Build, buy, or both? Designing a leadership framework to know how

Every revenue and technology leader I talk to is wrestling with some version of the same question: should we buy AI, build it ourselves, or do both?

There's no universal answer, and anyone who offers one should make you suspicious. Every company's data, workflows, and goals are different. But after working with some of the largest enterprise buyers of AI, I've seen clear patterns in what turns AI investment into measurable business results and what doesn't. Those patterns add up to a decision framework any leadership team can use.

It starts with understanding how we got here.

Phase 1: Everyone could build. About a year ago, AI capabilities exploded, both inside the platforms we use and in our ability to build our own. Suddenly anyone could create an agent. That sparked a real builder mindset, which was good. It also set off a frenzy. Organizations pushed everyone to use as much AI as possible, and boards called for bold moves to signal progress, often before anyone had defined the plan or the objective.

Phase 2: Fragmentation. Within about six months, many teams hit the limits. Small groups spun up hundreds of their own agents and skills. Use cases multiplied, and the same work got done a dozen different ways. Leaders lost visibility. Managers couldn't tell whether people were following the behaviors they coached, using the enablement they'd invested in, or sticking to agreed best practices. Everyone had their own AI, and the organization lost its ability to steer.

Phase 3: Intentional architecture. That's where most enterprises are now. They want a deliberate approach built around business outcomes: one operating rhythm that lifts the performance of whole teams, not just individuals. Efficiency still matters, but effectiveness matters more. And the plan needs to be designed from the start with leading indicators that tie to financial return.

What is emerging from these lessons is not a naive and simplistic one-size-fits-all plan, but rather a data-driven leadership decision framework that can guide leaders to AI investment that truly delivers and transforms.

At Gong Celebrate ‘26 I had the chance to sit down with two leaders who recently led their organizations through this evaluation to build or buy: Oscar Canon, AI COE Leader at Cisco, and Austin Reichart, Sales Enablement Manager at ConnectWise. During our panel session: "Decide: Build, buy, or both," we talked through how to approach that decision through real examples from each of their teams.

The pressure to keep up with AI pushes teams to buy or build before they've defined the problem. Start instead with the business outcome you want and how you'll measure the financial impact once you get there.

"You truly need to understand the work," Oscar said. That means understanding the people, their workflows and the specific tasks inside them. Only then can you decide where AI should add context, where it should automate, and how people and AI should work together.

Next, take an honest inventory of what you already have. "Before you start building a duplicate solution, truly understand what you have," Oscar said. Capabilities often sit unused because they were rolled out poorly or never enabled. The cheapest capability you'll ever add is the one you already pay for.

ConnectWise found exactly that. Their honest assessment showed reps pasting Gong transcripts into general-purpose AI assistants to craft next best actions, something Gong’s AI does automatically. “It was an enablement challenge,” Austin said. “We had the capability, so we started there.”

Finally, map where you are today against where you want to be, and define the stages in between. Each stage should raise your maturity and effectiveness a step at a time. (How to design those stages is a topic for another post.)

The takeaway: Before any build-or-buy decision, write down the outcome, the metric that proves it, and what you already own that could get you there.

Every AI architecture rests on data. AI is only as effective as the data and context underneath it, whichever models or platforms you use. That much is table stakes.

What companies underestimate is the layer above raw data: the knowledge layer, or knowledge graph. It maps how information relates, what matters most to your business, and the rules around it, including governance, privacy, and security. It's what turns data into meaning.

Here's a sales example. Say I'm the seller on a call with Oscar, who's the buyer. Raw data can give me a summary and action items. That's useful, and relatively easy. A knowledge layer can tell me much more:

• What should I have said?
• Which key behaviors did I miss, and what worked?
• Where were the gaps, and what would have been more effective?
• How does the conversation connect to what Oscar cares about as an executive buyer, his key questions and his decision process?
• Have we looped in Austin, who wasn't on the call but is an important influencer?

The same idea applies in any function, from support to finance to operations. Data tells you what happened. The knowledge layer tells you what it means and what to do next. Together, they determine how useful everything built on top of them can be: the intelligence, the workflows, and the experience people actually use.

