Cases, Cody AI, AI platform, built at CodeSM Group
An AI assistant on a company's own documents, for 151,742 companies
The platform our approach comes from: turning a company's documents into an assistant that answers only from its own sources, with references.
151,742
companies served in 87 countries
Situation
Companies wanted an assistant that answers questions from their own manuals, terms and internal documents without making things up. At scale: hundreds of thousands of companies, 287,416 people a month.
What we built
- A layer that reads documents, splits them up and makes them searchable (74.6 million vectors)
- Routing across several language models by task, plan, context length and cost, with automatic failover
- Answers with source references and a human escalation route
- 112.4 billion tokens processed without customers noticing any model switches
Result
- 151,742 companies, 287,416 users a month
- Several providers behind one interface, so price and model changes did not become a project
- We now use the same building blocks for Custom builds for SMEs
Stack
Laravel, Vue, MySQL, Pinecone, AWS Lambda, multiple LLM providers
Ownership
This is work by Sergiu Neagu as founding engineer at CodeSM Group (2023 to 2025). It is shown here as proof of scale, not as a Wendig client.
Start with the Quickscan
Three minutes. You get a report with three processes, hours per week and a price. No obligation.
Or get to know us.