Case study · Solya
An AI operating system that forecasts demand and recommends restocks, transfers and markdowns for multi-store retail
+12% margin, +7% revenue
At a glance
- Client
- Solya
- Industry
- Retail
- Services
- AI Agent Development, Predictive Analytics & Forecasting, Data Engineering & Integration, SaaS Product Development, AI Discovery Workshop
- Stack
- LLM agents, Python, Forecasting, ERP & POS integrations
- Margin · client-reported
- +12%
- Revenue · client-reported
- +7%
The challenge
- Weekly planning ran in spreadsheets, where one wrong formula meant a wrong order.
- One store was overstocked while another ran out.
- Late or gut-feel markdowns destroyed margin.
- ERP, POS and the online store showed different numbers.
- Teams feared AI that changes orders or prices without anyone checking.
Our approach
- Built a data layer merging ERP, POS, e-commerce and supplier data into one view.
- Demand forecasting with stock-out and overstock risk per product, per store.
- Restock, store-to-store transfer and markdown recommendations with margin floors and business rules.
- Approval workflow with a full audit trail, then write-back to ERP, POS and Slack.
- Role-based apps for buyers, stores and planners, plus an AI agent for plain-language questions.
Results
- Buying, transfer and markdown decisions are based on forecasts and rules, not instinct.
- Less overstock, fewer stock-outs and better margins.
- Seasonal planning becomes a review of pre-built recommendations.
- Client-reported (unaudited): +7% revenue, +12% margin and +15% cash flow over twelve months on average.
Free resource · Scorecard
The AI Use-Case Scorecard
Score retail AI ideas on value, effort and data readiness before the next build.
Related services
Services used on this project
More work
Related case studies
Phonefilter · Voice & Telecom
AI call screening on web + mobile
An AI phone assistant that screens calls and messages across web and mobile
Web + mobilePlatforms
Speech → LLMPipeline
Fuzebox · SaaS & AI
Agent ops from prototype to production
An operations console to register, run, monitor and govern fleets of AI agents
Prototype → productStage
Registry · Tracing · GovernanceScope
Moonhub · HR & Recruiting
50% faster time-to-hire
AI recruiting that sources, matches and ranks candidates against each job's real requirements
−50%Time-to-hire · client-reported
Elasticsearch + AISearch
Next step
Want results like Solya?
Book a short call with a senior engineer. No pitch deck, just a clear next step.
- We reply within 1 business day.
- NDA available on request
- You own all code from day one