The takeaway: Whatever you build or buy, ask where its context comes from and how it connects to the business logic that matters to you.

Which AI should handle which job is probably the hardest question leaders face right now. In practice, every job to be done fits best in one of three buckets, and most companies use all three.

• General-purpose AI platforms handle a wide range of tasks and reason across many sources.
• Purpose-built AI platforms bring specialized depth in a specific domain, often embedded directly in the workflows where that work happens.
• What you build in-house covers what no vendor can provide. Cisco, Oscar explained, reserves building for work that runs on its own proprietary data or for use cases unique to the business.

These aren't competing choices. General-purpose and purpose-built platforms work best together, each extending the other, and open standards like MCP make it easier to connect them with what you build.

The connections matter as much as the choices. Sellers don't want to jump between tools to get through one workflow, Oscar said. Without integration, you end up with five or ten disconnected solutions and a harder job, even with AI everywhere.

ConnectWise shows what a well-justified build looks like. With more than 35 products, Austin's team needed sellers to spot upsell opportunities and help new hires ramp faster. But reps were working across five interfaces and piecing together each account's story by hand.

So the team built an internal agent called Account Brain that pulls from five third-party platforms, Gong among them. Before a meeting, reps see their internal data, recent customer conversations, product usage and what customers say about ConnectWise on sites like Reddit and G2, all in one place.

The takeaway: Build where you're unique, buy where others have depth, and invest in connecting the two.

Building has never looked easier. But the build is the cheap part, and the total cost of ownership surprises many enterprises.

When you build, the costs keep coming after launch: engineers, maintenance, new releases, and the solution's full lifespan, Oscar said. You own the upkeep, and you own the roadmap. He also weighs speed to value, because every month of delay is a missed opportunity.

You own every outage and every troubleshooting cycle, too. "If the system goes down, the buck stops with you," Austin said.

ConnectWise kept costs in check by starting small. Austin's team ran a low-code pilot in a controlled environment to learn how reps used it and where costs ran higher than expected, then optimized from there.

The takeaway: Compare the full lifetime cost and time to value of building, not just the cost of version one, and pilot small before you scale.

The last move is the one teams often overlook: treat your vendors as true partners, not one-time purchases.

Talk to your vendors about more than renewals. Austin tells his vendors what his team is struggling with and what they're thinking about building. Often the answer is some version of, "Can you wait two or three months? It's probably on the roadmap." Those conversations have kept ConnectWise from building things it was about to get anyway.

That's why Oscar advises against competing with your vendors on their core strengths. They have the expertise, and they can usually move faster. Save your build effort for what's unique to you.

Share your plans with the platforms you rely on, influence their roadmaps, get early access where you can, and consolidate overlapping point tools over time. You build less and carry less risk, without giving up capability.

The takeaway: Before you build, ask your strategic partners what's coming.

So build, buy, or both? Most companies are architecting a combination of the two. Most importantly, every company can build a framework for making decisions that take a cohesive, mature approach to achieving their business outcomes.

That framework is taking shape in this third phase. When everyone was spinning up their own AI, roughly 80 percent of the effort went to individual experimentation. Now the balance is flipping. Leading organizations put about 80 percent into core workflows and best practices standardized across teams, and leave the rest for creativity at the edges. The numbers are directional, not a formula, but the shift is real.

The capabilities available to you will keep changing, so you'll be making these calls again, Oscar noted. Take stock of what you already own, put your core workflows where your data lives, and build only what makes you different.

Or, as Oscar summed it up: buy the capability, build the differentiation, and integrate it into the flow of work.

1. What business outcome are we solving for, and how will we measure it?
2. Do we already own something that could do this if we enabled it properly?
3. Is this work unique to us, or does someone already do it with more depth?
4. What will it cost to own over its full life, and how fast does each option get us to value?
5. Have we asked our strategic partners what's on their roadmap?

If you're weighing these decisions now, see how Gong's Revenue AI fits into your stack.

Watch the full "Decide: Build, buy, or both" session on-demand now, along with more sessions from Celebrate ’26.