Skip to content

Latest commit

 

History

14 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Business Loan Onboarding & Verification — Workshop

A hands-on, end-to-end workshop for building an agentic business-loan onboarding process on Camunda 8. You will orchestrate parallel verification checks, an AI Agent operating within defined process boundaries, a CRM update, and a customer outcome email — all driven by a BPMN model.

This workshop spans two repositories — each works on its own:

Repository What it teaches
Camunda workflow — you are here Orchestrate an agentic business process on Camunda 8.
Langflow — no-code AI agent builder Build no-code AI agents, independent of any workflow engine.

Learning environment. Use synthetic test data only. Do not enter real customer data, production credentials, or regulated information at any point in this workshop.


About this workshop

This repository accompanies the code-along webinar “Build a No Code AI Banking Agent.”

Session Build a No Code AI Banking Agent (free to join)
When Monday, June 29, 11 AM ET
Presenter Anjali Jain — Enterprise AI Architect at Metro Bank, CTO at Erdos Research, and Senior Tutor in AI & Machine Learning at the University of Oxford. Author of AI-Assisted Programming for Web and Machine Learning and the forthcoming Enterprise Architecture in an Agentic World.

What you will take away

  • How to build an AI agent using no-code tools for financial-services use cases.
  • Where AI agents add value in banking and other regulated environments.
  • How to architect agents that meet performance, compliance, and reliability requirements.

AI agents are transforming financial services, but building them in a regulated environment requires careful design, strong controls, and knowing where no-code tools can safely accelerate delivery. This workshop combines speed with responsibility: the BPMN model keeps the AI agent inside explicit, auditable boundaries.


Do I need the integrations?

No. A mocked process variant is provided so you can run the entire example end-to-end with no external integrations — no Companies House, Salesforce, or email accounts. The Companies House, CRM, and email steps are replaced by built-in simulations, so you only need a free Camunda account and an OPENAI_API_KEY for the AI Agent.

Two skills in this workshop also stand entirely on their own:

  • Build no-code AI agents in Langflow — a no-code agent builder that works independently of any workflow engine.
  • Design and run agentic process flows in Camunda — workflow orchestration with an AI Agent inside auditable boundaries.

The external integrations — Companies House, Salesforce, and email — are only needed if you want to run the full variant against real systems instead of the mocks.

Goal What you need
Learn a no-code AI agent builder Langflow on its own — no workflow engine required
Run the full example with no integrations A free Camunda account + OPENAI_API_KEY — use the mocks BPMN
Run the example against real systems Camunda plus Companies House, Salesforce, and email — use the full BPMN

Two BPMN variants are provided in camunda/processes/: business-loan-onboarding-verification_demo_mocks.bpmn (mocked, recommended starting point) and business-loan-onboarding-verification_demo.bpmn (full, live integrations).


Table of contents


What you will build

                Application received
                        │
                        ▼
            Create or reuse CRM Lead (Salesforce)
                        │
                        ▼
              Two checks run in parallel
        ┌───────────────┴───────────────┐
        ▼                               ▼
 Company registry check        AI document analysis
 (Companies House)             (Camunda AI Agent / Langflow)
        └───────────────┬───────────────┘
                        ▼
            Set a controlled verification route
        ┌───────────────┼───────────────────────────┐
        ▼               ▼                           ▼
   Hard fail       Passed / exception          AI Agent selects a
        │          → AI Agent next action       permitted next action
        │               │
        ▼               ▼
   Update CRM + send applicant outcome email + complete

The BPMN model controls the path at all times. The AI Agent can only recommend or trigger actions that are already defined in the process.

Process name: Business Loan Onboarding & Verification Provided BPMN files: business-loan-onboarding-verification_demo_mocks.bpmn (mocked — no integrations) and business-loan-onboarding-verification_demo.bpmn (full — live integrations)


Learning objectives

By the end of this workshop you will be able to:

  1. Deploy and run a process on a Camunda 8 SaaS cluster.
  2. Configure Connector Secrets and reference them safely from a process.
  3. Run parallel service tasks and combine their results into a decision.
  4. Use a Camunda AI Agent within explicit, auditable process boundaries.
  5. Integrate a CRM (Salesforce) and an email channel.
  6. Test happy-path, exception, and escalation scenarios, and inspect them in Operate and Tasklist.

Repository structure

business-loan-onboarding-workshop/
├── README.md                     # This guide
├── LICENSE
├── docs/
│   ├── prerequisites.md          # Accounts and tools you need before you start
│   ├── troubleshooting.md        # Common errors and how to resolve them
│   ├── glossary.md               # Key terms used in the workshop
│   └── regional-notes.md         # Adapting data sources by region (e.g. US)
├── camunda/
│   ├── README.md
│   ├── docs/saas-setup-guide.md  # Set up, deploy, and run on Camunda 8 SaaS
│   └── processes/                # Provided BPMN variants (mocks + full) and forms
├── integrations/
│   ├── README.md
│   ├── companies-house/setup-guide.md
│   ├── salesforce/setup-guide.md
│   └── email-gmail/smtp-app-password-setup.md
└── workshop/
    ├── README.md
    └── business-loan-flow-guide.md   # Full walkthrough and test scenarios

Prerequisites

A short summary is below. See docs/prerequisites.md for the full checklist, supported versions, and account sign-up links.

