As the final project of my UC Berkeley School of Information course in Artificial Intelligence strategies, we had to submit an AI strategy canvas. Today, I will attempt to share important aspects of the canvas, so you and your team may have a template to work from and consider. Let’s begin. We will look at both a strategy and operations perspective both internally and externally. Part one will include Internal and Operational Strategies, while Part II will examine external processes, threats, and procedures.

Urgency: Strategy, Internal – It’s now or never to employ AI strategies for your organization. To stay competitive in the trajectory of digital transformation we are experiencing today, an AI roadmap is essential. Within this roadmap is your strategy. The consequences of not developing such a system will make your organization obsolete. Attempting an AI strategy involves 100% buy-in from all stakeholders and, more importantly, a mindset to take small steps, fully understand the questions you want AI to pursue, and the likelihood of completing that initial goal.

Requirements or Needs: Strategy, Internal – First is a commitment by all stakeholders and the recognition that AI is a fundamental cornerstone for discovering more about your company and, more importantly, about what your clients require. The second is financing. What is your budget for AI, and can you predict or show proof of ROI from a unique implementation? You can use data modeling from cohort companies to extract the necessary data to complete your ROI model. Third, are subject matter experts. There are severe challenges in finding, retaining, and managing top AI talent. It is undoubtedly true that companies such as Google, Amazon, Microsoft, Facebook, Tencent, Alibaba, and Baidu have budgets for AI talent beyond the reach of most organizations. You’ll need subject matter experts in your field and data scientists who can answer the tough questions about your data.

Identify Goals: Strategy, Internal – The most important goal for a company in committing to an AI strategy is that everyone associated with the project must all be dedicated to the implementation. One dissenter can rake havoc on the project. Pick one achievable goal to start and build momentum from there. Reduce Churn by 10% monthly, for example? Increase conversational conversions by 25%. Don’t put all your eggs in one basket. You will want to use a “portfolio approach” as you implement new AI strategies. Further, the utilization of SMART goal methodology. Make the goals, Specific, Measurable, Attainable, Relatable, and Time-sensitive.

Approach: Strategy, Internal – Begin by building value and predicting ROI for the AI strategy. First, start by training initial data sets, using a segment of your audience, opening it up from there, building on this approach, and double down when it is successful. Small wins at the outset and scale across the company as you gain momentum… OK, now that we know conversational commerce is working at your B2B marketing site, and conversions have increased by 25%, you can expand horizontally to other channels. You might want to broaden the chatbot recommendation engine to social. Perhaps your social media channel needs a lift in conversational commerce, develop that angle, and employ the same strategies you used on your site chatbot.

Data: Internal, ResourcesThere is a vast universe of data available to set the groundwork for AI implementation, including but not limited to academic, opensource, governmental, and corporate data sets, which most can be accessed from your cloud provider. There are public and private data sets and data that you might likely have to pay for. When modeling, you might include data sets that will complement your internal data set, which is structured or even unstructured. When introducing third-party data sets, make sure it is cleaned and tagged. Most, if not all, of the data sets, come from cloud computing solutions, like Amazon, Google, and Microsoft Azure. Probably crucial that if you are contemplating moving from lakes to warehouses, the centralization of data makes good sense. This should be a primary discussion point before introducing any machine learning concepts or data. The idea of focusing and concentrating AI techniques makes good sense. Keep in mind that you require the governance of these data sets immediately. Privacy will be discussed in Part two.

Talent: Internal Resources – Large Data Sets are hard to manage, but getting more accessible with applications like ObviouslyAI a pretty cool startup. Since there is a distinct shortage of data science experts and costs to hire and manage and, more importantly, keep them are very high. But if you manage to have kept them for “some time,” and they are paying dividends, the acquihire opportunity presents itself in the form of valuation for your company. Basically, that means “buyers” of your app will pay up for the right experts. As an AI strategist, you don’t have to be a data scientist to understand your corporation’s needs through an AI strategy; you just have to understand the methodology, and more importantly, how the data will be interpreted to stakeholders. There is a trade-off between time and accuracy and size of the data set. If you are lucky enough to find one recognized subject matter expert and combine his/her heuristic knowledge with data science experts, you’re in good shape. It turns out that experience is powerful when a system knows nothing. Subject matter experts are usually less biased, providing the dataset to be introduced more worthy. The familiarity around AI can come from a diverse set of backgrounds. They are centralized or decentralized. Pioneering companies like SnowFlake, for example, offer increasingly powerful AI cloud products, talent pools for the automated AI space, which results in ample opportunities for organizations to thrive, not only through cutting edge AI research but strategically using the available tools and products while remaining focused on business goals and core competencies.

Time: Internal Resources – Time is a critical factor when implementing your AI strategy. You’ll need time to create the plan—usually 30-45 days. Then use a subset of your traffic for six months to determine the validity of your AI strategy. Time is one of your SMART Goals. Each AI process must have a time element before implementation. If you see engagement and incremental lift in this time frame, you are on the right path to success. Suppose you’ve used the time wisely to create a solid foundation of your AI strategy, collecting, tagging, and organizing your data. In that case, the initial data set will be much more useful, and you have generated an excellent starting point.

Technology: Internal Resources – Pioneering companies, like Amazon, Google, and MSFT, offer increasingly powerful products in the automated AI space. You can access their datasets from within AWS, BigQuery, Azure, etc. You’ll have to determine which AI strategy to deploy. There are five subsets: including but not limited to ML, Robotics, and ComputerVision. Datasets and technology are available to anyone who wants to implement an AI strategy. The know-how and the talent pool is what is scarce.

Value Proposition: Anything AI-related should have an impact on consumer behavior. One question you might want to ask is, how do I measure consumer behavior? How often are they engaging with my app, are my DAUs seeing m/m lift, are my users spending more time on the app, etc.? So, the critical factor would be customer engagement. Monitor the delta in the process and, just as importantly, monitor the delta in time to satisfaction. Did you solve the customer problems in a timely matter? The delta has to be measured by total lift. Then through customer satisfaction surveys, you’ll prove that the AI strategy was worthy based on a higher NPS score. How can we do better?