Chapter 14. An AI Strategy Example

Now that we’ve covered information and considerations necessary for creating an AI strategy in detail, let’s create one using a hypothetical example. Recall that an AI strategy guided by AIPB should take the form of a solution strategy and prioritized roadmap.

Podcast Example Introduction

Let’s assume that I want to build a simple, podcast listening app that allows users to get podcast recommendations based on their preferences and past listening history. Users with unpaid accounts will be shown ads, whereas users with paid accounts will not see ads. Stakeholder-wise, I represent the business because it’s my company and app, the users are those who use the app to find and listen to podcasts, and the customers are the companies that use my app for advertising to unpaid users.

All three (business, users, customers) can have many goals for the app, and each goal could have multiple initiatives that can help achieve them. Let’s assume an AI vision exists for which this strategy is based, that defines the why, how, and what, as previously discussed. (I cover this example in greater depth in Appendix B, so we leave the vision aspect as is in this chapter in order to focus on AI strategy development.)

AIPB Strategy Phase Recap

Before diving into our discussion and development of a solution strategy and prioritized roadmap for our example, recall from Part I that all expert categories are recommended for developing an AIPB strategy. This is because developing an end-to-end, AI-based innovation strategy from the AIPB North Star through to production solution optimization may require some expertise across many functional areas of a business. All expert categories should therefore be collaboratively involved, as applicable and as needed, with oversight and management as the responsibility of the manager group of experts (especially product managers given the prioritized roadmap output).

For the AIPB methodology strategy phase and from the AIPB process categories introduced in Chapter 2, I recommend the following:

Ideation and vision development (e.g., design thinking, brainstorming, five whys)
Business and product strategy (e.g., strengths, weaknesses, opportunities, and threats [SWOT], cost-benefit analysis [CBA], Porter’s five forces, The Product-Market Fit Pyramid)
Roadmap prioritization (e.g., cost of delay, CD3, Kano model, importance versus satisfaction)
Requirements elicitation (e.g., design thinking, interviews)
Product design (e.g., design thinking, UX design, human-centered design)

Each of these recommended process categories and specific methods will be applicable to certain respective steps covered as we progress through this chapter and example.

As mentioned in the AIPB AI vision example chapter, HiPPo (highest-paid person’s opinion) methods and design by committee are not allowed, and concepts like the flipped classroom are highly recommended. Use all sessions as highly productive, effective, actual working collaboration sessions, as opposed to teaching and learning sessions. Let everyone know in advance how to get up to speed, preferably in a way that requires minimal time and effort to accommodate busy schedules.

Now let’s discuss creating the AIPB strategy phase outputs. Both are guided by an AI vision, which again defines the why, how, and what to which the strategy should be based and aligned.

Creating An AIPB Solution Strategy

Creating an AIPB solution strategy can range from simple to extensive, and involve many different disciplines and considerations. Recall that the solution strategy should define the people, processes, and resources required (i.e., the plan) to do the following:

Make your AI vision a successful reality
Execute your assessment strategy initiatives while executing your AI strategy
Iteratively execute the Five Ds

Iteratively executing the Five Ds includes answering all questions and addressing all considerations outlined in the AIPB methodology strategy phase section of Chapter 3. Refer to that chapter and section for a detailed refresher.

Our example solution strategy will be kept at a very high level given the vast depth and types of solution strategy deliverables that could be involved. Deliverables could be in the form of one or more documents, diagrams, or presentations, for example, and the relative nature of each could be business level, technical (e.g., software architecture, data models), product related, design related, data- and analytics-related, and so on.

For the purposes of this example, our high-level solution strategy provides our plan, as represented by the following list:

Develop and execute a prioritized roadmap (representing a MVP) leveraging the Five Ds process.
Conduct user research in the context of podcasts and our app vision to help guide design and development.
Take advantage of best-in-class designers and modern design methods to ensure the best UX possible.
Use an appropriate, modern tech stack and software architecture to build mobile apps for both iOS and Android devices.
Develop and integrate a new recommender engine that will be built and deployed as a cloud-based API endpoint.
Ensure proper monitoring and analytics after the apps are made available on the iTunes and Google Play app stores.
Properly assess app usability and UX, and make changes according to our findings.
Develop success metrics and a plan to evaluate them to help drive future data-driven decisions and improvements.
Develop a data feedback loop that can guide ongoing recommender engine improvements and optimizations.

Creating an AIPB Prioritized Roadmap

For creating an AIPB prioritized roadmap, my recommended approach is to create a tiered roadmap, in which each tier is aligned to the tier above it. The tiers that I recommend using are shown in Figure 14-1, and your prioritized roadmap will ultimately look something like this. A product manager should be able to help with this process.

