After I shared my experience building AI GTM agents, a few people reached out with a question about my learning of the model: changes once the agents were in place, and where the real opportunity to refine things might be.
I found myself finding that answer in the Foundation Capital's blog “AI’s trillion-dollar opportunity: Context graphs” by Jaya Gupta and Ashu Garg.
Their piece articulated that modern systems are remarkably bad at preserving reasoning.
We store the final state of a deal, a campaign, or a renewal, but not the logic, exceptions, and tradeoffs that produced that state.
In GTM, this gap is structural.
Sales, marketing, and customer success are constantly making judgment calls based on incomplete information.
Two opportunities with similar numbers receive different treatment because of a nuance that never makes it into a CRM field.
A discount is approved once because of precedent, but that precedent is never recorded.
A deal stalls not because the funnel is broken, but because the context that justified urgency lived in a conversation and then vanished.
What building agents surfaced for me is that when an agent gathers inputs across systems, evaluates policies, surfaces exceptions, and routes approvals, it briefly sees the full picture.
That moment is usually ephemeral. Once the decision is made, the system only remembers the result, not the reasoning.
The idea of a context graph says that instead of treating decisions as disposable moments, it treats them as durable artifacts.
Over time, those artifacts form a living map of how judgment has been applied across accounts, segments, and scenarios.
For GTM teams, this turns tribal knowledge into institutional memory.
In practice, many decisions still need humans in the loop.
But when those decisions are recorded with their rationale, the organization learns instead of forgetting.
So the context lives on to not only help humans do more but also to remember better.
Curious how others are encountering this in their own teams and products.
