Six shapes of AI agent we actually build.
An agent is a system that turns a goal into a sequence of tool calls. These are the six concrete shapes we keep shipping — each with anatomy, costs, and real examples.
Code & Integration Agent
Best for · API plumbing, schema mapping, OpenAPI client generation, internal tooling
- Generate typed clients from OpenAPI / GraphQL schemas
- Schema-map between two evolving systems
- Reverse-engineer OpenAPI specs from real traffic
Conversational Agent
Best for · Internal or customer chat grounded in your knowledge base with citations and escalation
- Customer support chat with citation and warm transfer
- Internal ops Q&A across policies, runbooks, and code
- Sales assistant with CRM context
Document Processing Agent
Best for · Invoices, contracts, receipts, and forms → structured data with confidence-tier human review
- Auto-classify and extract line items from supplier invoices
- Pull clause obligations and renewal dates out of contracts
- Process expense receipts photographed by mobile
Research & Synthesis Agent
Best for · Web and internal source aggregation with structured summaries and citations
- Competitive intelligence briefs from public sources
- Internal research summaries with source citations
- Investment / due-diligence dossiers on demand
Voice & Phone Agent
Best for · After-hours bookings, lead qualification, customer service overflow, FAQ lines
- After-hours appointment booking with calendar write-back
- Inbound lead qualification with CRM logging
- Order status, FAQ, and warm transfer to humans
Workflow Orchestrator Agent
Best for · Cross-SaaS triggers — Microsoft 365, Slack, Sheets, HubSpot, Stripe — with idempotency and approvals
- New lead → enrich → CRM → Slack alert → calendar invite
- Closed deal → Stripe → invoice → Drive folder → onboarding flow
- Stockout → supplier email → PO → ERP
Not sure which agent shape fits your workflow?
Send a short description of what you're trying to automate. We'll come back with a clear recommendation within one business day.