After you have identified the available resources, data, people, skills, and the potential areas suitable for AI deployment, specific tasks, processes, or jobs, you stand in front of three major choices; to buy a pre-made solution, to build your own AI system, or a combination of the two. One, buy ready-made AI. There is an increasing number of companies that specialize in building customizable AI solutions for specific domains. In addition, the big tech giants, IBM, Oracle, Salesforce, Microsoft, and many more, are proposing AI dedicated solutions that fit with their other software packages. This high availability of ready-made AI tech should make it easy for an organization to jump on the AI train and introduce some Automation quickly. However, before committing to the purchase of new and probably rather expensive tech, any organization must answer some questions. The first and most important question is, which problem is the new AI system intended to solve? This goes back to one of the original points made this week. Mainly, that we should avoid AI for AI's sake. The technology needs to have a purpose and a specific one at that. If a specific problem is suitable for automation, the purchase decision will include a list of features that a ready-made system will deploy to solve the identified issue. However, no AI can successfully solve a problem in the absence of good data. The second question to ask before purchasing an AI solution is, which data is required by the system? Does my organization have already this data? If not, how can we get it? Supposing the organization already needs the technical requirements of the new system, the next data related question is, what happens with the output of the AI tech? Where will this data feed into? How will the data insights be used to maximize its utility? Where will the data be stored and who else will have access to it? For example, would the company that provides the AI tech, have access to the output generated by your organization. Two, build your own AI. If buying AI solutions was the quick and maybe less expensive option, building your own AI system is the opposite. It is more costly and requires a bigger investment in terms of personnel and data, but it also is likely to give the best return on investment and it is entirely fitting your context. Your organization needs to be AI ready or to have reached AI maturity for this to work. But if your evaluation of internal resources and processes shows this degree of maturity, then you are ready to give it a try. The one question to ask is, how necessary is it to build your own systems? Can you achieve almost the same results by buying the ready-made software and/or can you perhaps combine ready-made and in-house solutions? Three, build on existing solutions. This is the middle path in terms of cost, as it takes out-of-the-box solutions and attaches them to in-house software. Which does not have to be AI, but regular digital services would be enough. Some of the major players in providing these bricks that can construct modular solutions are Microsoft and Amazon, AWS, both of who offer services such as recommendations, forecasting, image and video analysis, advanced text analysis, document analysis, and many more. An example of pre-made software that can be integrated in various architectures use speech recognition or aspects of computer visions, such as face recognition or image recognition. These functionalities have many different applications. Speech recognition can build interfaces for a wide variety of customer services. Age recognition can be used in other machines to make age sensitive purchases, such as tobacco. The biggest challenge is to reach the best combination between the external and the internal Solutions which need to talk to each other and together provide one integrated solution.