The Buy vs Build Decision for Enterprise AI

The Decision Should be Tailored to the Needs of Each Organization

Roughly half of all businesses are buying off-the-shelf AI systems and products today, and that trend is expanding as products become more refined and efficient, but it’s a bit of a minefield with a fair amount of fear, uncertainty, doubt and desperation, so I thought we should dive into the decision in this edition of the Enterprise AI newsletter.

Let’s start with a few rules of thumb in the shallow end and before jumping off for a deeper dive with questions on the big decision, followed by our approach at KYield.

• Everyone has conflicts, particularly in EAI, including vendors, consultants, and internal teams. Even some CEOs have conflicting incentives. As I’ve shared several times in LinkedIn posts, beyond competency, conflicts are the greatest cause of investment failures in AI. Managing conflicts well is therefore essential for EAI investments.

• The larger and more complex a company is, the more custom AI work will likely be warranted. However, even the largest and most complex organizations should be blending off-the-shelf systems with custom work. No organization has everything they need in-house. For small business, few can even consider custom AI due to the cost of talent, and it’s not an option for most mid-markets either. A few exceptions exist in special situations, such as smaller companies that offer AI products that require custom work.

• A good starting point (or restart) is a disciplined assessment process, which should be continuously monitored and reconsidered on an annual basis for most. AI introduces more strategic and operational issues than any other technology in generations, if ever. Moreover, it’s still full of uncertainties and risk, so it’s a moving target.

Five general questions to consider

1. How would a SCOTUS decision on copyright impact your AI plan?

2. Should your business consider developing your own AI product?

3. Is the business under serious risk from competitors adopting AI?

4. Will the existing data quality and structure require replacement?

5. Is your business a good candidate for a strategic partnership?

Five buy questions to consider

1. Do the AI vendors under consideration have sustainable business models?

2. Does the off-the-shelf product offer interoperable software and data?

3. Can the AI systems under consideration provide a competitive advantage?

4. Do they offer a good foundation from which to build new value?

5. Do the leaders of the company have integrity? (are they trustworthy?)

Five build questions to consider

1. Does the enterprise have the resources to build custom AI?

2. Is it essential for the organization to become an AI leader to survive?

3. Can the team convert the AI strategy to an efficient system design?

4. What is the percentage of redundancy in the custom build design?

5. Will a blended buy/build be optimal for the specific situation?

We could dedicate an edition of this newsletter to each of the above questions. Teams in most organizations should tailor questions as needed during the assessment process. Industry-specific questions may represent the majority for some businesses, for example, and those who pursue a custom build or blended path will need to plan infrastructure carefully. And of course don't forget to review for regulatory compliance and change as needed.

Given the stage of maturity and understanding in EAI, it makes sense for most large enterprises to pursue a blended EAI system design. Depending on the situation, some upper mid-market businesses may also find it optimal to adopt a blended system design, primarily to provide additional competitive advantage on top of vendor efficiency.

Software economics haven’t changed much with AI—it’s still much more efficient to spread costs across many customers than it is for each enterprise to build one-off custom systems, unless there is an identifiable reason. The primary exceptions are companies who have sufficient resources and need a unique product offering. Few SMEs are in a position to consider custom AI, unless they are native AI companies with their own technology like our company (deep-tech ventures don’t resemble SMEs other than in size).

On the importance of talent and a systems perspective

It’s surprisingly challenging for some organizations to think about enterprise AI as a refined system, but that’s precisely the lens they should be looking through. Refined systems are necessary to be competitive, whether acquired, built or blended.

I recently posted a note in response to the recent flurry of news about record compensation offers in the AI talent war. One of the ten suggestions is to prioritize AI product development, which is an important consideration in the build vs buy decision.

“The biggest single overlooked opportunity in AI by far is new products, and that's due almost entirely to the lack of understanding and expertise (e.g., Apple - $10s of billions at stake in rev).”

Although the talent war may be waged in an unwise manner, talent is rare. We have a strange phenomenon occurring where large numbers of people experiment with LLM chatbots and then think they understand AI, when actually very few understand how to design and build AI systems (not be confused with executing GenAI apps).

It’s taken me nearly 30 years of applied and theoretical work to invent, design and refine two AI systems from the theorem stage, and I was considered a fast learner. Even the most brilliant specialists typically require a decade or more of intense study combined with hands-on experience to become competitive. Suffice to say that no one should underestimate the multi-disciplinary learning curve in AI—it’s steep, long and expensive.

Our approach for the KOS

The KOS is an enterprise-wide organizational operating system enhanced by multiple types of technologies under the AI umbrella. The most accurate technical description is a human-centric neurosymbolic AI system, as it’s a hybrid of rules-based symbolic AI and neural network algorithms.

Central to the KOS is our patented AI system core that includes an adaptive data management system (ADMS). The ADMS essentially enables end-to-end data management for the digital workplace, and includes strong security and governance that can be easily tailored by each organization and individual in natural language. The KOS ADMS is laser focused on creating and maintaining high quality data, which enables much more accurate and secure AI functions as well as increased productivity.

We recommend that every employee in the organization install and use DANA, our AI assistant, which is tailored to each individual in a semi-automated manner (each person can tailor to their needs). DANA includes eight default functions for increased learning and productivity. The goal of the KOS is to achieve a continuously adaptive learning organization (CALO). A CALO is an organization that not only learns efficiently, but executes on that knowledge in an adaptive and timely manner.

With regard to the buy vs build decision, the KOS is focused on human work in the digital work environment, not everything in the enterprise. It’s necessary to provide end-to-end precision data management in order to be competitive or achieve a CALO, hence our focus. Custom AI can be either built on top of the KOS or integrated with our APIs.

For example, we have worked extensively on industry-specific versions of the KOS that require integration with vertical software. We believe the KOS is a good example of a blended AI system that provides the foundation for human workflow and governance of enterprise AI. One option we've collaborated on with customers in multiple industries is to build custom AI apps on top of the KOS. A strong, secure, adaptive foundation provides many interesting and potentially very valuable options.

Related reading recommendations:

• Exceptional article by Deborah Perry Piscione essentially about how to invest in enterprise AI systems (HBR).

• “When It Comes to AI, It Still Comes Down to Make or Buy”, by Dr. Erik Linstead.

• “The Evolution of Build Vs Buy in Artificial Intelligence”, by Leanne Allen (KPMG UK).