to follow-through.
Explore the stages of the implemented CRM workflow. This diagram contains no customer data.
A website form validates the request and passes the agreed fields through a Cloudflare Worker to the private CRM API. The browser never receives the shared server token.
Phylia CRM & AI workflows
An independent technical project demonstrating connected workflows for inquiries, proposals, customer responses, invoicing, and recurring follow-up.
- The problem
- Customer requests, correspondence, quotes, and follow-up need a consistent path through a small business.
- My contribution
- I developed the website and connected CRM workflows using AI-assisted tools, working through configuration, integration, testing, and documentation.
- A deliberate decision
- AI can prepare a proposal, but a person reviews and approves outbound actions. Generated content never becomes a customer commitment automatically.
Explore the implementation
- A public intake Worker validates submissions and forwards known fields to an authenticated private API.
- PostgreSQL stores inquiries, pipeline stages, quote responses, invoice states, and audit events.
- AI proposal drafts use structured output, validation, editable text, and local fallback templates.
- Inbound email enters a review workflow. Scheduled jobs handle reminders and suppress duplicate failure notifications.
- Private receipt storage, migration scripts, focused workflow checks, and deployment notes support ongoing maintenance.


