Without a proper strategy, an organization will never become AI-driven. What are the elements of a winning strategy? Are sneakers and hoodies, and a smartwatch around the wrist required? Rest assured, they are optional. But culture, amongst other things, is an essential ingredient for success with data and AI. In this article I will answer the following questions:
- What is an AI Strategy?
- What are the AI capabilities that leading organizations focus on?
- Which tradeoffs have leading organizations made to define their AI strategy?
Apply the practical experiences from this article to get one step closer to an AI strategy that is ambitious yet executable for your company!
What is an AI Strategy?
As a consultant, I have seen many senior executives struggle with the definition of (their) AI strategy. Their pondering went far beyond the question if they should change their attire.
Often, business leaders start exploring their data sources to find new business opportunities. Monetization of data assets is an important goal of a successful Data & AI strategy, but it's always intertwined with other objectives and aspects.
In my experience, a winning AI strategy describes the right directions for organizations to turn their AI ambition into reality. It is the most general resource allocation plan, and can be described as follows:
AI strategy is your plan for acquisition, organization and allocation of AI resources, to turn your AI ambition into reality.
Let’s take a closer look at what these AI resources are and the topics that make an AI strategy.
Core ingredients of the AI capability
The components of an AI capability of a company are its AI resources, being:
- People & Skills
- Tools & Technology
This blog focuses on the contents of an AI Strategy and not on its execution, so please have a look at our AI Maturity Journey whitepaper for best practices and learnings from large Dutch companies like Randstad, KLM, ING and Dynniq on how to develop your AI capability.
AI Strategy topics
The AI strategy contains a roadmap on how and when to acquire, organize and allocate AI resources to turn the company’s ambition into reality. This roadmap basically is a course of action, a set of choices on many topics, setting a clear direction to guide day-to-day decision making. Each topic comprises a trade off between short- and long-term investments and benefits.
Below you will find example trade offs of every AI strategy. Please note that this list is not exhaustive, because many other generic or industry specific trade offs may be relevant for your company.
In practice you will most likely not settle for either side of a trade off, but somewhere in between. Hence, I have decided to present the trade offs with “&” instead of “vs". It's the challenge to find an acceptable level of drawbacks from both ends that brings you the most relevant benefits.
Business Demand & Data/Technology Push
AI solutions may be initiated from both demand from business teams and push from data experts. While initiatives from business stakeholders are preferably derived from strategic objectives and aligned to it as a consequence, initiatives from just data or technological opportunities may lead to innovative new ways of generating value from existing data assets that would not have been found working from business demand only.
Yet, demand from business exists for a reason, so the AI Strategy needs to set the balance between innovation from data/technology and developing solutions according to business demand.
At our clients we often see great ideas from business stakeholders on innovative solutions. Without initial exploration, we don’t know whether these innovations will be feasible. The AI Strategy should both facilitate time for research and experimentation and guidelines on how and when to stop this if things don’t work out as anticipated.
Day-to-day business & Business strategy
While working on AI solutions data experts may get many ad hoc questions from business stakeholders related to day-to-day operational matters. Sure we prefer to work on AI solutions addressing structural needs and benefits. Sure we are enabling business people more and more to perform small analytical tasks by self-service BI. But in practice we see that very valid ad hoc questions do exist, which cannot be answered by the business people involved (yet). These questions may be very well balanced with respect to effort needed and its benefits and deserve attention by your data experts.
The resource allocation plan from the AI Strategy should allow this. Moreover, it should contain guidelines on how to handle these ad hoc questions and whether to invest time now to democratize analytics even more to reduce the number of unplanned questions to your scarce data experts in the future.
Business opportunities & Ethical behavior
In general, there is never a valid reason not to comply to regulations on data privacy and data security. But knowing what you are not allowed to do with your (or the customers') data does not mean that you want to do everything else that is allowed. In the allowed zone, there are still potential business opportunities of data assets that are less wanted. This is because of ethical considerations, company values or e.g. some expected stricter regulations.
It is this grey zone in which the AI Strategy should be clear on the code of ethics and the risk appetite of the company. This topic is strongly related to data governance. In fact, while implementing data management processes for your company, it is your data & AI strategy that sets the balance between enabling (offensive) and restrictive (defensive) data governance policies.
By the way, by positioning data governance as an enabling process, by means of democratising access to data sources in a controlled way, we have been able to get much more positive attention to data governance at one of our clients.
Data driven & Business driven
At GoDataDriven we help our clients on their journey to become a more data driven organization. In practice, this means, amongst other things, that business teams and data experts have to collaborate more continuously in their daily work. They have to become familiar with each other’s language and each other’s objectives. So one can say that in becoming more data driven as a company, data experts also need to become more business driven.
To support this process, we have trained people at several clients to become take on the role of Analytics Translator. They take the lead on development of AI solutions from ideation to production and are the liaison between business and data experts. Please see our Analytics Translator whitepaper for more information on this role. Your AI strategy has to identify the need for training programs, AT roles, and has to describe the organizational implications to support this process.
Make & Buy
Data Scientists share an insatiable hunger for knowledge. They are always eager to learn new methods and technologies. They love the challenge of working on a variety of use cases from different business domains.
The development of AI solutions takes considerable time and effort. Before you can start to think about a solution, you need to become familiar with the business domain. This includes getting to grips with all the specific data.
Data science teams are usually very limited in size and capacity. Moreover, the number of people at your company able to do this (the data scientists) is very limited. Organizations need to be selective on the use cases they can work on.
A good rule of thumb here is to develop domain-specific solutions in-house, and to use generic solutions for non core business applications. It is very important to define this clearly in your AI strategy.
Enabling technology & Hiring expertise
Data experts need proper tooling to develop AI solutions. There is a wide range of tools available from free up to very expensive, from open source to proprietary software and from full flexibility for adopting new technologies to this will be the single tool for the next 5 years. Another aspect of tooling is the amount of training and software engineering skills you need to use the software.
Next to tooling data scientists and data engineers need a strong background in e.g. statistics, machine learning, software engineering and cloud engineering. A theoretical and practical foundation that continuously needs to be updated as methods and technology evolve.
Your data team should consist of the right mix in seniority, theoretical expertise and business experience and that’s why you need a talent strategy. Decide on the budget for both hiring data and AI experts and nurturing this talent continuously.
You need to make choices on the AI enabling technology that you are willing to invest in and accept the additional requirements on knowledge and experience of your new data hires as a consequence.
We have made finding the right people a lot easier for one of our clients by offering a managed data platform so the hiring manager doesn't have to look for the skills to set up and maintain advanced cloud data technology.
Your AI strategy should describe your ideal mix of talent and technology including the way you will acquire, organize and nurture these resources.
Your AI strategy
This blog covered several potential topics for your AI Strategy. However, there is no single standard AI strategy that will be applicable to every company. It will heavily depend on e.g. the size, culture, AI maturity and the industry of the company.
If your company is small and you are just experimenting how AI might be of value for you it is probably smart to direct your AI Strategy towards flexibility in people and technology and start with small showcases. When you have advanced a bit more on your AI journey you need a broader AI Strategy focusing on acquiring your own AI resources and organizing them for future scaling.
Would you like to join us and help companies to define and execute their AI Strategy? Please have a look at our Careers page on the AI Strategy Consultant position!