals.

Evaluating Readiness

The next step is to assess your current data maturity in the context of incorporating generative AI. While your data and analytics department is very capable at handling structured data, you recognize that generative AI would require dealing with unstructured data, natural language processing, and the ability to integrate large language models within your data products. These are new skills for your data team. While the chatbot team has experience with traditional rule-based chatbot technology, they are new to the potential and risk of using generative AI in a chatbot context.

Within your organization you have already formulated a policy on what your employees should and should not do with tools like ChatGPT and CoPilot. It is also clear that some teams are enthusiastic about the use of these tools and want to use it more intensively. Because you are aware of the potential, limitations, and risks of generative AI, but otherwise have limited experience with it, it is fair to assess your level of maturity as low in this context. We call this maturity level ‘aware’. The next level of maturity as formulated in the above data strategy ambition would be ‘knowledgeable’, a situation where some part of the organization has some capabilities to leverage generative AI effectively, but still in a pilot setting. The overall data organization is not yet affected by generative AI and many aspects of leveraging it effectively still need to be ironed out.

Once you have assessed your maturity in the context of generative AI, it is time to make the strategy concrete. This brings us to the different dimensions that require attention.

Covering All Bases

Considering the organizational dimensions, you realize that implementing generative AI requires a multi-faceted approach. You outline the five key dimensions and assess your next steps in each.

In the people dimension, you decide on the team composition of the pilot team. You also identify the need to hire external talent to train and coach the team on the use of generative AI ranging from prompt engineering to best practices when using large language models in software products. From an organizational perspective you ensure active support from the leadership as well as ensure that the right people from different departments are collaborating together in a cross-functional pilot team. This end-to-end team setup increases the changes for success. In terms of process you agree on an agile way of working. Within each development cycle clear goals are set, solutions are designed, documented, and implemented, and both the end result as well as the team collaboration and process are evaluated and improved where needed. This regular feedback loop helps in adapting to the unforeseen obstacles, while keeping focus on reaching the end goal: a working prototype.

You also make the necessary technological choices. You decide to use the Azure Open AI service to combine a high out of the box model quality combined with all the enterprise features of your existing cloud platform. Last but not least, you identify the data governance requirements. For the pilot you decide to initiate a test program with friendly customers. All participants will sign for an opt-in that their interaction with the pilot chatbot can be used to improve the prototype. This reduces the risk of the pilot if the prototype would behave in unanticipated ways.

Once you have settled on the above strategic choices in each dimension, you go back to your defined purpose and maturity and double check if they are all aligned. This ensures that your generative AI data strategy fits into your overall strategy and increases the data maturity of your organization across all dimensions.

The usefulness of this strategic approach is twofold. On one hand it signals a deliberate organizational commitment to start the generative AI chatbot pilot and build up initial generative AI capabilities. After all, an equally valid strategic decision could have been to not become an early adopter and wait until other organizations show the added value of generative AI in a chatbot pilot. In that case it might also make more sense to rely on third party services in the future and not build these capabilities within your organization. An essential characteristic of a strategy is to navigate such trade-offs consciously.

A second advantage of this approach is that this type of strategizing provides a structure to identify blind spots in your organization’s capabilities. Although strategizing itself does not guarantee the success of your customer service ChatBot pilot, it does increase EverGreens chances to advance their position as a customer and environmental friendly energy company.

Conclusion

In the modern world of data, technologies like generative AI, machine learning, dashboarding, big data processing, and cloud computing, are just tools. This is why driving impact through data and analytics begins with a well-defined purpose and data strategy. Since every organization is unique, there’s no one-size-fits-all solution. However, adopting a structured strategy framework when addressing strategic questions specific to your organization can be immensely beneficial.

At BigData Republic, we have a lot of experience in helping organizations increase their data maturity. Besides providing technical expertise, we assist organizations in assessing their current data maturity and devising a data strategy that aligns with their aspirations. To make that next step first try out our online data maturity scan. The resulting personalized report provides insight into your organization’s current maturity and suggests concrete advice on what to focus on next in your data and analytics journey. If you require even more customized advise in your strategic journey, reach out to us to discuss how we can help your organization. You don’t need to be a tech giant to deserve an effective data and analytics strategy.