Thanks for getting in touch. It sounds like there could be an opportunity to improve both your website and the way incoming enquiries are handled…
✓ Ready for human review
The system prepares the work. You make the decision.
Design + development + workflowDaniel Palfi Forms / Independent project
THE PROJECT
My own enquiry platform
MY ROLE
Journey, form, backend & automation
THE PRINCIPLE
Prepare automatically. Review personally.
01 / The starting point
The form is only the start.
A new enquiry should start a conversation, not a round of copying, pasting and working out what matters.
I built Daniel Palfi Forms to connect the customer-facing form with the work that follows: secure storage, an organised lead record, an AI-assisted assessment and a draft response for eligible enquiries.
The original enquiry remains the source of truth. The final decision stays with a person.
02 / Follow one enquiryInteractive walkthrough · Fictional sample
Same enquiry. A better starting point.
Explore the information at each stage. Nothing here is sent or processed live.
01
Useful information. From the first step.
The form asks for relevant business details, service interest, priorities, budget and timing. Questions can change with the service selected.
Validation and spam checks happen before the accepted enquiry is stored securely in D1. Confirmation follows storage.
If a later connection fails, the accepted enquiry is still there.
Daniel Palfi FormsEXAMPLE
A new enquiry.
Name
Jamie Carter
Interested in
Website + AI Automation
Budget
£2,000–£5,000
Timeline
1–3 months
“We currently have an older WordPress website and most of our enquiries come through email. We’d like something more modern and ideally a better way to organise and respond to incoming leads.”
Fictional sample · Original answers preserved
02
The context. Without the noise.
Google Sheets receives an organised lead record. The OpenAI API then reviews selected business information and prepares a summary, provisional service recommendation, potential concerns and suggested next action.
Contact details and private CRM notes are excluded from the AI input.
The assessment supports a decision. It does not replace the original message or set the price.
AI-assisted assessmentILLUSTRATIVE OUTPUT
AT A GLANCE
A website refresh. A connected follow-up.
An established business wants to modernise its WordPress website and improve how incoming enquiries are managed.
Service interest
Website + automation
Timing
Near-term
Next action
Review requirements and scope
Prepared information · Human judgement required
03
A head start. Not an automatic reply.
An internal notification highlights the enquiry. Eligible enquiries receive a personalised, unsent Gmail draft.
I review the assessment, decide the appropriate scope and price, and check the reply. The draft can be edited or discarded before anything reaches the customer.
The workflow stops at preparation. Customer communication stays under human control.
Gmail draftUNSENT
To
Jamie Carter
Subject
Re: Your enquiry
Hi Jamie,
Thanks for getting in touch. It sounds like there could be an opportunity to improve both your website and the way incoming enquiries are handled…
✓
Ready for your review.Check. Edit. Decide.
Illustrative draft excerpt · No email is sent here
Validate
Store
Organise
Assess
Prepare
Review
03 / Built to behave properly
The less glamorous bit. Just as important.
A workflow needs to be useful when things go wrong, too.
01
Keep the original.
AI-generated information sits alongside the customer’s answers. It never replaces them.
02
Don’t do it twice.
Unique references prevent duplicate imports. Processing records help avoid repeated AI analysis, notifications or draft responses.
03
Save first. Retry safely.
Accepted enquiries are stored before confirmation. Pending records can be retried, and a submission is marked as synced only after its Sheet record is confirmed.
04
Keep the human decision.
AI supports review. It does not make final pricing decisions or send customer replies automatically.
Privacy, measurement and the tools underneath
The form includes a privacy acknowledgement and Turnstile spam protection. It can capture campaign information such as UTM parameters. Optional analytics follows the visitor’s consent choice; contact details and enquiry answers are kept out of analytics events.
The form interface, service-specific questions and submission backend connect to my existing Google Sheets lead pipeline. Google Apps Script retrieves pending submissions through a private connection. Additional forms can use the same storage and integration approach, with questions tailored to each purpose.
Daniel Palfi Forms · Cloudflare Workers · D1 · Turnstile · Google Sheets · Google Apps Script · OpenAI API · Gmail
04 / Evidence, not estimatesControlled functional test · 5 October 2026
Tested with an enquiry. Not a made-up saving.
1Enquiry stored and synced
1Unsent draft verified through a separate eligibility test
0Duplicate records or actions on the repeat run
A controlled live test stored one fictional enquiry in D1, synced one lead record to Google Sheets and produced an AI assessment and internal notification. The AI recognised the enquiry as a test, so draft eligibility was exercised separately with that same record to verify one unsent Gmail draft. A repeat run produced no duplicate records or follow-up actions.
These are functional test results from my own system, not client results or measured claims about time saved.
Your process
What happens after your next enquiry?
Start with what currently happens when a customer gets in touch. We can work out what would make it simpler.
Daniel Palfi FormsEXAMPLE