As generative AI evolves into agentic AI, the build-or-buy decision has become increasingly complex, hinging on factors such as business size, use cases, and strategic priorities.

Enterprises and small to mid-sized businesses (SMBs) have long grappled with a fundamental question: whether to build AI capabilities in-house or purchase them from a vendor. This decision shapes not only how organizations adopt technology but also how cost-effective their approach proves to be.

When OpenAI first introduced ChatGPT roughly four years ago and companies began experimenting with generative AI, the path forward seemed straightforward. Purchasing large language models from OpenAI or Anthropic, or leveraging open source models from Meta, appeared to be a safe bet for organizations that did not want to fall behind.

However, with the maturation of generative AI and the growing adoption of agentic AI, the build-versus-buy dilemma has grown more nuanced. Unlike generative AI tools that respond to individual prompts, agentic AI systems are designed to plan and autonomously execute multi-step tasks — interacting with external tools, databases, and APIs to complete workflows such as processing claims, managing inventory, or scheduling. This shift from content generation to autonomous action introduces new requirements around reliability, monitoring, and system integration that earlier AI deployments did not demand. Key considerations now include business size, specific applications, and what an enterprise or SMB identifies as its "moat" — the defining value proposition that separates and protects it from competitors.

Focus and Flexibility

For Home Depot, customer experience (CX) ranks among the most critical elements of the business. Ningyu Chen, the retailer's senior vice president of technology, emphasized this point during a session at the Ai4 2026 conference in Las Vegas.

"We would never outsource that piece to others," Chen said.

Nevertheless, Home Depot maintains partnerships with prominent AI technology vendors, including Anthropic, OpenAI, Google, and Microsoft, and actively incorporates their technology into its operations.

Chen advises businesses weighing the build-versus-buy question to prioritize identifying the right use case and to remain flexible about partnerships and vendor selection.

"I will leave the partnership very open," he said in an interview with AI Business. "We are still early, and there are a lot of things that can happen. Look at OpenAI, it was dominant; now it is not anymore."

The model provider landscape has indeed diversified considerably since ChatGPT's debut. Anthropic's Claude, Google's Gemini, Meta's open-weight Llama family, France's Mistral, and China's DeepSeek and Moonshot AI are among the alternatives now competing across different performance tiers and price points, giving buyers more options — and more complexity — than at any prior point.

"You have to have an abstraction layer to prevent yourself from these changes," he added.

Santhi Ramesh, CEO and chief AI strategy officer at advisory firm Future Propel, noted that businesses looking to build should carefully weigh the time and personnel they are prepared to invest in integrating generative or agentic AI into their platforms.

Ramesh, who previously led AI transformation initiatives at organizations including The Hershey Company and Ferrero, the Italian chocolate company, said the decision also hinges on a business's core competency.

"If your core competency is not a technology company, you are better off partnering and buying," she said in an interview. "If you have intense data, security, massive amounts of data and a custom solution that you want, then you must build on your own. I have done both."

These data and security considerations carry particular weight in regulated industries such as healthcare, financial services, and insurance, where obligations around data residency, auditability, and compliance can constrain the use of third-party AI platforms.

For enterprises and SMBs that require greater flexibility, Ramesh recommended testing, learning, and purchasing a platform or allying with a vendor under terms that allow an exit from the partnership — for example, after a beta trial.

The Cost of Buying or Building

Buying can lead to operational debt, cautioned Rachel Ibarra, vice president of data and innovation at Cardinal Group Companies, a property management and real estate organization. Cardinal Group adopts a hybrid approach, deploying an in-house AI agent called Stan.

Ibarra said that enterprises choosing between in-house development and external procurement must consider "what a cohesive system looks like."

"It's not just the ecosystem of technology," she said during a conference talk. "It's the ecosystem of how the organization works."

For businesses whose data confers a tangible competitive advantage, building may be the preferable route, Ibarra said. She also suggested that organizations consider building when they have a fractional need.

"So, you only need a part of something that's relatively small [and] where it's more internal facing and gives you a lot of internal advantage," she said in an interview. "That's what I would say are better built for nontechnical companies that don't necessarily have the resources to go and hire a huge engineering team."

Infrastructure represents another critical consideration, according to Markus McKay-Fleisch, professional services enablement director at Smartsheet.

"The real question for me is, do you feel like you have the right infrastructure to build yourself?" he said in an interview. "Is it a citizen development program where people in business are building, and if so, what are the guardrails you put around that?"

Businesses must also ensure they possess robust hosting infrastructure and the capacity to maintain AI agents or systems when they break or undergo changes. The major cloud providers — AWS, Google Cloud, and Microsoft Azure — have each introduced managed agent platforms intended to reduce this burden, offering pre-built orchestration and deployment tools that occupy a middle ground between full custom development and off-the-shelf products.

Furthermore, building an AI agent or system requires navigating the nuances of selecting the most suitable AI model for each use case.

Phenom, an HR technology and applied AI company that builds software to help organizations recruit candidates and develop skills, relies on multiple models. These include open models from Mistral and the low-cost Kimi model from Chinese vendor Moonshot AI.

According to Phenom CEO Mahe Bayireddi, organizations building AI agents or agent-based systems sometimes need to switch models to match the appropriate use case — certain models work well for HR applications, while others are better suited for finance.

Companies are "not understanding the nuances of how to be able to switch, not at your company level, but at an industry level," Bayireddi said. "Your models will switch on the fly."

Just a Bite Better

Steve Toy, founder and creator of Just a Bite Better, an AI nutrition app, opted to build AI agents directly within his application rather than purchasing a system to implement them. The app employs multiple agents that work in concert to provide consumers with a comprehensive overview of their nutrition.

"We don't use any one provider or any one model," Toy said in an interview. "We built our own model garden, if you will, which is merely just here is the code that accesses the APIs of all these things."

For Toy, the decision to build rather than buy or rely on a model garden from Google, AWS, or Microsoft stems from his technical expertise and the specific requirements of his business application.

"I'm not dealing with passing lots of money back and forth," he said, noting that consumers pay for the app through Apple, Google, or Stripe. Additionally, consumers can delete any personally identifiable information, and their data is encoded so it does not identify a specific individual.

"We're low stakes in many respects, whereas IBM and PayPal have to care about different things," Toy said.

For businesses still evaluating whether to buy or build, Toy recommended first understanding exactly what they need to purchase, particularly in a market saturated with AI vendors, products, and services.

"To figure out how to buy, you need to know what problem you are trying to solve and remember that every business is decomposed into pieces," he said. "When you're buying, you're buying to solve a problem, a part of your business."

Building, Toy added, can also help control costs. For Just a Bite Better, he implemented a system that imposes a hard cap on the number of tokens or the cost each model can consume monthly. While per-token inference costs for leading models have declined substantially since 2023, variable API pricing remains difficult to forecast as usage scales — making cost controls like Toy's increasingly relevant for organizations of all sizes.

"That's how you can prevent disaster, and that's really important because it's really easy to wake up in the morning and find a $50,000 bill," Toy said.