Our work
Case studies
Real generative AI projects with real business impact. Here’s what we built, how we built it, and what changed.
AI support agent for a Turkish e-commerce platform
The problem
15,000+ monthly support tickets with 70% being repetitive questions. A 12-person team couldn’t keep up, response times averaged 18 hours, and CSAT scores were dropping. The company needed to scale support without scaling headcount.
What we built
We built a RAG-powered AI support agent connected to their order database, product catalog, and return policies. The agent handles Turkish and English, resolves common issues autonomously, and escalates complex cases with full context to human agents.
What changed
- 67% of tickets resolved automatically by AI
- Response time dropped from 18 hours to under 30 seconds
- Support team refocused from 12 to 5 people on complex cases
- CSAT improved from 3.2 to 4.6 out of 5
RAG knowledge system for a legal firm
The problem
A 50-lawyer firm had 200,000+ legal documents across scattered drives and systems. Junior lawyers spent 3-4 hours per case just searching for relevant precedents and clauses. Knowledge was trapped in senior lawyers’ heads.
What we built
We built an AI-powered legal research assistant using RAG. It indexes all firm documents, case law, and internal memos. Lawyers ask questions in natural language and get answers with exact source citations and document links.
What changed
- Document research time cut from 3-4 hours to 15 minutes per case
- 85% accuracy on legal Q&A benchmarked against senior lawyers
- New associates onboard 60% faster with AI-assisted learning
- Firm estimates $400K+ annual savings in billable hour efficiency
AI content pipeline for a digital marketing agency
The problem
A 30-person marketing agency was producing content for 40+ clients manually. Each blog post, social caption, and newsletter took hours. They couldn’t scale without proportionally scaling their team and burning margins.
What we built
We built an agentic content generation pipeline: one agent researches topics, another drafts content in the client’s brand voice, a third optimizes for SEO, and a human reviewer approves the final output. Connected to their project management and publishing tools.
What changed
- Content production capacity increased 5x without new hires
- Average blog post creation time dropped from 6 hours to 45 minutes
- SEO rankings improved — 28% more pages ranking in top 10
- Agency margins improved by 35% on content retainer clients
Agentic document processing for a logistics company
The problem
A logistics company processed 2,000+ invoices, bills of lading, and customs documents daily — all manually. A 20-person data entry team worked overtime, error rates were 8%, and processing delays cost the company penalty fees.
What we built
We built an AI document processing pipeline: OCR + LLM extraction for structured data, classification agents to route documents, validation against their ERP, and exception handling with human review for edge cases.
What changed
- 85% of documents processed fully automatically
- Error rate dropped from 8% to under 1%
- Processing time cut from 48 hours to 2 hours per batch
- Data entry team reassigned to higher-value operations work
Have a similar challenge?
Let’s talkOwn an AI app? Building one?
Sell it to us, or build it with us. Write to [email protected]. You’ll have an answer within days, not months.
Get in touch