06 · Retail tech
AI answers from the field guides
Service staff dug through shared drives for install and troubleshooting guides. I built an assistant inside the CRM that answers from the managed guides and articles, with citations and the right page images, scoped to one customer when it matters.
- Client
- Installs and services checkout tech for national retailers and restaurant chains
- Built with
- Salesforce
- Apex
- Lightning Web Components
- Python
- Azure
- OpenAI
Screens
Rendered from the real code with made-up data.
The chat panel answers a technician's receipt printer question with numbered steps and the guide pages and article it used.
A follow-up answer that shows the matching figure from the service guide, with its caption, page and source.
The admin page for the guides and articles the assistant can use, with each one's customer tag, synced content, last update and status.
Tagging an install guide to one project and also showing it to that project's customer.
The scope picker: all knowledge, general only, or one customer or project, with source counts.
Past conversations grouped by day, each with a preview, its customer and message count.
After a thumbs-down, a short form asks what went wrong.
What it changed
- 01
Answers come back in a strict structure, and any citation or image the model invents that was not retrieved is rejected.
- 02
Tag a guide once on a parent account and it reaches all 3,900+ of that customer's stores, and no one else's.
- 03
History is partitioned by user, so one person's conversations can never be read by another.
How it fits together
- Built by me
- Platform or outside system
- People
More of my work
See all
Contact
Need something like this built?
I take on full-time roles and focused projects. Email is the fastest way to reach me.
- Email[email protected]
- LinkedInin/seanbfoster
- LocationOrlando, FL · Remote