Build vs. Buy: How Code and Theory Navigates AI Adoption at Scale

A conversation with David DiCamillo, CTO at Code and Theory

The AI landscape is evolving faster than most enterprises can keep up. Every week brings new models, new tools, and new possibilities. For enterprise leaders, this creates a critical question: when should you build your own AI solutions, and when should you buy?

We sat down with David DiCamillo, CTO at Code and Theory, to explore how a leading digital transformation agency navigates this decision. With 20 different AI deployments across their organization, Code and Theory has learned valuable lessons about experimentation, adoption, and what actually drives value.

The Reality of Building Your Own AI Tools

Many companies are tempted to build their own AI solutions. The appeal is obvious: complete control, custom features, and the excitement of working with cutting-edge technology.

But DiCamillo's experience reveals a more nuanced reality.

"We've definitely had scenarios where we've started up on things that we're going to incubate internally and just turn around and say, whoop, that's on the market and it's a lot cheaper," DiCamillo explains. "There's a delicate way to talk to the engineers who have built this. This is their equity, right? They put their time into this."

The challenge isn't just the initial build. It's what comes after. As Sachin Kamdar, CEO of elvex, points out: "Everything's exciting. It's greenfield like you can choose whatever you want and like you start to get into it and it's very, very fun. There are no bugs yet. There's no real users of the product. So everything's fun to work on."

But once you have real users, feature requests pile up. Competitors keep shipping new capabilities. And suddenly, your two-week prototype becomes a maintenance burden that pulls engineering resources away from your core business.

When Building Makes Sense

That doesn't mean building is always the wrong choice.

Kamdar identifies clear scenarios where custom development delivers value: "The stuff that tends to be successful is when it's more aligned with the core. So if you're already a software company in some way, shape or form, you're adding AI features to something that's a complicated use case."

Code and Theory has successfully built custom tools for specific client needs—particularly when they identify gaps in the market or need capabilities that don't exist in available products. But even then, they're strategic about it.

"We may spend the effort and time there to go off and design and build a product, beta test, quick, you know, how can we do the POC before we invest more time," DiCamillo notes. "But what we have spent a lot of time to do is also research and talk with partners out there and see what is the best products hitting the market."

The Hidden Cost of "Control"

One of the most common arguments for building is control. But DiCamillo challenges this assumption.

"You have to always balance what's the upside of having that control, how much effort you're spending to maintain that versus the benefits for the organization," he explains.

In fact, Code and Theory actively cost models their build decisions, comparing engineering time and resources against available solutions in the market.

The control argument also overlooks a key advantage of today's AI landscape: vendor responsiveness. Unlike trying to influence the roadmap of established enterprise software, many AI startups are eager to hear customer feedback and quickly iterate.

"What's great about it is that there's, you know, these are all companies that are striving to take grab land in this sort of big land grab of this new world," DiCamillo observes. "And they are open to hearing from their partners, from their customers about where they need to go and what the use cases are that are driving their business."

The Power of Experimentation

Perhaps the most important insight from Code and Theory's approach is their commitment to experimentation.

"We have about 20 different AI deployments right now within the organization," DiCamillo shares. "And that's everything from just access to OpenAI and Claude and Gemini through very specific production products."

But this isn't about giving everyone access to everything. It's strategic experimentation with clear evaluation criteria.

"We have 20 tools, but some of them are only available to five or six people for very specific reasons," DiCamillo explains. "A research team needs access to a perplexity pro at a very different level than other people."

This culture of experimentation has revealed unexpected insights. For example, Code and Theory discovered that Anthropic significantly outperforms OpenAI for parsing client performance data—something they would never have known without testing both.

Why Code and Theory Chose elvex

When Code and Theory evaluated AI platforms, they looked for specific capabilities that would enable broad adoption across their organization.

"Part of the reason why we found elvex so fruitful for our organization is that the ability to build custom AI assistants, the team has been doing that with OpenAI for a while," DiCamillo explains. "But now we have the control and we have the ability to embed them into other core products that we are using like Slack."

This integration capability proved critical. As DiCamillo notes: "The hardest part about all of these tools is the change management around it. Everyone's stuck in their ways."

By embedding AI capabilities directly into Slack, elvex eliminated the adoption barrier of asking people to visit yet another tool. The AI comes to them where they already work.

Code and Theory also valued elvex's consumption-based pricing model. "What I have a hard time paying for is a license that gives people free range to do whatever they want but they're not going to use it," DiCamillo says. "If someone's going to sit down and use the tool, I'm happy to pay for it every single time."

Key Principles for AI Adoption

Based on Code and Theory's experience, several principles emerge for enterprise AI adoption:

Start small and prove value. Don't make massive upfront investments. Test, learn, and scale what works.

Enable experimentation without chaos. Create space for teams to explore new tools while maintaining governance and security standards.

Meet people where they work. The best AI tools integrate into existing workflows rather than requiring behavior change.

Focus on high-leverage activities. Look for areas where small productivity improvements create significant business impact.

Align pricing with usage. Consumption-based models ensure you're paying for value delivered, not unused licenses.

Build relationships with responsive vendors. In a fast-moving space, vendor partnership matters more than feature checklists.

The Future of Enterprise AI

As we look ahead, the conversation is shifting from "should we use AI?" to "how do we use AI effectively across our entire organization?"

The companies that will succeed aren't necessarily those with the biggest AI budgets or the most engineers. They're the ones that create cultures of experimentation, make strategic build-versus-buy decisions, and focus relentlessly on adoption and value creation.

As DiCamillo puts it: "If we stop that thirst for exploration around new tools, then we're going to die as a company."

Ready to Transform Your AI Strategy?

Code and Theory's journey demonstrates that successful AI adoption requires the right platform, the right approach, and the right partner.

elvex makes it easy and safe for anyone in your organization to create and deploy AI assistants—without requiring technical expertise. Our platform integrates seamlessly with tools like Slack, offers consumption-based pricing, and provides the security and governance enterprise organizations require.

Want to learn how elvex can accelerate your AI adoption? Get in touch with our team to explore how we can help you unlock the power of AI across your organization.

‍