Chapter 9: A Salesforce AI Decision Guide

This chapter will present a summary of all the key architectural decisions and trade-offs that are relevant to the technologies discussed in this book. We will start by introducing the guide and how to use it, then move on to discussing how to approach the decision-making process from three perspectives:

• A functional perspective grounded in common use cases
• A structural perspective grounded in technical and non-functional requirements
• A strategic perspective grounded in the long-term evolution of a company's enterprise architecture

In this chapter, we're going to cover the following main topics:

• Using the decision guide
• Choosing the right feature based on functional factors
• Choosing the right feature based on structural factors
• Choosing the right feature based on strategic factors
• Applying the framework in practice

After completing this chapter, you will have learned how to make...

Using the decision guide

This chapter is structured as a decision-making guide to help you ask the right questions and make the right decisions, given your specific requirements. It contains three sections with factors that should be considered during the solution design phase of your AI project.

These factors will guide you based on the type of features we have considered in this book. In general, we can consider the features along a spectrum of more out-of-the-box to more custom, as shown in the following diagram:

Figure 9.1 – AI feature spectrum

At the far left, we have pure out-of-the-box features such as Email Insights, while at the far right, we have entirely Custom Code-based solutions. For this guide, we will group these into four categories: out-of-the-box, declarative customizations, third-party APIs, and custom code-based solutions.

To use this guide, make a shortlist of different solution options that you might consider for implementation...

Choosing the right feature based on functional factors

Functional factors include everything that is needed to implement the actual user experience of a given AI feature. That is both the UI, the data model, and the associated automation and code that is required for the whole process to work. In the following section, we will go through some of the key concerns that you may have, as an architect, in terms of functional factors and discuss how the different types of AI solutions stack up.

We will start by looking at functional fit, which is an obvious starting point for evaluating a feature to see if you wish to include it in your solution.

The following diagram shows where we are in the overall framework:

Figure 9.3 – Functional factors

Functional fit

By functional fit, we mean how well a given feature supports the use case that you want to apply it to. For instance, if you are considering a lead scoring requirement, you would evaluate the...

Choosing the right feature based on structural factors

Structural factors include technical and non-functional requirements that don't directly stem from the functional requirements of business users. In a broader setup, this includes things such as performance, maintainability, testability, and many more. However, in contrast to our discussion of functional factors, we will limit ourselves to items that are of special relevance to AI features.

Don't think this means you shouldn't consider performance, maintainability, or testability when designing for AI; we just don't want this guide to go on for hundreds of pages. Also, most of the considerations for these non-functional requirements are the same when you're designing for AI as when you're designing other complex features. You can see where we are in the overall framework in the following diagram:

Figure 9.4 – Structural factors

We will start by considering the need...

Choosing the right feature based on strategic factors

By strategic factors, we mean the high-level business-related considerations that we need to consider when designing an AI feature. AI projects are rarely small or inexpensive, and considering the business consequences of the decisions we're making is a necessary corrective for any conclusions we might draw from our functional or structural analysis. You can see where we are in the overall framework in the following diagram:

Figure 9.5 – Strategic factors

We will begin this analysis by considering the size of investment required to implement different types of features.

Size of investment

The initial investment covers the amount of effort and money, including licensing costs, that will be required to implement an AI feature. We will include in this the full life cycle cost of the implementation and the ongoing maintenance cost of running the solution.

Considering our feature categories...

Applying the framework in practice

The most important thing to understand about applying this framework for decision-making is that you will never have a perfect fit. There will be different considerations pointing in different directions and you need to assess which factors are the most important.

For instance, when considering a credit scoring use case, you might find that a third-party API provides you with a near-perfect functional fit and a very strong time-to-value. You might even have the skills to implement it. However, on closer examination, the model may not meet your compliance and explainability criteria and thus have to be discarded for legal reasons.

Additionally, you may run up against general architecture and security guidelines within your organization that may override the best fit based on this framework. For instance, you may have strong architecture governance against heavy customization that prohibits a heavily code-based approach, or you may be unable to...

Summary

In this chapter, we looked at 14 factors that are important in deciding what kind of AI feature to use. By going through these, you will avoid the common pitfalls that haunt many AI feature implementations and ensure that you get to the right solution for your organization.

We saw that factors will often point in different directions. For instance, you may have very detailed and complex user requirements coupled with hard compliance requirements pointing toward a custom code-based solution built on a cloud platform. However, you may not have the necessary skills to implement this, nor the requirements to be able to realize value cannot be met with a custom solution that takes all requirements into account.

While that can be dispiriting if you are an architect, that is your job. You need to make the difficult trade-offs between competing technical and business considerations that result in a solution that is, if not perfect, at least adequate to the needs of your organization...

Questions

1. What are the different kinds of factors to consider in relation to AI feature decisions?
2. Before evaluating different feature designs, what should you do?
3. Having a limited budget for implementation is relevant to what factor?