Opinions expressed by Entrepreneur contributors are their own.
Key Takeaways
- Coding agents are thriving, but GTM agents are barely scratching the surface.
- It’s not because AI isn’t smart enough, but because sales data is fragmented, duplicated and disconnected from the external signals that actually drive commercial decisions.
- The fix won’t come from writing better prompts or buying newer software wrappers. It’ll come from doing the foundational architecture work — unifying internal systems, anchoring them to verified external intelligence and giving agents a coherent view of the world.
If you look at where enterprise AI dollars are flowing, the disparity is stark. Software engineering teams are adopting autonomous agents almost overnight, while revenue operations — sales, marketing and go-to-market (GTM) — are barely scratching the surface.
As a CEO who spends time coding in Claude, building in Cursor and prototyping in Vercel, I understand why developers have embraced these tools so quickly: they’re incredible. Yet, when I talk to other executives about the relative quiet across their sales organizations, I find they usually draw the wrong conclusion.
They assume large language models (LLMs) simply aren’t mature enough to handle complex commercial motions. But that diagnosis misses the real bottleneck. The reason coding agents thrive while GTM agents struggle isn’t an intelligence problem. It’s a context problem.
The context trap: Codebases vs. commercial reality
In order to understand the gap, you have to look at the environments these two agents live in. A coding agent runs over a codebase. That codebase is self-contained, machine-readable and fully accessible inside a single repository. Every piece of context the model needs to write the next line of code exists right in front of it. The agent doesn’t need to consult external systems or guess what a third party thinks about its architecture.
A go-to-market agent, by contrast, faces a fragmented reality. Building an actionable account plan requires synthesizing past conversation histories, buyer profiles, executive tenure, funding rounds, technology stacks, earnings signals and open job postings.
Even if an enterprise centralizes its internal data across calls, emails and CRM records, that first-party view represents only a fraction of the necessary picture. For decades, companies assumed they were capturing account context by forcing sellers to log details into CRM fields. But reps rarely log complete information, and whatever does get entered is filtered through what revenue leaders call “happy ears” — the natural tendency of salespeople to interpret prospect interactions far more favorably than reality warrants.
More importantly, critical external signals like funding events, executive turnover and tech stack changes sit entirely outside internal systems. Without that external intelligence, an autonomous agent is operating blind.
The fragmentation and identity resolution nightmare
Solving that context gap isn’t as simple as plugging external data streams into your CRM. You first have to confront a messier internal reality: Revenue data inside most enterprises is notoriously chaotic. CRMs are routinely crippled by duplicate entries, inconsistent records and messy naming conventions. A single enterprise customer might appear as “Cisco” in a CRM, “Cisco WebEx” in call transcriptions and “AppDynamics” inside an outreach platform.
If an AI agent attempts to reason across this disconnected dataset without an identity resolution framework, it inevitably draws flawed conclusions. It might pull conversation notes from one entity, apply financial metrics from another and deliver a next-best action that is confidently wrong.
Look at how vertical AI has succeeded in sectors like the legal industry. Specialized platforms like Harvey and Legora don’t rely on generic LLMs alone; they ground their models in domain-specific reference architecture and verified legal datasets. GTM AI requires the exact same foundation. A generic model does not understand B2B commercial logic out of the box. To generate real value, an agent must be anchored in a unified reference data layer.
Democratizing the infrastructure layer
Historically, unifying first- and third-party data required massive engineering teams and multi-quarter custom implementations stuck behind IT backlogs. But as intelligence layers mature, that dynamic is evolving. When underlying data architecture is exposed through flexible APIs and Model Context Protocol (MCP) integrations, even non-technical business leaders can construct custom AI workflows in an afternoon.
Recently, I spoke with the CEO of a 50-person mid-sized business who reached out regarding a quick API integration question. He wasn’t a software engineer or a RevOps builder. Yet, using Claude Code paired with ZoomInfo’s API infrastructure (GTM.ai), he was able to build a custom account-scoring and enrichment application tailored specifically to his team prospects.
A few years ago, he would have been forced to rely on whatever rigid software interface a vendor built for him. Instead, he was interacting directly with a unified data layer inside Claude to automate his team’s specific commercial logic. This reflects a massive structural shift: moving away from traditional software applications where hundreds of thousands of users log into a single interface, toward millions of tailored, natural-language interfaces grounded in live data.
Grounding the future of AI GTM
In the end, the advice I give to other revenue leaders is always the same: Don’t confuse a slick demo with a viable enterprise strategy. Right now, dozens of lightweight AI sales tools are stalling out because they built polished interfaces without a durable data foundation beneath them. An autonomous agent is only as intelligent as the context layer feeding it. If you bolt an agent onto fragmented data, you get unreliable outputs every time.
The real unlock in go-to-market won’t come from writing cleverer prompts or buying newer software wrappers. It comes down to doing the foundational architecture work — unifying internal systems, anchoring them to verified external intelligence and giving agents a coherent view of the world. The leaders who capture the promise of enterprise AI won’t be the ones waiting for foundation models to magically solve B2B complexity. They will be the ones who build that prerequisite context layer today so their agents have the complete picture required to deliver.
Key Takeaways
- Coding agents are thriving, but GTM agents are barely scratching the surface.
- It’s not because AI isn’t smart enough, but because sales data is fragmented, duplicated and disconnected from the external signals that actually drive commercial decisions.
- The fix won’t come from writing better prompts or buying newer software wrappers. It’ll come from doing the foundational architecture work — unifying internal systems, anchoring them to verified external intelligence and giving agents a coherent view of the world.
If you look at where enterprise AI dollars are flowing, the disparity is stark. Software engineering teams are adopting autonomous agents almost overnight, while revenue operations — sales, marketing and go-to-market (GTM) — are barely scratching the surface.
As a CEO who spends time coding in Claude, building in Cursor and prototyping in Vercel, I understand why developers have embraced these tools so quickly: they’re incredible. Yet, when I talk to other executives about the relative quiet across their sales organizations, I find they usually draw the wrong conclusion.
They assume large language models (LLMs) simply aren’t mature enough to handle complex commercial motions. But that diagnosis misses the real bottleneck. The reason coding agents thrive while GTM agents struggle isn’t an intelligence problem. It’s a context problem.
This story originally appeared on Entrepreneur
