AI Strategy for Executives

Why Do Executives Need to Know About AI?

It may seem like artificial intelligence is rapidly taking over all aspects of business right now, but the truth is that it’s still early for AI: Most of the value it will generate still lies in the future. This does not mean that every business will benefit equally from adopting AI, though: The early adopters will reap the vast majority of the benefits, as a recent McKinsey study shows.

AI is not a magic black box. It is a set of technologies and practices that can help make your business more data-driven, so you get deeper insights into your customers and can deliver better products and services more quickly and more efficiently.

Key Factors to Ensuring Success

While AI has enormous potential, digitalization projects focused on AI often don’t deliver the desired results. The fundamental reason for this failure is that those at the executive level have not made the necessary organizational and cultural changes to set up AI projects for success. The following are key actions to consider:

  • Take a Whole-Business Approach: To get the most value out of AI, it needs to be a technology that touches different business units and functions, not just IT. AI affects areas as diverse as product design, operations, finance, and marketing, and they all need to play together.
  • Understand Push Vs. Pull: Most successful AI projects go from the bottom up rather than being directed from the top down. Empower your people to propose their best ideas and then run with them.
  • Embrace a “Fail Fast” Culture: In traditional business projects, failure is undesirable and avoided, so leaders whose projects fail are penalized. The opposite is true in a robust engineering-driven culture where AI thrives. Multiple projects are advanced fast and rapidly evaluated for their potential to be successful: Fail fast, and you learn fast.
  • Beware of Hype and Overpromising: AI has, unfortunately, been promoted as a technology that routinely delivers a 10× return on investment, which is not realistic. Where AI shines is in making your product or service better at those margins that matter for beating your competitors.
  • Emphasize Good Communication: The gap in knowledge between technical people working directly on AI and their peers in other business units needs to be closed via dedicated “analytics translation.” This business role is now being developed to bridge that gap.

AI Journey Steps

What does a typical AI Journey look like? 

AI Readiness Assessment and Business Case Development

The first step in realizing an AI project is to develop a solid business case:

  • What’s the objective?
  • How will it be measured? 
  • Which KPIs does a solution need to meet to be considered a success?

These questions must be answered ahead of time by a team representing the various stakeholders, not just IT. For this purpose, management must have a clear strategy and the full understanding of potential use cases, business units’ projects and requirements. It is of paramount importance to build an action plan and prioritize projects with high impact and high feasibility, to start achieving those “quick wins” which create consensus around AI adoption. Management needs at the same time to allocate appropriate budgets, reflecting these strategic decisions, to the relevant business units and encourage the implementation of a clear governance  structure where best practices and standards are adopted consistently across business units.

Before starting a major development effort, make a first review of the existing data. There is a danger here: Upon realization that their existing data is scattered and of low quality, businesses often start investing in major data infrastructure projects without a clear goal. Such processes can be extremely costly and often do not support the actual use cases that the business is trying to tackle. Instead, an initial analysis of the existing data often reveals low-hanging fruit that can be reached without a major investment. This analysis can also set the direction for what kind of data is currently not collected but should be.

Complete an AI Proof of Concept

The next goal is to get one or more proofs of concept (POCs) completed as fast as possible. The most important blocker to rapid development and iteration is often the quality of the data available, so manual work is required to get the data in shape to be used for training machine learning models.

Modern automated machine learning platforms like Modulos can get POC models done for you within days—not weeks or months, like models developed by hand using conventional methods—thereby enabling the “fail fast” approach. The POC models allow you to rapidly reach the point where you can perform a gap analysis relative to your target KPIs to see whether it is worth continuing development and, if so, what additional resources are required, particularly on the data side.


Taking a POC model to production is often the longest and most difficult part of the AI journey. This is the phase where most AI projects fail—usually not for fundamental technological reasons but because the project was not set up for success by management. 

Of course, IT needs to make the model part of its operations. To deliver value, an AI model needs to scale, and it must be continuously monitored and refined. However, this means that change management across business units is essential. Processes need to be adapted and created to support the flow of essential information and feedback from all areas of the business. 

Use Cases

Analytics translation means taking a complex business task and reducing it to a question answerable by an AI model. In reality, many different business tasks map down to a relatively small number of AI models, which is why modern AI platforms can rapidly build useful models and easily adapt to customer data and needs. 

Hiring a team of data scientists, machine learning engineers, and data engineers and integrating them into a mature business can take a long time and require many adjustments to company culture and processes. Modern software platforms that enable building AI models can shorten the process and reduce upfront investment. 

We present some common use cases from various industries on our website:

You can also find video recordings of our team going through use cases here:

AI Regulation and the EU AI Act

When it comes to artificial intelligence, executives need to pay attention not only to technological/innovation concerns but also regulation. Until recently, AI was not regulated at all, was covered by existing legislation, or was merely subject to voluntary codes of conduct and policies. The goal of these policies so far has been generally to increase transparency in the use of AI and to guide the development of ethical, non-discriminatory AI systems.

This “light-touch” regime is now almost over, as indicated by the imminent passage of the European Union’s Artificial Intelligence Act (the “EU AI Act”). In the works for many years, this Act has the support of the European Commission and is now in the final stages of negotiation, with a vote in the EU Parliament scheduled for the fall of 2022. 

Why should you care about the EU AI Act? 
Here are its main points, in brief:

  • The EU AI Act will regulate all uses of AI in the EU and defines “AI” in extremely broad terms. Its definition includes not only sophisticated, deep-learning systems for applications such as self-driving cars but also basic mathematical and statistical techniques in common use today. 
  • The penalties for violations are severe: 6% of global turnover or €30 million, whichever is higher.
  • Its reach will be global: just as GDPR set the worldwide standard for privacy, the EU AI Act will set the global standard for AI. This means that even companies not based in the EU will, as a practical matter, have to comply with the Act.
  • The Act has many requirements and differentiates AI systems by the risk they pose to society. A common theme is that data quality is essential and that models must be fair and robust—something that we at Modulos fully believe in as well.

All your AI initiatives need to be future-proofed based on the EU AI Act. Any AI project that you start now—whether at the ideation or POC stage or close to production—should already conform as much as possible to these new requirements. Starting the cultural and organizational shift to adjust to this new regulatory landscape now will give you an edge over your competitors that have not yet started.

Key Takeaway Points

AI is not a topic just for the IT department but is the responsibility of executive leadership. For AI to deliver ROI, executives must lead the organizational and cultural shifts in their companies.

Companies that have started their AI journeys so far have had mixed results—some have been able to reap massive benefits, while others wasted budget and time on projects that did not deliver. By utilizing a state-of-the-art AI platform like Modulos, companies can deliver and iterate faster.

Our two key takeaway points are as follows:

  1. Take into consideration the full scope of organizational and cultural changes needed to adopt AI.
  2. Adapt and optimize technology to rapidly launch projects that deliver ROI.