Context Graphs for GTM: Why Your CRM Isn't Your Source of Truth Anymore

A fascinating conversation is emerging in the GTM space right now. It started with Foundation Capital’s thesis on context graphs as AI’s trillion-dollar opportunity, got picked up by builders like Kirk Marple and Animesh Koratana, and is now landing squarely in GTM territory with Anshul Gupta’s piece on why Salesforce can’t build context graphs.

The core insight: Systems of record tell you what happened. Context graphs tell you how and why it happened.

This distinction matters enormously for GTM teams. And it explains why the best reps consistently outperform average ones, even with identical territories and resources.

The Decision Trace Problem

This is what gets lost in every B2B sales cycle:

Your CRM stores outcomes. But the reasoning, the exceptions, the “we tried X but Y worked better” knowledge? That lives in your top rep’s head. In Slack threads nobody will ever find again. In call recordings that nobody has time to review.

Foundation Capital calls this the difference between rules and decision traces. Rules are general: “Always multi-thread above €50k deals.” Decision traces are specific: “On the Acme deal, we multi-threaded because the VP went dark for two weeks, and reaching the CFO directly saved the quarter.”

The best sales teams don’t just follow playbooks. They accumulate decision traces that compound into organizational intelligence.

Why GTM Is Uniquely Broken

Koratana identifies the “two clocks” problem: organizations have built infrastructure for the state clock (what’s true now) but almost nothing for the event clock (how things became true).

GTM might be the most extreme example:

People fragmentation: SDR researches the account, hands off to AE, who closes and hands off to CSM, who eventually hands back to AM for expansion. Each handoff loses context. Each person rebuilds understanding from scratch.

Systems fragmentation: The full picture of an account lives across Salesforce, Gong, LinkedIn, email, Slack, product analytics, and that one Google Doc from the original discovery call. No single system sees the whole story.

Tribal knowledge: Your best AE knows that Acme’s procurement process takes 6 weeks, that their CTO hates vendor lock-in, and that deals always stall in August. None of this is written down anywhere.

The result? As Marple puts it: Salesforce is technically a system of record, but it’s no longer the source of truth. Reps consult Slack and call transcripts before they check the CRM.

What Context Graphs Would Change

Imagine if every account had a living memory that captured:

• What was tried: Which messaging resonated, which fell flat

• Why decisions were made: Not just “moved to closed-lost” but the specific objection that killed it

• Who matters: The real org chart, updated as relationships evolve

• What’s changed: New funding, leadership changes, competitive moves, the moment they became ready to buy

This isn’t science fiction. The building blocks exist. Kirk Marple’s work on operational context layers shows how to resolve identity across systems. Koratana’s “agents as informed walkers” approach discovers organizational structure through use. And the entire GTM AI wave is generating decision traces as a byproduct of automation.

The question is whether these traces get captured or lost.

Where Compelling Fits

We’ve been thinking about this problem from a different angle: transparent, explainable sales intelligence.

When Compelling ranks a lead, we don’t just say “this account scores 87.” We show why: recent funding, tech stack match, buying signals from three sources, similar profile to your last five closed-won deals. That’s a decision trace. It’s auditable. It compounds.

When we automate account research, we’re not just saving time. We’re creating a persistent layer of context that follows the account through its lifecycle. The research doesn’t disappear into a rep’s mental model. It’s captured, searchable, and builds over time.

We’re not claiming to have solved context graphs. Nobody has, yet. But the direction is clear: GTM teams need systems that capture why decisions are made, not just what happened.

The Trillion-Dollar Shift

Foundation Capital’s thesis is bold: the next platform giants won’t be built by adding AI to existing systems of record. They’ll be built by capturing decision traces.

For GTM specifically, this means:

1. CRM becomes a sync target, not the source of truth. The real intelligence lives in context layers that understand relationships, timing, and causality.

2. AI agents need memory, not just tools. An SDR agent that can’t remember why a similar deal failed last quarter is just a faster way to repeat mistakes.

3. Handoffs become transfers, not restarts. When context persists, the CSM doesn’t need to rebuild understanding of how the deal was won.

4. Forecasting becomes causal, not correlational. Instead of “deals at this stage close 40% of the time,” you get “deals with this decision pattern close 70% of the time.”

The GTM teams that figure this out first will have a structural advantage that goes past better data into better reasoning: the ability to learn from every deal, not only the ones their top rep happened to work.

What to Watch

This conversation is moving fast. The technical pieces are coming together: identity resolution across systems, temporal modeling, agent memory frameworks, and the MCP protocol for cross-system context.

If you’re building GTM infrastructure, the question isn’t whether context graphs matter. It’s whether you’re capturing decision traces or letting them evaporate into Slack threads and forgotten call recordings.

We’re betting heavily on the former. And based on how quickly this conversation is spreading, we’re not alone.

Further reading:

• Context Graphs: AI’s Trillion-Dollar Opportunity (Foundation Capital)

• The Context Layer AI Agents Actually Need (Kirk Marple / Graphlit)

• How to Build a Context Graph (Animesh Koratana)

• Building Context Graphs for GTM Agents (Anshul Gupta)