Requirement Why you need it
Your own Camunda 8 account You sign up yourself — no account is provided. A free trial covers this workshop
Your own Camunda 8 SaaS cluster You create this yourself in Camunda Console
Modern web browser Use Console, Web Modeler, Operate, and Tasklist
The provided BPMN and forms Included in camunda/processes/ — you upload them into your own Camunda project
An OpenAI API key Required by the AI Agent in both variants
Integration accounts Only for the full variant (Companies House, Salesforce, Gmail) — not needed for the mocks variant
Windows or macOS Only if you run the optional local Langflow track

You create your own Camunda account and cluster. A free trial (signup) cluster is sufficient.

The Camunda process files (two BPMN variants and three forms) are included in this repository under camunda/processes/. You create your own Camunda project and upload them before deploying.


Getting started

git clone https://github.com/CodeDaim0n/business-loan-onboarding-workshop.git
cd business-loan-onboarding-workshop
  1. Confirm you meet the prerequisites, including your own Camunda account.
  2. In Camunda Web Modeler, create a project and upload a BPMN variant from camunda/processes/ together with the three .form files. Start with business-loan-onboarding-verification_demo_mocks.bpmn to run with no integrations.
  3. Follow the Camunda 8 SaaS setup guide to create your own cluster, add the OPENAI_API_KEY secret, and deploy.
  4. Work through the business loan flow guide and run the test scenarios — the mocks variant runs end-to-end on its own.
  5. (Optional) To use real systems, complete the integration guides and deploy the full variant instead.
  6. Build no-code AI agents in the standalone Langflow track — usable on its own or connected to this process.

Learning path

Follow the steps in order. With the mocks variant you can go straight from step 1 to step 5 — the integration steps (2–4) are only needed for the full variant.

Step Topic Guide
1 Platform setup, upload, and deploy Camunda 8 SaaS setup
2 (Full variant only) Company registry check Companies House setup
3 (Full variant only) CRM Salesforce setup
4 (Full variant only) Email channel Gmail SMTP setup
5 Process walkthrough Business loan flow guide
No-code AI agent builder (standalone) Langflow track

Camunda Connector Secrets

Reference secrets in connector fields with {{secrets.SECRET_NAME}}.

Mocks variant — create just one secret:

Secret Used for
OPENAI_API_KEY AI document analysis and AI Agent tasks

Full variant — also create the integration secrets:

Secret Used for
COMPANY_HOUSE_KEY Companies House company and director checks
SFDC_DEMO_BASE_URL, SFDC_DEMO_CONSUMER_KEY, SFDC_DEMO_CONSUMER_SECRET Salesforce OAuth
INBOUND_MAIL_USER, INBOUND_MAIL_PASSWORD, INBOUND_MAIL_SERVER, INBOUND_MAIL_ADDRESS Outgoing (SMTP) email
INBOUND_MAIL_IMAP_SERVER, INBOUND_MAIL_IMAP_PORT, INBOUND_MAIL_FOLDER Incoming (IMAP) applicant-response listener

The provided process calls OpenAI directly; it does not use LANGFLOW_API_KEY. Add a Langflow secret only if you adapt the process to call a Langflow flow.

Full details are in the Camunda setup guide.


Security and safe data

  • Store every credential in Camunda Connector Secrets — never in BPMN XML, process variables, form fields, screenshots, or shared documents.
  • Never commit .env files, API keys, passwords, or client secrets to Git.
  • Use synthetic test data only. The credit-score step in this process is simulated and is not a real credit-bureau integration.
  • Rotate any credential that is accidentally exposed.

Troubleshooting and FAQ

Most setup and runtime issues are covered in docs/troubleshooting.md, including connector errors (401/403/404/422), secret resolution, deployment problems, and human-task visibility.


Regional notes

This workshop uses a UK company registry (Companies House) for the verification step. If you are running it in another region (for example, the US), see docs/regional-notes.md for what to use as-is and how to swap or mock the registry check.


Companion track — Langflow (no-code agent builder)

Want to learn a no-code AI agent builder that is independent of any workflow engine? Use the Langflow track. It stands on its own — build, run, and iterate on agents with just a model provider — and can optionally connect to this Camunda process through Cloudflare Tunnel and the REST connector.

langflow-local-setup-workshop


Reference documentation


Star and contribute

If you like this repo, give it a star :) — it helps others find it.

If you find issues, please help others by providing corrections :) — open an issue or a pull request.


License

See LICENSE. These materials are provided for educational use as part of the workshop.

About

Hands-on workshop: Business Loan Onboarding with Camunda 8, Langflow, and integrations

Resources

Stars

6 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors