Trace, a London-based startup from Y Combinator’s 2025 summer cohort, has raised $3M in seed funding to tackle one of enterprise AI’s thorniest challenges: getting autonomous agents to actually work inside complex organizations. Rather than build yet another agent, the company is betting that context is the missing layer — the connective tissue that lets AI understand a company’s people, processes, and systems well enough to deliver reliable results.
The round includes participation from Y Combinator, Zeno Ventures, Transpose Platform Management, Goodwater Capital, Formosa Capital, and Wefunder, alongside angel investors Benjamin Bryant and Kevin Moore. The pitch is direct: agents are promising, but without organizational context, they act like smart interns who don’t know where anything is. Trace wants to be the manager that gives them the right information at the right time.
A Context Graph for Agentic Work in Enterprises
Trace builds a dynamic knowledge graph from the tools enterprises already use — think email, Slack, project trackers, knowledge bases, and structured data in platforms like Airtable or Salesforce. That graph encodes who does what, how work flows across teams, and where relevant documents and records live. With that in place, a user can issue a high-level instruction (“Launch a new microsite” or “Assemble the FY27 sales plan”), and Trace decomposes the goal into a step-by-step workflow, routing sub-tasks to AI agents or human owners as appropriate.
The technical thesis is a shift from prompt engineering to what the founders call context engineering. Instead of perfecting a single prompt, Trace focuses on gathering just-in-time, task-specific context — the exact brief, data, permissions, and constraints an agent needs to act. In practice, that means agents are invoked with the right snippets from a knowledge base, the latest numbers from a CRM, and the relevant process checklists, reducing the brittle behavior that often derails pilots.
Crucially, Trace also treats orchestration as a first-class problem: it models dependencies, sets handoffs between humans and agents, and tracks progress. That matters because most real enterprise work is multi-step, multi-system, and compliance-bound. Without orchestration and visibility, even strong underlying models can produce work that’s misaligned or non-auditable.
Funding Signal in a Crowded Agent Market
The $3M seed arrives as heavyweight players move aggressively into enterprise agents. Anthropic recently introduced agent capabilities with department-focused plugins. Workplace platforms from Atlassian to Microsoft are embedding agents directly in products like Jira and Copilot. Automation leaders such as ServiceNow and UiPath are pushing deeper into AI-powered workflows. Trace’s bet is that a neutral, context-first orchestrator can sit above these tools, unify them, and prevent vendor lock-in.
Competition will be fierce. If incumbent platforms deliver sufficiently capable, context-aware agents within their own ecosystems, the surface area for an independent orchestrator could shrink. On the other hand, most large enterprises run dozens to hundreds of systems; a cross-tool brain that reduces integration overhead and enforces consistent governance has clear appeal.
Why Agent Adoption Has Stalled in Large Enterprises
Enterprises haven’t lacked model quality as much as operational fit. AI agents need identity, permissions, data provenance, and process understanding to act safely. They also need observability and escalation paths when confidence is low. Deloitte’s State of AI in the Enterprise reports have repeatedly ranked integration, data readiness, and risk management among top scaling barriers, and Gartner has emphasized governance and policy controls as table stakes for production deployments.
The opportunity remains enormous. McKinsey estimated in 2023 that generative AI could add $2.6T to $4.4T in annual economic value, and that activities representing 60–70% of employees’ time are now potentially automatable with current technologies. Yet translating that potential into durable productivity gains requires more than a powerful model; it requires system-level context so outputs are accurate, compliant, and actionable inside real processes.
Adoption data underscores the gap. IBM’s 2023 Global AI Adoption Index found that a substantial share of organizations were experimenting with AI but had not operationalized it widely, citing skills, integration, and governance as obstacles. Agentic systems magnify those concerns because they don’t just generate content — they take actions across tools.
What to Watch as Trace Scales Its Enterprise Platform
Three proof points will determine whether Trace’s approach sticks.
- First, measurable business impact: reduced cycle times, higher throughput per employee, and fewer handoff failures across multi-step workflows.
- Second, enterprise-grade controls: robust access management, auditability, and policy enforcement that let security teams sleep at night.
- Third, breadth and depth of integrations: the more systems the context graph covers — and the fresher that context — the more reliable agents become.
If Trace can convert disconnected systems into a living map of how a company works, it may turn today’s flashy demos into dependable operations. The $3M seed won’t outspend platform giants, but it doesn’t need to. In enterprise AI, the winning layer is often the one that quietly makes everything else make sense — transforming smart interns into trustworthy teammates.