1. Bring specialists and AI experts to the same table right from the outset
IT projects in general and AI projects in particular thrive on the exchange of information. From the outset, specialists and experts on the user side and on the developer side must talk to those involved with technology and to the people with data and domain expertise. This is the only way a project team can strike a balance between what is possible from a technological viewpoint and what makes economical sense.
2. Have data
AI cannot function without data. We cannot stress this enough. A team needs data to develop models, evaluate results and obtain learning results. If there is a lack of data, those involved must first solve this problem before they can focus on other issues in the project.
3. Know your data
Getting to grips with the available data is an absolute must for AI success – and that’s before resources and budgets are poured into modelling, expertise building and equipment. There must first be clarity on matters such as availability, structure and usability. Only then can the team take their first steps towards AI development.
4. Select application scenarios deliberately
From chatbots to predictive maintenance, and from image recognition to fraud detection, there is no limit to the range of AI applications. But limits abound when it comes to resources and budgets. The main task for those responsible is to choose the right scenario out of all the options. This involves a systematic selection and evaluation process.
5. Visualise the AI possibilities
AI processes are new for many companies. There is a lack of empirical figures on which projects can be based. Before getting down to development, those involved in the project should make themselves familiar with the potential of the technologies. Regardless of whether the project involves application scenarios from your own industry or looking beyond to other sectors, the first step is to develop a feeling for what is feasible.
6. Involve top management
The effects of AI projects often stretch beyond the actual scope of the work. They require new sets of skills, affect organisational aspects and shift responsibilities. With management behind them, convinced of AI initiatives, project teams can more easily initiate and implement necessary changes.
7. Don’t relocate to a development loft
AI applications should be built where they are also needed: in companies, with the people who will also be using them. Hip development studios in trendy neighbourhoods may get media coverage. But there is a real risk that the teams in those places could develop for the showroom – and not for real life.
8. Take IT operations seriously
AI is not an island unto itself in the world of IT. The applications fulfil defined tasks within business processes. Developers must take this co-existence and interaction into account right from the outset. Regardless of whether they are interfaces, UIs or updates, only those solutions that are seamlessly integrated can contribute to the business. Individual initiatives, on the other hand, fizzle out without really leaving a mark.
9. Define responsibilities clearly
Companies collect, condense and process data, then AI applications work with it. The flow of data is not oriented towards existing departmental boundaries or reporting channels. Organisations must take this fact into account – with new job descriptions and adapted financial incentive models that do justice to the importance of data.
10. Be courageous
Trial and error, making the wrong choice, finding new solutions, binning supposedly good ideas: despite all the planning – and all the commandments – they are all part of developing AI applications. Larger organisations in particular find it difficult to grapple with this level of uncertainty. Such projects need managers who take risks and scopes that encourage this.
