Choosing Between Building and Buying AI: A Leadership Decision Framework

The core tension for today’s revenue and technology leaders is whether to buy an AI solution, build one internally, or pursue a hybrid approach. Gong’s latest research outlines a decision framework that moves past generic advice and anchors the choice in measurable business outcomes.

Start with the problem, not the technology

Leaders often feel pressure to adopt AI before clarifying the problem they need to solve. Gong’s interview with Oscar Canon, AI COE leader at Cisco, stresses that teams must first articulate the desired business outcome and the metric that will prove its impact. An honest inventory of existing capabilities is the next step; many organizations already own under-utilised tools that can be repurposed at minimal cost. ConnectWise discovered that its reps were manually feeding Gong transcripts into generic AI assistants—a capability already embedded in Gong’s own AI—prompting them to focus on enablement rather than new development.

Classify every AI job into three buckets

Gong proposes a three-bucket model for assigning AI responsibilities:

1. General-purpose AI platforms that handle a broad set of tasks across sources.
2. Purpose-built AI platforms that deliver deep, domain-specific functionality within existing workflows.
3. In-house builds for use cases that rely on proprietary data or are unique to the business.

These categories are not mutually exclusive. Cisco reserves in-house development for proprietary data scenarios, while leveraging both general-purpose and purpose-built solutions for the rest of its stack. ConnectWise’s “Account Brain” agent illustrates a hybrid approach: it aggregates data from five third-party platforms—including Gong—into a single interface, enabling reps to surface upsell opportunities without juggling multiple tools.

Account for the true cost of ownership

The perception that building AI is cheap is challenged by Gong’s findings. While initial development may be inexpensive, the ongoing expenses—engineer time, maintenance, upgrades, and downtime—can outweigh the upfront savings. Oscar notes that every month of delay erodes potential value, and Austin Reichart of ConnectWise emphasizes that “the buck stops with you” when an internally built system fails. A low-code pilot, as ConnectWise employed, can surface hidden cost drivers before scaling.

Treat vendors as partners, not one-time purchases

A final insight from the research is the importance of partnership with AI vendors. Rather than viewing vendors solely as product suppliers, teams should engage them in ongoing dialogues about challenges and upcoming builds. Austin’s conversations with vendors often revealed that a requested feature was already on the vendor’s roadmap, sparing ConnectWise from duplicative development. Oscar advises against competing with vendors on features they already plan to deliver, reinforcing the value of collaborative relationships.

Our read: Gong’s framework cuts through the hype by insisting on outcome-first thinking, a clear taxonomy for AI job allocation, and a realistic appraisal of total cost of ownership. The real-world examples from Cisco and ConnectWise demonstrate that hybrid strategies, underpinned by strong vendor partnerships, can deliver measurable results without the hidden expenses that often plague pure-build initiatives.