+ case study
SimplyBox
AI unified inbox for the Meta ecosystem
- year
- 2025
- type
- Product, competition
- role
- Problem structuring, architecture, full build
- status
- shipped
Six apps to answer one message. SimplyBox unifies the Meta ecosystem into one inbox, grounded in the company’s own knowledge by RAG. Response time down 80%+. Top 7, Llama AI Accelerator.
01+ the problem
What they came in with
Support teams answering customers across the Meta apps were switching between inboxes all day. The brief was speed: answer faster.
02+ what we found
What it actually was
Speed was the symptom. The real constraint was that answers were not grounded: every agent answered from memory, so the same question got different replies. A faster inbox would only have produced wrong answers sooner.
03+ the build
- 01
One inbox
Every Meta channel lands in a single queue, so an agent works one screen instead of several apps.
- 02
Their knowledge, ingested
The company’s own documents become the knowledge base, so the system knows their products and policies, not the internet’s.
- 03
Grounded answers
Replies are retrieved from that knowledge at answer time (RAG), so they cite the company’s material and hallucination drops.
04+ architecture
Written down, so your team can extend it
05+ the result
- 80%+
Response time
Cut for support teams running SimplyBox
Measured against the client's own pre-launch baseline
- Top 7
Llama AI Accelerator
The only active student team to advance
Meta Llama AI Accelerator, 2025 cohort standings
Next case study
02RoyaleCard ArenaOn-chain competitive strategy game
Have a problem shaped like this one?