We've been working on something interesting at the studio: a reservation bot that actually understands what people are asking for.
The Problem with Forms
Anyone who's made a restaurant reservation online knows the drill. Click the date picker. Select a time from the dropdown. Choose party size. Fill in contact info. Maybe there's a comments box if you're lucky.
But that's not how people naturally communicate. What they actually want to say is: "I need a table for four next Friday evening, somewhere quiet because it's my wife's birthday."
Traditional online forms can't handle that kind of request. They break it into discrete fields and hope you fill everything out correctly.
The Conversational Alternative
We're building reservation functionality for RSVP-System (our open-source Drupal platform) that uses local AI to have actual conversations with guests. Here's how it works:
The bot asks natural questions, understands natural answers, and extracts the structured data it needs - date, time, party size, contact information. It handles fuzzy inputs like "next Friday evening" or "around 7-ish" and converts them into concrete reservation details.
The interesting part? It runs on local hardware using open-source models. No sending customer data to external APIs. The AI lives on your server, understands your specific needs, and gets better over time.
Human + AI: The Right Balance
Here's what makes this practical rather than just technically interesting: **every reservation goes to a moderation queue for staff review**.
The bot collects information and creates a draft reservation in Drupal's content moderation system. Staff see all the context - the special requests, the conversation notes, the specific timing. They approve it, maybe adjust something, and the confirmation goes out.
This is what we call the "centaur approach" - AI handling the tedious data collection, humans making the final decisions. Neither trying to replace the other, both doing what they're good at.
Why This Matters
We're at an inflection point with AI where it's finally practical to build tools like this without massive budgets or cloud dependencies. A few years ago, this would have required enterprise infrastructure. Today, it runs on the same server that hosts your website.
For RSVP-System users, this means reservation handling that works the way your guests actually think, with the quality control you need to run your business properly. No lost details, no misunderstood special requests, no guessing what "near the window" means.
Built in the Open
This is part of our larger RSVP-System project - a cookbook of Drupal recipes for building reservation and event management platforms. The whole thing is open source, version controlled, and built on standard web technologies.
The bot component will be another recipe in the collection: install it, configure it for your needs, and you've got conversational reservations feeding into your existing Drupal workflow.
We're developing this in public, documenting the architecture decisions, and making it all available on GitLab and Packagist. Because the interesting technical work shouldn't stay locked in proprietary systems.
What's Next
We're currently in the prototype phase - bot conversation working locally, Drupal integration established, moderation workflow configured. Next comes refinement: better natural language understanding, availability checking before submission, multi-language support.
If you're running a reservation system and interested in what conversational AI could do for your workflow, let's talk. This is exactly the kind of forward-thinking infrastructure work we do at Gluebox.
---
*Stephen Groundwater runs Gluebox LLC, a technology design studio in Philadelphia. RSVP-System is an open-source project providing Drupal-based tools for reservation and event management platforms.*