Context Graphs: Startup Opportunities and Market Signals
Three Strategic Paths for Startups
Path 1: New Systems of Record for Decisions
Startups begin as orchestration layers but evolve into systems of record, persisting what enterprises never systematically stored: decision-making traces.
Progression:
- Start with automation of workflow (L2/L3 support, customer success decisions, etc.)
- Build context graph as byproduct of decision tracing
- Replayable lineage becomes the authoritative artifact
- Agent layer becomes where business goes to answer “why did we do that?”
Example: PlayerZero in production engineering
- Domain: Intersection of SRE, support, QA, and dev (a “glue function”)
- Problem: Humans carry context that software doesn’t capture
- MVP: Automate L2/L3 support decisions
- Real asset: Context graph of how code, config, infrastructure, and customer behavior interact in reality
- Becomes source of truth for: “Why did this break?” and “Will this change break production?”
Why this works:
- These are glue functions where decision logic is too complex for traditional automation
- Rich decision traces naturally accumulate
- Context graph becomes more valuable than the automation itself
Path 2: Observability for Agent Decision Quality
As context graphs accumulate and decision traces grow, enterprises need infrastructure to monitor, debug, and evaluate agent behavior at scale.
Positioning: Just as Datadog became essential infrastructure for application monitoring, observability for agents becomes essential infrastructure for decision quality.
Core capabilities:
- Visibility into how agents reason
- Where they fail in decision-making
- How their decisions perform over time
- Impact of policy changes on behavior
- Precedent usage and drift
Example: Arize in agent observability
- Provides visibility into agent reasoning
- Tracks decision performance
- Helps teams improve agent quality
- Becomes essential for managing context graphs at scale
Why this works:
- Context graphs create massive operational complexity
- Enterprises need tools to understand agent decisions
- Observability becomes as critical as the agents themselves
Path 3: Vertical-Specific Orchestration Platforms
Specialize orchestration to specific domains where decision traces are particularly valuable.
High-value verticals:
- Deal desk orchestration – Complex discount/ARR decisions with exception-heavy logic
- Support escalation – Routing, triage, and resolution with high precedent value
- Underwriting – Complex rule application with exception override patterns
- Compliance review – Policy evaluation with regulatory lineage requirements
- Customer success management – Renewal, upsell, churn decisions with precedent value
Why verticals matter:
- Each has “glue functions” that exist precisely because no system captures cross-functional context
- Each has exception-heavy decisions where precedent matters
- Each has high labor cost if manual
- Each has regulatory/operational need for decision lineage
Key Market Signals for All Paths
Signal 1: High Headcount in Manual Workflow
If a company has 50+ people doing a workflow manually:
- Routing tickets
- Triaging requests
- Reconciling data between systems
- Making exceptions to rules
- Escalating decisions
What it signals:
- Labor exists because decision logic is too complex for traditional automation
- Decisions are exception-heavy and precedent-dependent
- Manual process preserves context that software doesn’t capture
- Opportunity to capture context graph while automating
Example: RevOps teams exist because someone has to reconcile sales, finance, marketing, customer success. DevOps exists to bridge development, IT, operations. These “glue functions” are signals of orchestration opportunities.
Signal 2: Exception-Heavy Logic
Look for workflows where:
- Routine cases = 70%, exceptions = 30%
- Logic is “it depends” (precedent-based, not rule-based)
- Same exception patterns repeat quarterly/annually
- Decisions require cross-system context
What it signals:
- Traditional automation won’t work (too many edge cases)
- Human decision-making is what you need to systematize
- Rich precedent will accumulate quickly
- Context graph becomes valuable immediately
Example workflows:
- Deal desk pricing (90% of requests are exceptions to policy)
- Support escalation (each escalation depends on different factors)
- Underwriting (rule application depends on combination of factors)
- Budget approvals (policy caps exist but precedent exceptions are normal)
Signal 3: Cross-Functional “Glue Function” Organizations
Look for companies with dedicated teams that exist at intersections:
RevOps – Sales + Finance + Marketing + Customer Success
DevOps – Development + IT + Operations
Security Ops – IT + Engineering + Compliance + Risk
Partner Enablement – Sales + Product + Marketing + CS
What it signals:
- No single system of record owns the workflow
- Organization had to hire humans to carry context across systems
- Context that matters is distributed and hard to access
- Ideal place to insert orchestration platform
Why this matters for new systems of record:
The existence of the glue function role means you’ve already identified:
- Cross-system workflow that matters
- Complex decision logic in that workflow
- Labor cost that could be reduced with better context capture
- Natural boundary for a new system of record
Market Size and Opportunity
The Trillion-Dollar Question
The last generation of enterprise software created trillion-dollar ecosystems by owning systems of record:
- Salesforce owns customer data
- Workday owns employee data
- SAP owns operations data
New frontier: Systems of record for decisions
- Decision traces (not just object state)
- Precedent (not just policies)
- Lineage (not just current facts)
- Why (not just what)
If decision-record platforms reach same TAM as object-record platforms, this is a trillion-dollar opportunity.
Incumbent Limitations
Existing platforms can’t build context graphs because:
Operational players (Salesforce, ServiceNow, Workday):
- Inherit current-state architecture
- Can’t replay state at decision time
- Can’t be in execution path for cross-system workflows
Data warehouse players (Snowflake, Databricks):
- See data after ETL
- Read path only (writes are done elsewhere)
- Miss the moment of decision
Only orchestration startups can capture traces because:
- They’re in the write path
- They execute workflows
- They see full context at decision time
- They can record in the moment
Strategic Advantages for Early Movers
- First-mover access to precedent – Early customer contexts become competitive moat (harder to replace once encoded in decisions)
- Network effects – Richer context graph → better agent decisions → more enterprises want to plug in
- Vertical defensibility – Once you own decision records for a vertical, replacement friction is high
- Data as asset – Decision traces become valuable for training, consulting, benchmarking
Timeline and Adoption Phases
2026: Early adopters (high-labor, exception-heavy workflows)
2027-2028: Mainstream adoption (move from 80% to 99% automation)
2029+: Strategic necessity (enterprises that don’t have decision lineage fall behind)
Critical inflection: When context graph goes from “nice to have” to “required for competitive autonomy.”
Key Insight for Founders
The billion-dollar question isn’t “how do we automate more?” It’s “how do we systematize the decision-making that humans currently do?”
Answers to that question become trillion-dollar platforms.
References
- Foundation Capital: “AI’s Trillion-Dollar Opportunity: Context Graphs”
- Authors: Jaya Gupta, Ashu Garg
- Example referenced: PlayerZero (production engineering orchestration)
- Example referenced: Arize (agent observability infrastructure)