CHAPTER 8
Conclusion
Artificial intelligence has the potential to transform all organizations. The process by which this transformation happens can vary, but the steps will tend to follow the roadmap we have listed in this book. Following all the steps outlined in the previous chapters will enable your organization to implement and excel in the use of AI technology. AI holds the key to unlocking a magnificent future where, driven by data and computers that understand our world, we will all make more informed decisions. These computers of the future will understand not just how to turn on the switches but why the switches need to be turned on. Even further, they may one day ask us if we need switches at all.
Although AI cannot solve all your organization's problems, it has the potential to completely change how business is done. It affects every sector, from manufacturing to finance, bringing about never before seen increases in efficiency. As more industries adopt and start experimenting with this technology, newer applications will be invented. AI will bring a change even more widespread and sweeping than the introduction of computing devices. It will change the way we transact, get diagnosed, perform surgeries, and drive our cars. It is already changing industrial processes, medical imaging, financial modeling, and computer vision. We are well on our way to tapping into this enormous potential, and as a result, the future holds better decision-making potential and faster, better analytics for all.
Technology has a way of making the once impossible become possible. The key is to recognize which technologies can make a difference in your organization and determine whether they are ready to be used in building real systems. This is a skill learned over time and sometimes requires prototyping to know for sure. Prototyping will help bolster innovation while keeping research costs low. Mistakes are good, as long as they are small and can be easily corrected. Implementing new technology can be a daunting task, but making consistent improvements in small increments will go a long way in keeping your organization ahead of the curve. Staying abreast of AI announcements can also pay dividends in terms of organizational optimization.
The Intelligent Business Model
After following the paths laid out in this book, you will be looking at a changed business. This business will now need a new model to understand and follow—a model that relies on making small errors, then recovering quickly, to realize sustainable and handsome returns. Such a model would imply that you might incur occasional losses as you adjust and cope with newer strategies. Regardless, you will still be innovating, growing, and marching into the future at full speed. Applying Agile frameworks to your organization will beckon growth and large improvements in efficiency while keeping errors and costs low.
This new business model is focused on innovation and growth. With artificial intelligence helping you choose the best actions for your company in this forever unpredictable and uncertain world, you will have a mighty ally as you grow. AI will help us all overcome the drawbacks of the human decision makers of yesterday, who were able to hold only a few variables in mind at a time. This is not to say that AI will replace all human decision making, but rather that it will assist its human counterparts in arriving at sounder, safer, and more sustainable conclusions.
The Recap
The process of AI implementation laid out in the book is a malleable one. It is not written in stone or intended to be followed to the letter. The process must be customized to suit the needs of your organization. Customization can take the form of skipping a step or adding new ones based on your perception of how your organization operates. A blind copy-paste application of the steps listed in this book would probably give your organization a culture shock and hinder growth rather than supporting it. Do your best to avoid that.
All organizations are different, and their methods of operation are different as well. There can never be a one-size-fits-all implementation methodology for bringing changes on a massive scale. Such changes require careful comparison of the present state of the organization to the expected state, identifying the deltas and then figuring out how best to modify the process to better suit your needs. There may be areas where a simple copy-paste could work, like the implementation of an idea bank, but even there you will need to figure out how best to organize ideas, arrange review meetings, and store your idea bank to ensure that this process does not become a drag on employees. It is important that everyone in your organization understands and is able to implement changes correctly.
Keeping in mind its flexible nature, a short summary of the entire process follows.
Step 1: Ideation
Every project starts with an idea. It is the same for projects involving AI. Bringing ideas to the forefront of your organization will involve inculcating a culture of innovation. With innovation-based focus groups and idea banks, your organization will be equipped with the best tools to start generating new ideas on a consistent basis, increasing your chances of being the first mover. At this stage, it is also ideal to study available technology to learn the challenges involved. Keeping an open mind and ensuring that all ideas receive an audience before being adjudicated upon will go a long way in fostering a collaborative, innovative, and support environment where good new ideas regularly emerge.