AIPB Prioritized Roadmap Tiers and Alignment

Goals are the why, benefits, and outcomes that we’ve been talking about throughout this book, and should be given for people and business, one of the primary unique aspects of AIPB. An example of this was presented when creating our example AIPB AI vision statement in Chapter 10.

Initiatives are specific, individual strategies for achieving an aligned goal. We referred to these earlier as goal-aligned initiatives. Themes (also called epics) are high-level product features that can include many lower-level features. Both themes and individual features should be defined through discovery, ideation, and creation of resulting Agile requirements that will guide design and development work.

In this example, let’s assume that the primary goals on the business side are to increase customer acquisition, engagement, and retention. These are actually multiple goals, but we’ve rolled them into one for simplicity. We can come up with multiple aligned initiatives, themes, and features to achieve these goals.

For example, we can create preference-based, personalized incentives and promotions to encourage users to sign up for an account as opposed to using the application anonymously. In this case personalization is the initiative, while incentives and promotions are themes, each of which could have multiple specific features. We could also strategically create high-quality content (e.g., podcasts, images, copy, features) and recommendations that are personalized to each individual user. Personalization helps make things more relevant to the individual. These approaches should help achieve our goals of customer acquisition, engagement, and retention.

From the user perspective, one goal for users is to be able to quickly find new podcasts that they’re likely to enjoy. This should include podcasts that have similar characteristics (e.g., length and format) and topics as those that they already listen to, while also introducing users to new, nonsimilar podcasts. This allows users to discover new content that they’d otherwise not know about. Initiative-wise, we could accomplish this with personalization, as discussed, to achieve our business goals. This is a great example of “for people and business,” for which the same strategic initiative and outcome is a win-win for both.

Lastly, the customer’s primary goal is to produce the highest ROI possible on their mobile app ad spend. We can accomplish this through a variety of methods such as segmentation and targeting. Figure 14-2 shows a generalized visualization of a goal-aligned prioritized roadmap.

Goal and Initiative-aligned Prioritized Roadmap

Here is a more specific bulleted version of a prioritized roadmap, tailored to our example, and assume that this is a work in progress. Also assume that anything related to segmentation, personalization, and recommendations is intended to be AI-powered. Note that I’ve used more words than I normally would to describe some items (e.g., goals and initiatives) for clarity, and also that some roadmap items will need design and non-AI software development work as well since we’re integrating AI solutions into an app with a user interface and non-AI powered features.

Aligned Goals, Initiatives, Themes, and Features

Business goal: Increase customer acquisition, engagement, and retention
- Initiative 1: Personalization
  - Theme 1 Incentives
    - Feature 1: Incentivized signup
  - Theme 2: Promotions
    - Feature 1: Segmentation-based targeted promotions
    - Feature 2: Preference-based targeted promotions
User goal: Quickly find new podcasts likely to be enjoyed that are both similar to podcasts I currently listen to, and also introduces me to nonsimilar podcasts that I might not know I would like until giving it a try.
- Initiative 1: Personalization
  - Theme 1: Podcast feed
    - Feature 1: Personalized messages
    - Feature 2: Personalized images
    - Feature 3: Personalized layout
  - Theme 2: Recommendations
    - Feature 1: Recommendations based on similar users
    - Feature 2: Recommendations based on similar podcasts
    - Feature 3: Recommendations based on nonsimilar podcasts or users
Customer goal: Achieve above average ROI on mobile app ad spend
- Initiative 1: Targeting (increase likelihood that ads are relevant to each user)
  - Theme 1: Segmentation-based, targeted advertising

Notice that certain themes and features might align; that is, roll up to more than one initiative or goal. That’s perfectly fine. Requirements-wise, I recommend defining requirements at the theme and feature level, which will guide the design and development process. In the case of software, requirements should be associated with Agile and not waterfall methodologies.

Lastly, prioritizing all levels is key because you can’t accomplish everything at once (unless you have a huge team and budget). I recommend starting at the goal level—prioritize goals at each of the people and business levels and then collectively (there will be some overlap as in our example). Then, prioritize the initiatives aligned to each goal, followed by prioritizing the themes aligned to each initiative, and finally by prioritizing each feature aligned to each theme. This will result in an optimal, prioritized roadmap to drive the rest of the AIPB downstream process of build, deliver, and optimize. I recommend cost of delay, CD3, Kano model, and importance versus satisfaction methods for prioritization.

From an Agile perspective and my own experience, I find that roadmaps with nonexistent (or very loose) time ranges (for long-term planning and/or scientific innovation) or those covering at most 90 days (roughly estimated) worth of work are best. Anything beyond 90 days is likely to be highly inaccurate and will improperly set expectations (again, this is Agile, not waterfall).