Business ConsultingClaims Teammate
Reads bills of claims, validates line items, scores fraud and drafts settlements.
Agentic AI teammatesMeet your new teammates.
Agents that own a workflow, ask when it matters and learn every time. Together, they plan, delegate and deliver.
What is an AI teammate?
An AI agent with a job description.
One workflow. One KPI. A person on every call that matters.
SeeKnowAssistOptimizeLearnBuild trust
AgentOwns one workflow end to end
SkillsClassifiers, extractors, recommenders
Data setsEnterprise data, documents, feeds
KPIsThe outcome it is measured on
Platform functionsML, CV, NLP and GenAI services
Human engagementGates, reviews and escalation
How a teammate gets work done
Trigger, perceive, reason, act, ask, learn
1
Trigger
An event, schedule or request: a new claim, a camera frame, a ticket, a log spike.
2
Perceive
Read documents, images, video, voice, logs and system data.
3
Reason
Neuro-symbolic reasoning: LLMs and predictive models, checked against rules learned from your de-identified data.
4
Act
Call APIs, update systems, draft responses, raise alerts.
5
Human gate
Pause for approval whenever a decision carries risk.
6
Learn
Feedback retrains models and improves the next decision.
Teammate catalog
Nine teammates. Ready to hire.
Referral Teammate
Matches employee referrals to open roles and screens candidates.
CX & Lending Teammate
Tracks CSAT, triages tickets and pre-qualifies loan requests.
Vision Inspector
Watches conveyor belts and flags defects and safety breaches in real time.
Fleet Teammate
Monitors vehicles, driver alertness and maintenance needs.
Energy Analyst
Reads equipment log data and finds energy waste and anomalies.
Brand Listener
Measures share of voice and sentiment across social and news.
Player Teammate
Predicts churn and personalizes offers for real-money and freemium games.
Legal Teammate
Analyzes contracts, summarizes cases and drafts documents.
Agentic teammate swarm
One teammate is good. A swarm is better.
Big work needs a team. So do agents.
Single teammate
One agent, one workflow
- Good for focused tasks
- Limited context and skills
- One point of failure
Crew
Specialists with a shared goal
- Roles, goals and tasks per agent
- Sequential or manager-led delegation
- Agents check each other's work
Swarm flow
Crews, rules and people
- Triggers, routing and shared state
- Many crews run in parallel
- Human gates and full audit
IntelliMesh swarm architecture
Experience
Vera WorkspaceSwarm Studio (visual)Python SDKREST APIsMCP ServerChat & Voice
Flows
Event TriggersShared StateRoutersParallel BranchesHuman GatesSchedules
Crews
Manager AgentSequential ProcessHierarchical ProcessConsensus VotingDelegationCrew Templates
Agents & tasks
Role · Goal · BackstoryTask & Expected OutputGuardrailsContext PassingAsync TasksRetries
Tools & skills
ML & Intelligent ML ModelsVisionNLP & OCRConnectorsBrowser & RPAMCP Tools
Memory & knowledge
Short-TermLong-TermEntity MemoryKnowledge / RAGVector StoreKnowledge Graph
Models
Multi-LLM RouterPrivate LLMsFine-Tuned SLMsEmbeddingsSpeechVision Models
Swarm modes
Four ways to work together.
Sequential
Each agent completes its task and passes context to the next.
Best for document processing, onboardingHierarchical
A manager agent plans, delegates to specialists and validates results.
Best for claims, investigations, researchParallel consensus
Several agents work the same problem; a judge reconciles answers.
Best for fraud scoring, coding, diagnosis supportEvent-driven flow
Triggers start crews; routers branch on results; gates pause for people.
Best for plant alarms, fleet incidents, churn signalsSwarm Studio
Drag. Connect. Deploy.
No code for business teams. Full code for engineers.
COMPONENTSAgentsTasks & CrewsModels & LLMsIntelliMesh ToolsMemory & KnowledgeData LoadersRouters & LogicHuman GatesOutputs
TriggerNew claim in inbox
Extractor AgentOCR + NLP tools
Policy AgentKnowledge: policy docs
Fraud AgentIntelligent Machine Learning risk model
RouterRisk > threshold?
Adjuster GateHuman approval
Settlement AgentDrafts payout letter
OutputCore system + audit
LearnCorrections retrain the extractor
Playground: chat with the swarm. Step through every agent's reasoning. Trace every call and its cost. Test models before go-live.
Deploy as REST APIExpose as MCP serverPublish to Vera WorkspaceExport as Python codeSwarm control plane
See everything. Control everything.
Every run traced. Every step governed.
Discover
- Mine process logs and tickets for automation candidates
- Rank by effort, value and readiness
- Baseline KPIs before go-live
Build
- Swarm Studio canvas or Python SDK
- Crew templates per vertical
- Test in the playground
Govern
- RBAC, SSO and audit trail
- Approval gates and intervention mid-run
- PII redaction and policy hooks on every LLM and tool call
Optimize
- Tracing with full cost accounting
- Evaluation on accuracy and KPI impact
- Swap models at runtime; roll back safely
Observability and evaluation integrate with tools such as Datadog and Arize.
Teammate swarm library
40 swarms. Eight industries.
Start with one crew. Prove one KPI. Scale to a swarm.
BFSIBill-of-ClaimsEmployee ReferralCSAT & LoanKYC & AMLInvoice Exceptions
HealthcarePrior Auth & ClaimsDenials & AppealsPatient SchedulingCare EngagementClaims Fraud
TelecomNetwork EnergyChurn RescueTicket TriageField SafetyNetwork Ops
GamingPlayer RetentionIntegrity & FraudResponsible GamingKYC & Age CheckLiveOps Insights
ManufacturingConveyor VisionGauge ReadingHazard WarningPredictive Maint.PPE Compliance
FleetFleet OperationsDriver AlertnessRoute & FuelPort OperationsVehicle Scheduling
Energy & UtilitiesShare of VoiceWater TheftDemand ForecastPower BiddingAsset Monitoring
Legal & eGovCase & ContractLegal ResearchPolicy AnalysisCitizen ServicesDocument Drafting
Start the 10·20·30 Pathway