Artificial Intelligence: A Senior Executives Action Plan
Jan 14, 2020 · 5 min read


The term ‘AI’ (Artificial Intelligence) has gathered momentum over the last few years and is arguably today’s most over-hyped business buzz phrase. The vast majority of Global 2000 CEOs are being challenged by their Boards to demonstrate operational and financial benefits from applying AI to their businesses. Almost every large company has established some kind of working team to brainstorm and prioritize AI applications, and many companies have funded some specific pilot projects that have emerged from this process.

In our experience, most of these initiatives will end up disappointing their sponsors. This is because they tend to veer in one of two directions: either they are so grandiose that they fail under their own weight, sometimes at great expense (example: create an integrated data architecture that breaks down silos and build a suite of AI analytical services for the collective enterprise), or they pursue interesting and useful technologies that in the end produce too little incremental profit to justify all the cost and management attention they require.

Our experience is that the following specific practices for executive management of AI initiatives will most reliably lead to success:

1. Work Backward From Business Problems, Not Forward From New Technologies

In theory, starting with a list of new technologies and then rigorously determining how much money is at stake in applying each of them to various business challenges should get to roughly the same place as starting from business problems and rigorously testing the business case for applying new technologies against them. But in practice we have found companies are much better off starting with the business problems, primarily because evaluating technical feasibility is a far more delegable task than judging where the profit opportunities sit in a business.

Specifically, senior executives should identify a short list of core decision processes with high-profit leverage that would be improved with better data utilization. We have rarely found them to be wrong about this. The work of the staff is then to estimate the value-at-stake for each process that is addressable with the AI technology of today (not the potential technology of five years from now). The senior executives should then select 1–3 of these as pilot initiatives and directly sponsor the pilot projects.

2. Focus Relentlessly on Financial Value

Every one of these AI pilot projects should create incremental free cash flow within 12 months, or else be terminated. Corporations pursue strategies that build value over time, and an overall AI strategy should do the same. But in this case, the “Trust me boss, we need to spend money for the next 4 years, but we will have a huge payout starting in year 5” approach almost always ends in disappointment.

Each pilot should have a practical theory of the case, including:

  • Clarity on the stream of decisions to be changed.
  • Identification of the analytical methods and at least some of the datasets that will be used to improve these decisions.
  • A simple analysis demonstrating that a sensible degree of decision improvement will create at least several million dollars per year of pre-tax operating profit gain.
  • A reliable method to measure the actual dollar value of business improvement created at the conclusion of the pilot. The measurement of value creation should be as rigorous as is consistent with the business process, and should ideally be an A/B test or other controlled experiment.

3. Minimize Change to Existing Business Processes

Any AI pilot will inevitably require some change in business processes in order to generate more profits. These changes are usually the greatest actual barrier to realizing value quickly. Process changes should be consciously kept to the minimum degree consistent with achieving a significant fraction of the available business benefit. This applies both to the operational units that are executing the target decision process and to the IT teams that support it.

The IT to support these AI applications should be lightweight, cloud-based and interact with existing corporate data stores via simple API. The AI system should also interact with preexisting operational software via API using established interfaces. These interactions should focus on directly driving decisions, rather than just providing yet more information that a human needs to analyze and process — nobody needs another dashboard, however well-intended.

With respect to data, the AI system should impose zero formatting demands on internal data systems and take data exactly as-is. It should also automatically extract and integrate external data sources, as these other data classes tend to be major sources of prediction advantage. No pilot should attempt to build or be part of a detailed IT roadmap, but if successful, it can help to influence the future roadmap.


While the current AI opportunity is enormous for almost any large business, addressing it as a senior executive can seem both daunting and exasperating. You listen to a lot of big talk, but getting answers to straightforward questions such as, “So, exactly what is this going to change, and how does that make me more money?” is surprisingly difficult. You don’t want to over-constrain a process that might be extremely valuable, but you also don’t want to give AI a pass from the need to generate clear benefits in excess of costs.

The race to capture real value from AI is officially on. If past is prologue, the difference between the winners and the losers will be how they approach the AI opportunity.

In our experience, real-world AI applications tend to consistently reuse software components in four key areas — machine learning, machine vision, natural language processing, and data integration. By working with experienced AI engineering and data science teams that have access to these core software components and an orientation to creating profits, a pilot AI software tool designed to address a specific decision challenge can be created, implemented, and have its value case proven through controlled experiments within 6–9 months.

Senior executives who have followed these guidelines have consistently succeeded in navigating the challenges of introducing AI into their large corporations. We hope they assist you in driving rapid and significant value for your organization as you consider the next step in your AI journey.

~ the team

Making AI Make Money

AI shouldn’t be a buzzword, that’s why we discuss its…

Making AI Make Money

AI shouldn’t be a buzzword, that’s why we discuss its practical, present-day applications that drive immediate, measurable, & recurring business improvements. -Publication written by members of, a technology studio that creates AI software companies w/ the Global 2000.