When companies want to use artificial intelligence (AI), the question arises: make or buy? A systemic approach is required. […]
From marketing to customer service to production control: companies are using AI applications in more and more business areas. In this way, they ensure, among other things, more efficient processes and relieve employees. In a recent study by consulting firm Deliotte, 79 percent of respondents describe AI technology as very important and critical to success. It is particularly effective in the IT sector.
The increasing prevalence and the wide range of possible applications pose an important question for companies: Is it worthwhile to develop AI applications independently? Or is it more profitable to acquire an external out-of-the-box solution?
In the rarest of cases, a company is able to handle the implementation of AI applications and the associated challenges from scratch on its own. In the best case, it should not have to face this task alone. Make-or-buy decisions should be approached systematically.
So far, however, most companies have neglected such an approach. It becomes critical especially if you leave this decision to a team that treats the acquisition and implementation of AI applications like the purchase of conventional software.
In order to be able to systematically approach the make-or-buy decision, companies should know two important factors. First, you don’t have to make a single decision for each use case for AI. Rather, each use case consists of several different levels: the actual application reflects the top level and consists of machine learning functions, data assets and the infrastructure.
Since the levels are interrelated and interdependent, companies must accurately assess at each level to what extent they influence the make-or-buy decision. For example, you should question the in-house development of an AI application if the data basis or the entire infrastructure is deficient.
In fact, the question of “make-or-buy” is not a binary one, in which the answer in the end is an unambiguous “develop yourself” or “buy”. Rather, it is about stepping between the two extremes. To what extent should in-house developers build AI applications themselves? Which components should companies leave to partners or buy in addition?
In some cases, companies will decide to develop the basic framework of the application independently. Nevertheless, you may use pre-trained machine learning models (ML) in order not to have to train your own models. On the other hand, external AI solutions usually use their own data, which requires efficient data exchange.
The make-or-buy decision can therefore be roughly distinguished into three approaches:
- the complete in-house development of AI applications including ML models;
- a hybrid approach;
- the purchase of an external out-of-the-box solution.
As soon as a company has been able to identify a suitable use case, those responsible should consider the make-or-buy decision against the background of various criteria. A first factor is the question of the strategic value of the use case. To what extent does the use of AI promote strategic positioning? Only when the strategic added value for a use case is high, companies should consider the independent development of AI applications. In the course of this, in a use case with a high added value, it is necessary to discuss whether companies have to develop (and own) the AI model themselves in order to maintain this strategic value in the long term.
If companies come to the decision that they do not necessarily have to own the AI model for value retention, they should question whether they have a specific advantage over competitors when they develop their own application. These advantages can be, for example, a unique data basis or subject-specific competencies.
Companies should also evaluate whether the performance of external solutions can meet the requirements of the use case at all and whether they are economical.
A final decisive factor is the total cost of ownership. The costs of development, deployment and maintenance also play a significant role in the make-or-buy decision.
The decision to “buy AI applications or develop them yourself” is not one that is comparable to the purchase and implementation of conventional IT solutions. Therefore, it is essential that companies systematically approach this decision-making process, consider each use case in its individual levels against the background of the make-or-buy decision and include the various influencing factors in their decision-making.
Every company has different requirements for potential AI use cases. That is why it is essential to develop your own AI strategy, in which the make-or-buy decision is an essential component.
*Phillip Hartmann is Director of AI Strategy at Applied AI. Prior to that, he worked as a strategy consultant at McKinsey & Company for four years and completed his doctorate at the Technical University of Munich on competitive factors in the use of artificial intelligence.