Making the AI Make-or-Buy Decision

How to find out what to build and what to buy.

Hey there,

I spent the first 4 months of my data science career working on a machine learning model that nobody needed. This experience taught me the importance of the make-or-buy decision in AI.

The reality is that you often don't know if you actually need a solution until you try it. What sounds promising on paper can flop with real users.

Still, many companies - even smaller ones - believe they need to be in full control over the solution right from the start. But there are other options.

Let's explore how to pick the best for your business!

It's Not a Binary Choice

The decision to make or buy an AI solution isn't simply black or white.

Instead, picture it as a continuum with "fully in-house development" on one end and "fully purchased solution" on the other.

Between these two endpoints lies a vast landscape of options. For instance, some companies choose a mix, where they use pre-built components or models from vendors but develop specific parts of the application in-house.

This hybrid approach allows for more flexibility and control. Of course, you would need a little more in-house talent for this.

By understanding these different options, you can find the right balance for your needs and benefit from AI without having to start from scratch.

Key Decision Factors

To determine the right balance, consider these key factors:

  • Use case development stage

  • Strategic value

  • External solution performance

  • Internal resources

Let's walk through them one-by-one.

Use Case Stage

There are two main stages in developing AI use cases:

The prototype, which often includes a proof of concept (POC) or minimum viable product (MVP), is about quickly testing feasibility and potential impact. The goal is to learn fast and fail fast. Prototypes are typically not integrated with your IT environment.

Once a use case proves its potential value from the prototype (typically less than 30% do), it moves to production. This is when you typically integrate your AI into your existing environment and processes, scaling it across your organization.

Consider this a whole different game.

When you're ordering pizza for the prototype team, it's time to order the big buffet for the production!

Strategic Value

Strategic value is how important and impactful the AI solution is to your overarching business goals.

Consider the example of autonomous driving for a company like Tesla. This use case alone has the potential to make or break the whole existence of the company. Therefore, it has a super high strategic value.

On the other hand, consider a document text extraction use case for an insurance company. Sure, this use case might be valuable, but the business will also probably survive without it.

The importance of an AI use case can change over time. Something that was initially seen as very important may become less important as the market changes. On the other hand, some AI projects that started small can become very valuable.

For example, the evolution of chatbots went from simple customer service tools to essential components of the customer experience.

External Solution Performance

No off-the-shelf solution will be a perfect fit. But the question is, how close does it come to meeting your needs?

It's essential to remember that the "good enough" threshold can vary.

Take, for example, an AI service that categorizes customer support tickets. If it offers a 75% accuracy rate, it might not seem impressive at first glance. However, if your manual process only achieved 60% accuracy before, then the AI solution represents a significant improvement.

And for some applications, even a small improvement in accuracy can lead to significant operational benefits or enhanced customer satisfaction.

Internal Resources

This refers to available technical resources such as data, infrastructure, and skilled teams. It's not just about whether you have the resources you need, but how effectively and efficiently you can use them to build the solutions you want. You may have a room full of developers, but they're all working on other projects.

While the decision factors above are the most critical ones, there's of course much more to consider - especially for larger enterprises: Factors like organizational readiness, cultural fit, regulatory compliance, and ethical considerations can also play a huge role in the adoption and implementation of AI systems.

For now, let's concentrate on our smaller subset and learn how to make decisions between "making or buying" in different scenarios.

The following flowchart helps you visualize this process:

Scenario 1: Prototyping a Low Strategic Value Use Case

Here the decision is simple - go with an off-the-shelf solution. Even if you have internal resources available, they are better spent on high strategic value use cases. Remember, your main goal in the prototype phase is to quickly test and validate ideas, not to reinvent the wheel.

Recommendation: Buy

Scenario 2: Prototyping a High Strategic Value Use Case

For high strategic value prototypes, first check if an off-the-shelf solution meets your needs. If so, use it to test feasibility and user feedback quickly. (Recommendation: Buy)

But what if none of the vendor solutions are good enough? Check if you have the resources to build a hybrid solution internally. For instance, you can build the main ML model yourself and use third-party components for the rest, or vice versa. (Recommendation: Hybrid)

If there are no internal resources, building a prototype is a tough call.

If the use case really has super high strategic value, you might consider building up some resources. This might require investments in infrastructure, data or hiring (temporary) staff. However, these things typically come with a lot of complexity and it's recommended to critically gauge the potential return on investment before making such a commitment. (Recommendation: Think Twice)

Scenario 3: Productionizing a Low Strategic Value Use Case

If you've successfully prototyped a "quick win" use case, it makes sense to keep using or build on the vendor solution for production if it meets your needs.

For instance, if you have created an AI prototype to automate manual data entry and discover that a ready-made software fulfills your needs and can be smoothly added to your current systems, it would most often be cheaper and faster to keep using the vendor solution for production.

You can always switch to building it yourself later.

Remember that every AI service needs ongoing maintenance and support.

And even if you need customization, you can still just rebuild parts of the workflow and adapt as needed.

For example, if you were prototyping using a chatbot service from a US vendor but want to switch to Europe for production, you could check if you can consume the core component (the AI model) as a service hosted in Europe and build the rest of the application on your own infrastructure using available open source or no code tools.

Recommendation: Hybrid or Buy

Scenario 4: Productionizing a High Strategic Value Use Case

First of all, if you managed to move a use case with high strategic value from prototype to production, then congrats!

You've already crossed a major hurdle.

If the use case is really paramount to your business it almost always makes sense to allocate some dedicated resources to it and build this use case yourself.

After all, this is what will differentiate you from your competitors and add significant long-term value to your business.

Recommendation: Make


From our scenarios, it's clear that the 'Make' option is usually not the best choice.

Your company's AI efforts will likely only produce a few strong use cases that are crucial for your competitiveness. It's okay to build those internally with a dedicated team.

But for most use cases, don't get too caught up in the details.

An AI-progressive company doesn't have a handful of AI use cases in production, it has hundreds. And most of them are small, low-impact. But their value adds up.

Especially if you’re a non-technical small or medium-sized business, the majority of your AI use cases should be handled with off-the-shelf or hybrid solutions. This allows you to focus on the core aspects of your business.

It's best to start small and seek expert help at the beginning.

Remember, the more you try, the more you learn.

So keep learning and see you next Friday!


Want to learn more?

  1. Book a meeting: Let's find out how I can help you over a coffee chat (use code FREEFLOW to book the call for free).

  2. Read my book: Improve your AI/ML skills and apply them to real-world use cases with AI-Powered Business Intelligence (O'Reilly).

  3. Follow me: I regularly share free content on LinkedIn and Twitter.