We've been observing patterns in what enterprises are building for go-to-market, and wanted to share a common framework:
1. They have a data lake (e.g., Snowflake, Databricks + 3rd party data sources, Segment, etc.)
2. Then an orchestration layer and semantic layer (e.g., Dust, Relevance, Claude Code, Clay, dbt, Dagster)
3. Then an agent layer where the work is happening (e.g., chat, LLMs, bought tools, and in-house built tools like a rep-facing sales UI) – most of them are investing in “Build” here.
Some things we're hearing repeatedly:
- Source quality:
- They all are investing significant money into clean, high-quality data sources for layer #1. They spend months cleaning data, aggregating all their data in a warehouse, capturing signals, buying 3rd-party data, etc.
-
- Plumbing:
- Setting up that foundation of #1 and #2 can take 6-12 months or more, but once they do that, they can stand up tools in layer #3 without rethinking the plumbing every time.
-
- End-users:
- For example, it might take a year to get an internal sales-facing tool working really well, then it’s faster to build for other end-users like solutions engineers and customer success.
-
- LinkedIn lies:
- Yes, it really does take this long. The ideas come easy, high-quality execution is hard.
-
- Build:
- The closer they get to the UX of a seller or a customer (layer #3), the more likely they are to try to build it in-house vs. buy good data at layer #1.
-
- EPD builds it:
- They are staffing full-stack engineers to build out these layers. They deploy full product teams on these projects. They think of them not as “just” GTM initiatives, but as products that teams around the company are accountable for, including EPD partners. Treating them like products also means that designers study sellers as end-users to create excellent UI’s.
-
- Few UI’s:
- They’re going full-product-UI, because chat-based UI’s (e.g., Claude Code) aren’t really ideal for sellers/CS/SE’s. The name of the game is consolidation – sellers are usually looking at ~10 UI’s as they strategize a deal, is it possible to reduce to 1-5? A sales-facing UI might show a daily deal briefing and pre-loaded actions to take.
-
- Leaders:
- The people leading these initiatives are sometimes titled CRO or CMO, but often they’re “AI GTM” leaders in deep partnership with EPD leaders (think: Asana’s Amrutha Suresh, Head of GTM AI & Innovation, who previously established their “GTM Product Management” practice). Either way, they’re systems-and-architecture thinkers.
-
- Companywide:
- It’s both top-down and bottoms up. Executives reinforce usage of internal tools (and again, developing them is a companywide XFN effort).
-
- Yet grassroots:
- Meanwhile, they’re clearing security blockers to allow employees to do grassroots experimentation.
- Picking an orchestration layer #2 is the unlock here – Claude Code is one option, but something that’s a little more pre-configured or “on rails” makes it even easier. (Vanta uses Dust; Canva uses Relevance).
- Valuable inventions come out of this, then the company can double down with EPD resources.
At Octave we support this stack in two ways.
One – Exceptional intelligence from conversations like calls and emails. Octave in your semantic layer #2 lets you wring more value & quality out of the data in layer #1.
We annotate conversations with more accuracy and precision than Claude or Gong can, because we’ve trained on the products you sell and the GTM motions you run.
Two – We have your company’s GTM strategy stored in a context graph (personas, product information, segments, etc.) to make your agents in #3 put out higher-quality content.
In a third-party test of 5 context systems across 9 GTM tasks (e.g., make an account plan), Octave’s context storage produced the highest-accuracy outputs at the best quality-to-token-cost ratio.
Want an account plan? It will be more accurate if it can pull from Octave. A 20% more accurate account plan can push a deal over the line.
Want 50 one-pagers for 5 personas across 10 industries? Octave can help you nail the message for each one – enterprise ABM at commercial scale.
Want to know if your reps are doing a good job? Most call analysis tools don’t know what they don’t know. They mis-grade and you don’t notice.
Their insights are generic: “Your AE’s talk too much.” What you really want is something that understands what you sell, the markets you sell into, and the motions you run. That way, it knows your AE missed the kill shot for a common objection on Monday, and another AE showed a healthcare case study to a financial services buyer on Tuesday.
Octave knows exactly what good looks like for your business because it stores your GTM strategy as context, and compares that to reality. This tension between hypothesis and reality, one improving the other, ensures that all 3 layers in the enterprise AI stack operate on a quality foundation for more closed deals.
