This chapter should give you a clear guidance on how to start your first project.
Rule #1: The project should be short and successful (max 3 months).
There is always skepticism for new ideas. And this skepticism or “naturally expected non-buy-in” needs to be addressed as a big risk for your first AI project. So try to find a project which can be realized within one quarter. Many companies plan from one quarter to another and you should be able to deliver good results so that a follow up project can easily be planned into the following quarter. This will keep up the momentum. Multi-quarter projects exponentially rise in complexity and you should always focus on lowering the complexity of your project wherever you can. Why? Because complexity means risk.
If this initial AI project fails, you might have lost the buy-in from key-management employees, staff and C-Level. So create a short, but tangible and successful AI project. In this way they will naturally ask for more.
Rule #2: The project has to create tangible value
This is not the time for playing around. If you want your employees and C-level to buy into the idea of artificial intelligence, you have to show them that AI can create value and not just waste money and time and ressources. So how do you create value? As discussed in chapter one, a company has three levels where it creates value:
- Voice of Customer
- Voice of Employee
- Voice of Business
Show clearly which dimension (Customer, Employee or Business) will be touched by your AI solution and show, expressed in cold cash, how much more value the new solution will deliver. I would advise to steer away from projects which might position the company better in the market or would “just” have strategic value, because this cannot be measured or clearly seen by the employees. We will describe in detail how to write an AI business case in the chapter “Process Management”, but for now I will just show you an example of how concrete your promise in value should be:
During FY 2021, the 1st Time Call Resolution Efficiency for New Customer Hardware Setup was 89% . This represents a gap of 4% from the industry standard of 93% that amounts to US $2,000,000 of an annualized cost impact. With our new AI improved call center software, we will be able to increase the Efficiency to 91% which will lower the annualized cost impact by $1,000,000.
Rule #3: Do not calculate the ROI for your first AI project.
Many people suggest to calculate ROI for your AI projects, which is great, but your first AI project should be an exception. Your first project has a big research component and many unknowns. In order to project costs, you need to know much time a project needs and how many hours people might work on them. But for your first AI project, all this projections would be vague and mostly meaningless assumptions. I do not mean to not do a rough project planning, but it will not be enough for a proper ROI projection.
Rule #4: Do not put your intern in charge of AI, because he had two courses of data science in university.
In order to deliver a successful AI project, you need an experienced team who knows data science, business, project management, data engineering and the whole machine learning pipeline. Thus work with credible external partners like Goldblum Consulting for the first project to ensure that the first project will be a success. The next chapter “How to build the best A.I. Team” will describe the roles of AI teams in detail.
Rule #5: Choose something company or industry specific
Always remember that every company or industry has its own problems and thus needs its own solutions. As a leader you can easily be the swayed to copy the solutions of others as you want to copy their success, but that might not be a good idea in general. Let me give you an example:
For this we will look at McKinsey Global Institute’s report called “The age of Analytics: competing in a data-driven world”. There, they analyzed which industries have which specific AI related problem. As it showed, the energy industry for example has a very high desire for forecasting (of energy consumption) but no desire for “radical personalization” or processing unstructured data, while the automotive industry was focusing on radical personalization and the agricultural industry on processing of unstructured data. As you see, each industry has their own focus, and within each industry, each company has its own share of problems. Thus finding your specific problem type might be a first step in brainstorming ideas for your first AI project. In the next section we will talk about eight major AI problem types.