The ML executive
Perspective
To make sure we are all on the same page, let’s remember that Machine Learning, a subfield of Artificial Intelligence, designs software that learns to perform computational tasks from patterns in data instead of being explicitly programmed by humans. While a couple of decades ago, companies started using robotic hardware to scale manual labor, today, companies have started using robotic software to scale mental labor (Davila-Chacon, 2019).
​
ML aims to perform cognitive tasks in a faster, cheaper, or better way than human operators — and sometimes it can even achieve all these advantages. Such automation helps companies to scale more rapidly, operate at lower costs, obtain deeper insights and more accurate results, identify better strategies, and execute them more efficiently. That being said, however, not every problem has to be solved with ML technology and knowing when it is –or is not– the right tool can save you plenty of time and resources.
​
​
| Not all ML initiatives will have
| strategic impact
​
Building and deploying ML solutions in your organization will create costs and tie up your human and financial resources. Naturally, as this needs prioritization and the buy-in of different stakeholders, you should focus on simultaneously maximizing the economic value and the strategic impact in every initiative. More concretely, directing your ML initiatives towards quality enhancements or process improvements might only yield minor benefits and lose the prioritization game. Instead, your ML initiative should always support at least one of two broader value-propositions: the first is generating new revenue streams, and the second is reducing costs. (Erichsen, 2020)
You can create new revenue streams by entering new markets. You do this, firstly, by introducing a substitute that performs the same or similar functions as the incumbent’s offerings. Secondly, by forward-integration into the buyer’s value chain—that is, by performing a subtask originally done by the buyer itself. In both cases, the product or service that generates new revenues must be justified by significant cost savings that will compensate new customers for the potential costs of switching. They will also mitigate your customers’ fear of giving away some core competencies.
​
In order to reduce costs, you should focus on productivity gains against your competitors, such as freeing skilled workers by automating preparatory or monotonous work, or on backward-integration—that is, to take over costly subtasks that were originally done by your suppliers.
​
​
| Try to excel only where it is worth it
In 2020, we coined the term Machine Learning Imperative to define the class of problems for which ML is the obliged solution. We aim to help our clients to understand that organizations should not try to excel in ML themselves when they can’t create an unfair advantage, nor should they build something themselves that won't deliver great business value to their customers. An unfair advantage can be reached if better or more training data, as well as customized models, can significantly boost the performance of conventional approaches. It can also be reached when optimizing models or applying more sophisticated ML architectures significantly lowers the operational costs.
​
Customers perceive great business value in three cases: First, if you share your cost advantages with them and lower the prices; second, if you can provide better insights or better results; and third, if you help them to automate processes more effectively or efficiently.
​
There might be exceptions to this imperative when open-source models are already good and stable enough to be used “out-of-the-box”. However, internal ML and project management resources are scarce, so from a business strategy and total-cost-of-ownership perspective companies should allocate them wisely.
​
​
​
| Full control is only guaranteed
| by ownership
We consider an internal solution as one where you define the product strategy and the roadmap by yourself, regardless of whether your internal team or external consultants build the solution. In contrast, external solutions are not owned by you, so you cannot be 100% certain that the product will be adapted to your changing needs.
Go for an internal solution when your problem has an ML Imperative and go for an external solution in the market when it can be used “out-of-the-box” and only needs a minimal effort to be integrated into your business. Both paths are distinct and have their own challenges and key success factors that we will describe in the later chapters.
​
​
| Transparency will become
| much more important
Whether the requirement of transparency comes from regulators or demanding customers, make sure that you get the following topics right at the beginning of your journey, since they will be exceedingly difficult to correct once you are in production. Ensure that the data underlying your internal or external ML solutions is of high quality and that it complies with all regulations. Make sure you document all designs and operations to enable traceability of the system's purpose, performance and results. Finally, make sure that your engineers evaluate the system for robustness, accuracy and security early on, and implement adequate risk assessment and mitigation systems appropriate to the corresponding risk classes of your use case.
​
| Understand the impact of
| Intellectual Property
Intellectual Property (IP) rights might be granted when a technical problem is solved by an innovative approach. However, the patents most worthy of consideration are the ones that tackle previously unsolved problems, or that achieve a considerable improvement over previous solutions. Even though software solutions are protected to some extent by copyright, patenting an innovative approach provides broader coverage that grants exclusivity and protects the IP holder from potential accusations of infringement.
​
As there seems to be an increasing trend to patent ML solutions, it is critical for executives to understand their industry’s IP landscape and its potential for impact or risks on a case-by-case basis. Therefore, before investing in acquisitions or strategic developments of ML technology, we encourage our clients to conduct a professional IP SWOT analysis where IP lawyers and ML engineers work hand in hand.
​
​
​