Step 2: Defining the Project
With the ideas fleshed out and chosen for implementation, it is time to break an idea into actionable steps. Three ways to draw out a project plan from an idea are design thinking, system analysis, and the Delphi method. Design thinking is about abstract ideas, finding a wholly new way of looking at existing processes. Systems thinking will help you improve the current systems, whereas the Delphi method is most suited to areas where expert opinion would be required. Using the incorrect technique can give you a bad project plan or doom the project entirely, though the techniques are not always mutually exclusive and it may suit your project to use them in tandem. Measurement criteria for the project should be decided at this stage. Without criteria for success, it is impossible to know what went right or wrong, because no concept of right and wrong will have been established.
Step 3: Data Curation and Governance
An AI system's prime component is data. Without data training and testing, the AI system will be useless. Data can be gathered from internal as well external sources. Internally sourced data will have the lowest hassle and should be able to be used readily. Internal data might need to be digitized if not already stored within computers. Only the data with perceived value should be converted to digital. As you start gathering larger and larger datasets, it becomes necessary to establish data governance procedures if they are not already implemented. Especially since the advent of GDPR regulations in the European Union, harnessing user data now requires establishing governance procedures before the data is collected. Care should be taken to stick to positive data collection techniques to avoid problems of legality or goodwill.
Step 4: Prototyping
Prototyping is the development of small iterations of the project plan, which can be used to demonstrate the project's early value. A prototype must necessarily be functional. Broken software with “pass” statements inside functions are not real prototypes. Before building a prototype, you should look at existing solutions available in the market, since reusing code is generally cheaper and therefore less risky. Defining a logical architecture diagram would be the first step to developing a prototype. At this juncture, some big decisions need to be made that will have a huge impact on the final outcome: technology selection, which programming language to use, cloud APIs, and microservices. Prototypes should be developed with stakeholder involvement using Agile methodologies.
Step 5: Production
After the prototype is showing value, it is time to scale it up and complete the system for release to end users. A hybrid approach should be adopted for reusing the code from the prototype in the production release, but that should only be done after organizing the code to make it easier to maintain. Automated testing and continuous integration should be implemented before the project is released to the users. Continuous integration techniques help avoid code conflicts, and automated tests alert you to problems before they are pushed out to the users. Implementing a continuous integration pipeline will provide the project with higher-quality code and better software overall. In production environments, hybrid AI systems that use humans to fill in gaps where the AI algorithm fails will give you augmented intelligence. Production environments need to be scalable to handle user loads without failing the system completely or making it unusable. Deploying in the cloud with computing services like IBM, AWS, Google Cloud Platform, Microsoft Azure, and other providers offers solutions to the scale problem with limited effort.
Repeat: Thriving with the AI Lifecycle
The project does not end upon its release. Bugs will need to be swatted. User feedback will need to be incorporated. Gathering user feedback via forms built into the software as well as surveys will help you to identify flaws that need to be resolved. Increasing the intelligence of the system by providing it with newer data and examples is another task that is required to keep the project in a usable state.
A review of your idea bank on a regular basis will help you to spot new opportunities. Based on the resources collected in the project (and resources collected in general), a knowledge base should be implemented in your organization and updated regularly, since it will aid all future projects. A model library implemented as part of your knowledge base will aid any long-term AI projects and their adoption. Open-sourcing the solution (or pieces of it) will invite the community to help you. Finally, ensuring that data is fresh and regularly updated will keep your models chugging along nicely. Outdated models will give bad predictions and might be detrimental to your business.
So What Are You Waiting For?
This book has covered the journey of adopting AI technologies. From the initial ideation to assembling user stories and available data, to implementing a prototype and then a final production system, your journey has been thoroughly mapped out.
Some of you may be just starting out on your journey whereas others may already be in the final lifecycle stage of your AI system. Even if you have already built your first AI system, there is value in following these adoption methods for future projects or for major system revamps. The roadmap provides a structure—a checklist of what must be accomplished along the way. We wish you and your organization success in making the future a reality today.