// work

Engagements we've shipped.

Every engagement runs to a measurable outcome. Numbers are real; client identities are anonymized per MSA unless explicitly released.

// case 01 · Hospitality · Fortune-5009-week engagement

8.2s → 480ms p95

7-step Dialogflow flow → 4-agent LangGraph topology

scope · 1,400+ properties · 78 markets · 22 languages

  • Multi-Agent Mesh
  • Platform Migration
  • LangOps & Guardrails
We expected a chatbot upgrade. We got an orchestration layer the rest of the org now builds on.
VP Engineering · attribution withheld per MSA
// beforelegacy stack
p95 latency
8.2s
fallback-to-human
31%
integration surface
14 disconnected APIs
// afterafter lorematics
p95 latency
480ms
fallback-to-human
4%
integration surface
1 agentic pipeline
// case 02 · P&C Insurance · Tier-1 carrier14-week engagement

47% → 8% manual triage

Adjuster intake queue → multi-agent FNOL pipeline with deterministic policy checks

scope · $2.4B annual claims volume · 11 lines of business

  • LangOps & Guardrails
  • Enterprise Integration
  • Hybrid Workforce
The guardrail layer is the deliverable we'll show to auditors. The latency drop is the deliverable we'll show to ops.
Head of Claims Technology · attribution withheld per MSA
// beforelegacy stack
manual triage rate
47%
median FNOL → assignment
6h 12m
policy-check violations / 1k
23
// afterafter lorematics
manual triage rate
8%
median FNOL → assignment
9m
policy-check violations / 1k
0
// case 03 · Vertical SaaS · series C8-week engagement

3.8x CSM coverage

Manual QBR prep + outreach → agentic CS pipeline with human-approval gating

scope · 12,400 enterprise accounts · 6 product lines

  • Multi-Agent Mesh
  • Hybrid Workforce
Every CSM now has an agent that drafts the work. The human is the approval node, not the assembly line.
Chief Customer Officer · attribution withheld per MSA
// beforelegacy stack
accounts per CSM
180
QBR prep time
4.5h
renewal-risk flags missed
~22%
// afterafter lorematics
accounts per CSM
680
QBR prep time
12m
renewal-risk flags missed
<3%
// case 04 · Global manufacturing · NYSE-listed11-week engagement

Agentforce ↔ LangGraph hybrid

Native Agentforce inside Service Cloud · LangGraph mesh for cross-system workflows the platform couldn't reach

scope · 4,200 Service Cloud users · 19 countries · 7 product divisions

  • Platform Migration
  • Multi-Agent Mesh
  • Enterprise Integration
We kept Agentforce where it earned its keep, and put LangGraph on the orchestration tier. The platform team and the AI team finally agreed on the architecture diagram.
Director of Service Operations · attribution withheld per MSA
// beforelegacy stack
service ticket → resolution
18h median
agents needed in 19 markets
1 per market
cross-system flows leaking outside Salesforce
11
// afterafter lorematics
service ticket → resolution
47m median
agents needed in 19 markets
1 federated topology
cross-system flows leaking outside Salesforce
0 — orchestrator handles

// pipeline

Pipeline open for 2026 Q3 engagements.