Wise Monks

Introduction
Development of a tailored, multi-agent, enterprise conversational system built on the Agno agentic orchestration framework, integrating internal systems (CRM, warehouse, BI, documents) to empower sales, purchasing, management teams to access scattered information, collaborate and make decisions more effectively.

Challenges

• Lestena had large amounts of enterprise-data spread across systems (sales reps’ CRM data, warehouse stock levels, BI dashboards, internal document archives) and needed a unified layer to quickly answer business-questions (e.g., “What’s our stock of hoki from Chile?” / “What’s the margin trend for cod from Norway last quarter?”).
• Different roles (sales, purchasing, management) each required tailored access and conversational workflows rather than being forced into generic dashboards.
• Opportunity to reduce time spent manually searching in multiple systems, improve proposal-preparation speed, and support management with clearer insight and actionable conversations rather than raw tables.

Solution

Designed a role-aware conversational layer:

• Sales-agent: assists reps with quick queries (“What’s available for order?”, “Show me last 3 months’ top-selling product in region X”)
• Purchasing-agent: gives purchasing team real-time view of warehouse, supplier lead times, upcoming needs
• Management-agent: summarises BI dashboards, gives trend alerts (“Margin dropped by 5 % in product Y”)
• Built on the Agno orchestration framework:
• Defined agents, tool-connectors and routing logic
• Integrated APIs with systems: CRM, warehouse system, Microsoft Power BI (for analytics) plus document-repository

UI/UX considerations:

• Clean chat-interface (desktop & mobile) with context-aware UI elements (e.g., quick-buttons for common queries)
• Role-specific dashboard entry points → if sales logs in they get chat + “Recent leads” summary; management sees “Recent alerts” + conversation entry
• Visual summaries in conversational replies (graphs, quick cards) rather than pure text

Impact
Lestena’s teams can now ask natural-language questions, get structured responses, and move faster in sales, purchasing and management decision-making.

Introduction

The client had accumulated a large repository of human-written credit opinions, each containing expert evaluations, financial analysis, and risk conclusions. Their goal was to transform this knowledge base into an automated system capable of generating accurate, consistent, and regulator-friendly credit opinions for new businesses based on financial statements.
Wise Monks delivered a full AI-powered solution that replicates expert-level reasoning while significantly reducing manual workload.

Challenges

• Thousands of legacy credit opinions stored in mixed formats and writing styles.
• Highly domain-specific terminology and regulatory structure.
• Need for consistent, unbiased output aligned with the client’s internal methodology.
• Strict accuracy and explainability requirements.
   

Solution

We built a complete AI pipeline optimized for financial decision-making:

✅ 1. Knowledge Base Extraction

Parsed and cleaned the entire credit opinion archive, isolating relevant sections (analysis, ratios, risks, recommendations).

Converted into structured datasets for training.

✅ 2. Custom LLM Fine-Tuning

Fine-tuned a large language model using the client’s internal credit opinions as ground truth.

The model learned:

Consistent structure and tone

Domain vocabulary

Analytical reasoning patterns

Risk evaluation logic

✅ 3. Automated Credit Opinion Generator

Developed an application where the user uploads financial data (balance sheet, P&L, ratios, cash flow).

The AI produces:

Complete credit opinion

Summary of the company

Key strengths and weaknesses

Risk assessment

Final recommendation

Clear reasoning and citations from data

✅ 4. Internal AI Agent Workflow

An agentic system evaluates financial statements, calculates ratios, and supplies them to the LLM with standardized prompts for maximum accuracy.

Impact

• 80–90% reduction in analyst time per credit opinion.
• High consistency across all generated reports.
• Significantly improved turnaround for new credit assessments.
• Regulator-aligned structure, matching internal documentation standards.
• Seamless integration into the client’s internal tools.

Introduction
We partnered with R-SM, a Vilnius-based sales and marketing agency with deep expertise in real estate (over 30 projects) and 360° business growth services, to implement identitycall.ai—a biometric AI platform for call transcription, speaker identification, emotion analysis, and performance scoring. As project developers, our team customized the system for R-SM's high-volume client consultations, enabling automated insights into sales calls with virtual numbers. This delivered enhanced personalization and efficiency across their services, from market analysis and strategic planning to real estate project sales and communication strategies.

Challenges
R-SM faced inefficiencies in manual call reviews for their dynamic sales processes, including difficulties in detecting client emotions, identifying needs during consultations, and ensuring consistent QA across teams handling real estate developments, marketing campaigns, and 360° growth plans. With a focus on real-time results in competitive markets like Vilnius, Kaunas, and the Lithuanian coast, legacy methods led to biased scoring, missed upsell opportunities in client interactions, and up to 40% longer review times, impacting their ability to scale services and achieve measurable ROI for clients.

Solution
In close agile collaboration with R-SM's marketing, sales, and operations teams, we leveraged identitycall.ai's features such as biometric voice ID, 98% accurate multi-language transcription, emotion/intent detection, and AI-driven QA rubrics tailored to their custom needs (e.g., compliance in real estate sales and communication plans). We managed full integration, including API exports to their CRM for client profiling, batch uploads for existing recordings, and real-time alerts for coaching—deployed within a 3-month timeline using secure, GDPR-compliant protocols to align with their 360° service model covering strategic planning, product development, and sales funnel optimization.

Impact
The system reduced review times by 80%, boosted sales conversion rates by 35% through personalized recommendations based on detected client needs and emotions, and improved QA accuracy to 97% with zero-bias scoring. R-SM achieved record-breaking results in projects like "Upės horizontai" and "Pušų parkas," with 25% more identified upsell moments in calls, enabling faster growth for their clients and a 40% increase in overall campaign ROI while maintaining their leadership in Lithuanian real estate and marketing services.

This structure keeps the description engaging, professional, and focused on value. Customize it with real details from your projects to make it authentic. If you upgrade to Sortlist+ (€129+/month), you can unlock more visibility features like top directory placement and lead opportunities.