MEAMA: a content factory and a multi-agent system that runs production
Client:
MEAMA
Segment:
Coffee and tea capsules, manufacturing
Result:
29 AI agents in 6 groups
Scale:
80+ SKUs, 85,000+ customers, 3 markets
Direction:
Content factories and automation
The task
MEAMA makes coffee and tea capsules at its own factory in Georgia. The brand has 80+ SKUs, 85 thousand customers in its base, 5 Shopify stores, vending, coffee shops and HoReCa, plus 7 international awards, including the Red Dot Design Award 2024.
The company is entering the Russian market and moving its management to AI. Sales reports were assembled by hand in Excel, and the factory needed a single demand forecast for Georgia, Russia and Germany.
What we did
- Designed the AI orchestra: 29 agents in six groups. Customers and CRM, growth, brand, content and channels, production and R&D, finance. Each group has its own human lead.
- Built the system on Paperclip: every agent has a role, a manager and an area of responsibility. The CEO agent sets priorities, and director agents distribute tasks and collect reports.
- Built a content factory for 15–30 content units a day: agents write copy in each channel’s tone of voice, make graphics within the brand system, edit Reels, TikTok and Shorts, and adapt one piece for every platform.
- Added a check before publishing: the brand agent matches every piece against the guidelines, and the quality scoring agent learns from examples approved by the brand director.
- Connected customers: profiles of 85 thousand buyers with LTV and churn risk, RFM segments, and a post-purchase sequence on days 1, 14, 30, 60 and 90.
- Linked sales to the factory: agents merge demand from Georgia, Russia and Germany into one forecast, track stock, and plan production and raw material purchases.
- Designed a data pipeline to replace manual Excel: 5 Shopify stores and the Fina accounting system, loaded into Supabase and Power BI dashboards.
- Finance agents calculate margin for every SKU, forecast sales for 30, 60 and 90 days, and raise an alert when the plan lags by more than 15% or a cost item exceeds the budget by more than 10%.
- Prepared a partnership proposal for a major marketplace: production, awards, products and a roadmap for growth in Russia through 2028.
Outcome
Instead of manual routine MEAMA runs an AI orchestra: 29 agents handle content, ads, customers, production and finance, while people set goals and approve decisions.
The rollout goes in stages: data first, then marketing, SEO and CRO, support and finance. Content, SEO, finance and CRO run with human approval until data quality is proven.
