The what and how when outsourcing your own intelligence
Utilizing and enabling Artificial Intelligence (AI) means that you, in one way or the other, augment or replace the human intelligence with that of a computer. According to Gartner’s CIO survey, almost 50 percent of the companies are either implementing AI or are planning to implement it.
Regardless of what type of AI solution that is considered, the computer power will either replace what a human used to do or enable companies to improve what was done with only human intelligence. The hype about these AI related technologies is well justified due to the fact that opportunities are virtually endless. However, as with all disruptive technologies, AI also means significant changes to the ways of working as well as major risks, such less control of IPRs, poor data quality etc. Therefore, a company investing in AI needs to carefully consider e.g. vendor selection, IPR handling and even future exit plans.
In any company, introduction of AI is likely to impact several different business processes, such as those regarding IT and HR. Other aspects are the implications related to partners and sourcing. Due to digitalisation, there is a constant move from “human capital” to” IT capital”, and with AI the move will be about the brain rather than the hands. The AI will learn and even develop intellectual properties which was previously developed by the staff or consultants. When you outsource to consultants today, traditionally you have a clear strategy on how and why you want to use consultants and how IPRs should be handled. A corresponding strategy regarding IPR handling can be applied when sourcing AI. The more advanced and business central the AI services are, e.g. deep learning or predictive analytics, the more crucial it becomes to get these services right.
There are several ways of sourcing AI and in our view, an iterative an agile process may generate the best result. For simplicity, the steps below are aligned with the IT Standard for Business framework, but regardless of your ways of developing and acquiring new services, there are a few additional aspects to consider related to AI.
- Defining a Vision. A study published by Harvard Business Review concludes: “the gap for most companies isn’t that machine learning doesn’t work, but that they struggle to actually use it.” You are not sourcing AI technologies. You are sourcing a technology to fulfil a business need and most likely it will require a change project to make it work. The starting point is to understand and define the business benefits that you want to address. Needless to say, you also need to consider that the AI solution is just a technical component of a complete change project and that the entire project needs to be managed.
- Trying out the concept. Many AI solutions may be so fundamentally different from what many stakeholders in the business unit are used to, so trying out the concept may be needed to understand the possibilities and not limiting the vision. For many people, you may not even know that you have a problem until a solution for it becomes available.
- Defining the requirements. The focus is on the business requirements and what you want to achieve and how it should be achieved.
- First, looking at your business needs. It is very tempting to look at AI as a way to achieve quick and hard cost reductions by replacing people. However Gartner recommends to initially focus on soft benefits such as process improvements and customer satisfaction, as well as augmenting people instead of replacing them. The business need will drive the selection of technology, not the other way around.
- Second, you need to consider whether you have the ability to utilize AI. An AI system needs to be trained based on data. If the data is not accurate, the model will not be correct. So, you have to carefully consider the data accuracy before you start implementing AI. Or as Katrin Ribant, co-founder of Datorama (now Salesforce) stated “AI is only as good as your data and goals allow it to be”.
- Third, you must consider how AI and Robotics will impact your existing IT costs, especially regarding SW licenses and terms. Are there any restrictions in your current license agreements? Will you require more or fewer licenses to cater for any AI systems or robotics installations?
- Fourth, you need to define your strategy regarding traceability and transparency. When a human recommends a decision, you usually ask “Why?”. Likewise, you need to consider what kind of transparency and traceability you need for your AI solutions.
- Fifth, you need to consider how the service will be delivered. It can be delivered in many ways where the level of sharing is changing. It can be a fully SaaS type of delivery where all the assets are shared, including the learnings. Or the system can be on premise and fully under control. The more open it is for sharing, the lower the cost tend to be and the faster the system will learn. On the other hand, the less control and ownership you will have. It is essential to establish a proper assessment of what is core for the company. For example, translation services may not be considered core while others e.g. predictive analytics regarding factory utilisation might be.
Figure 1. With cloud solutions, you will gain efficiency, but lose control. You need to consider what is right for you.
- Qualifying the partners and deciding the preferred one(s). The more dependent you will be on a supplier, the more careful you have to be in the selection process. And when you are transferring a company’s core assets – the intelligence – to a third party, you are definitely creating a strong dependency. Once you have started to train the system, you are creating a strong lock in effect.
- A contract for an AI based solution is to a large extent the same as for other software and services contracts.) However, there are a few things to pay attention to in particular:
- The SW will in fact develop its own intellectual properties or algorithms. It is essential to settle the IPR question.
- Closely related to the IPR issue, there are the exit clauses and procedure. When you enter a contract you may not think of exiting it, but with AI system contracts it is absolutely essential. What if essential knowledge is stored as algorithms in an AI system and you have no way of retaining it if you would change supplier?
Finally, the contract signing is just the beginning of the relationship. Since you are likely to have a high dependency on the supplier, you also have to make sure that you have the proper vendor management and governance framework in place.