Navigating the Build vs. Buy Dilemma in Enterprise AI
As artificial intelligence continues to reshape the business landscape, companies face a critical decision: should they build their own AI models or purchase solutions from vendors? This choice is not merely a technical consideration but a strategic one that can significantly impact an organization's competitive edge, resource allocation, and long-term innovation capabilities. Let’s explore the key considerations and strategies for making this crucial decision.
The AI Marketplace: A Spectrum of Options
The enterprise AI market has evolved rapidly, offering a range of solutions to meet diverse business needs. These options generally fall into four categories:
Large language models from major cloud providers
Enterprise software with embedded AI capabilities
Third-party point solutions for specific use cases
Open-source models and training data
Each option presents unique advantages and challenges, catering to different business needs and technical capabilities.
The Build vs. Buy Spectrum
Contrary to popular belief, the build vs. buy decision isn’t binary. Instead, it’s a spectrum of options ranging from off-the-shelf solutions to fully custom-built models. On one end, we have plug-and-play commercial off-the-shelf solutions, offering ease of implementation but potentially lacking the customization and complexity required for high-value enterprise use cases (e.g., customer service agents). On the other end, custom-built models provide full control over data and behavior but demand significant resources and expertise to develop and maintain.
In the middle of this spectrum lies a popular compromise: acquiring a private instance of an existing large language model from a major tech provider and fine-tuning it with proprietary data. This approach balances time-to-value and maintainability with a high degree of customization and control for enterprises.
Key Considerations for Decision-Making
When evaluating AI strategies, businesses should consider four crucial factors:
Business Strategy: Prioritize what matters most to your organization - time to market, data monetization, differentiation, or attracting top talent through technological innovation.
Risk Tolerance: Assess your comfort level with external biases, liability exposure, potential AI hallucinations, and long-term compliance requirements.
Total Cost of Ownership: Look beyond initial development or purchase costs to include ongoing maintenance and operational expense within the total cost of ownership.
Data Assets: Evaluate whether your proprietary data can create a differentiated solution or a new revenue stream. If so, a more customized approach may be warranted to serve your organization’s needs / intended use cases.
These factors should be applied to each use case, as the optimal solution may vary across different applications within the same organization.
Future-Proofing in a Rapidly Evolving Landscape
Given the breakneck pace of AI advancement, true "future-proofing" may be an unrealistic goal. Instead of trying to pick winners and losers in the technology race, companies should instead adopt a toolkit approach to their AI strategy.
This approach acknowledges that the build vs. buy decision is not a one-time choice but an ongoing series of decisions that will evolve with business objectives and technological capabilities. Over time, organizations are likely to accumulate a variety of AI solutions for different use cases.
To manage the complexity that comes with different AI solutions, businesses should develop a North Star reference architecture with robust governance (e.g., AI usage policies) and security measures (e.g., data privacy controls). This framework will allow diverse AI solutions to work together cohesively, delivering business value in a safe and controlled environment within your organization.
Embracing Flexibility and Continuous Evolution
As we navigate the exciting yet uncertain future of AI, flexibility will be key for organizations. Rather than seeking a single, perfect solution, successful organizations will instead cultivate a diverse AI ecosystem that enables them to rapidly adapt to changing business needs and emerging opportunities.
By carefully considering business strategy, risk tolerance, costs, and data, companies can make informed decisions about when is the ideal moment to build, to buy, and to pursue hybrid approaches such as partnerships. This thoughtful, nuanced strategy will enable organizations to harness the full potential of AI while managing the associated risks and resources effectively.
In the end, the enterprises that will “win” with AI will be those that closely align their AI strategy with their overall business objectives, leverage their unique organizational strengths, and remain agile in the face of a rapidly evolving technology. By embracing this approach, companies can turn the build vs. buy dilemma into a strategic advantage, driving innovation, unlocking value for their organization and customers, and enabling competitive differentiation in the AI-powered future.
About the Authors:
Baris is a technology strategist in the Telecom, Media & Entertainment, and Technology industry with more than 20 years of success in helping his clients transform their businesses through data-driven strategies and smart technology investments focused on business value. He has led large scale, complex programs that focus on maximizing return on technology assets, monetizing data, and scaling AI capabilities. Baris is currently focused on unlocking the potential of AI in the TMT industry as the AI Strategic Growth Opportunity (SGO) lead and growing Deloitte’s Generative AI business through bold plays, assets, and strategic partnerships.
Rohan is a Principal with Deloitte Consulting LLP focused on growth strategy and applied artificial intelligence (AI). In this role, he partners with clients to grow businesses and customer lifetime value by launching new products and services, designing multi-channel go-to-market strategies (across marketing, sales, pricing, partners, and customer success), and developing operating models that optimize cost-of-sale and cost-to-serve. Rohan also works with executives to evaluate, adopt, and scale new technologies Generative AI, cybersecurity, and cloud engineering.
Notes:
This post is adapted from AI360 Episode 1: Build or Buy Foundation Models, which you can check out below!
In addition to our blog, you can check out AI360 on YouTube, Spotify and Apple Podcasts!