# Agent Launch in 8

> Launch a custom enterprise AI agent in eight weeks with fixed scope, fixed price, deployment options, evaluation, integration, and handover.

Source: https://8fde.ai/offers/agent-launch-in-8

---

8 weeks · Fixed scope · **$75,000**

[Get an Agent→](https://calendar.app.google/ie9WbzuCCNpf6SF58)

## From idea to launch in 8 weeks

- 01 Define Weeks 1–2 Workshops, use case, scope, and architecture
- 02 Build & integrate Weeks 3–6 Agent development, data access, integrations, evaluation
- 03 Launch Weeks 7–8 UAT, deployment, and handoff

A production-ready agent, integrated, deployed, and handed off in eight weeks.

Weeks 1–2 confirm the users, workflow, constraints, and success measures, then define the target architecture and lock the delivery scope.

Weeks 3–6 validate data and access, build the agent and guardrails, connect required systems, and run evaluation and tuning.

Weeks 7–8 complete UAT and release checks, finalize runbooks, deploy the agent, observe initial use, and enable its owners.

## Deployment

Azure Databricks SaaS

Hosted in MS Foundry Agent Service

**Azure AI Agent**

**User** A business user starts the workflow and reviews the agent response.

**AI Agent** Custom agent A custom agent hosted in MS Foundry Agent Service coordinates model access, knowledge, tools, and responses.

**MS Foundry Models** Approved models Provides approved model deployments selected for the agent workflow.

**Azure AI Search** RAG Retrieves permission-aware enterprise knowledge to ground the agent response through retrieval-augmented generation.

**MS Foundry Toolbox** Tools & actions Connects approved MCP servers, OpenAPI operations, functions, and agent-to-agent tools.

**MS Foundry Observability** Trace & evaluate Captures traces and evaluation signals through MS Foundry observability and Application Insights.

**External Tools** APIs & systems Connects approved APIs, operational systems, and business tools to the agent workflow.

### Databricks deployment: Databricks AI Agent

Hosted in Databricks App.

- **User** — A business user starts the workflow and reviews the agent response.
- **AI Agent** (Custom agent) — A framework-based agent hosted in Databricks Apps coordinates model access, retrieval, Unity Catalog-governed data, and approved external tools through Unity Catalog connections or MCP Services.
- **Unity AI Gateway** (Models & policies) — Unity AI Gateway controls access to approved foundation models, endpoints, and model policies.
- **AI Search** (RAG) — Retrieves relevant governed knowledge to ground the agent response.
- **Unity Catalog** (Governed data) — Provides governed tables, volumes, functions, permissions, discovery, and lineage for the agent workflow.
- **MLflow** (Trace & evaluate) — Tracks traces, evaluation results, and production-quality signals for the agent.
- **External Tools** (APIs & systems) — Connects approved APIs, operational systems, and business tools to the agent workflow.

### SaaS deployment: LangChain + LangSmith Stack

Hosted in LangSmith Deployment.

- **User** — A business user starts the workflow and reviews the agent response.
- **AI Agent** (Custom agent) — A LangGraph or LangChain agent hosted in LangSmith Deployment coordinates models, knowledge, tools, state, and responses.
- **LangSmith LLM** (Approved models) — Connects the agent to approved model APIs selected for the business workflow.
- **Vector store** (RAG) — Supports retrieval-augmented generation with a provider selected during enablement, such as Azure AI Search, pgvector, or another approved service available in the target environment.
- **Tools & integrations** (MCP, APIs & A2A) — Connects approved MCP servers, APIs, functions, and agent-to-agent integrations.
- **LangSmith** (Trace & evaluate) — Captures traces, evaluates behavior, manages datasets, and monitors the deployed agent.
- **External Tools** (APIs & systems) — Connects approved APIs, operational systems, and business tools to the agent workflow.

## What’s included

### Define

- Prioritized use-case definition
- Solution architecture document

### Build & prove

- Working AI agent
- Evaluation benchmark and results

### Launch & enable

- Deployment package and operating runbook
- Handover and enablement session