CLIENT DETAILS
Emerging Mortgage Services Firm in the U.S.
The client is a mid-sized mortgage services provider based in Atlanta, GA. With operations across multiple states and a growing loan portfolio, the company manages a range of customized loan programs catering to homebuyers, property investors, and condominium buyers.
The firm employs 100-150 professionals, including a team of Loan Officers who act as the first point of contact for borrowers. The client is recognized for offering flexible portfolio loan programs with complex eligibility rules and documentation requirements.
Given the nature of their work, the company deals with frequent updates to program guidelines, regulatory compliance, and strict service-level expectations. Their sales and operations teams often struggled to keep up with changing policies and respond to borrower queries efficiently.
Project Challenge
A Smarter Way for Loan Officers to Access Program Information
The client approached Flatworld.ai with a clear but complex objective: to enable its Loan Officers to quickly access the most up-to-date information on various portfolio programs and condo requirements without having to dig through static documents or rely on manual back-and-forth with internal support teams.
Inputs Provided by the Client
- A set of documents, including Portfolio Program Guidelines and Condo Requirements
- Access to their Microsoft 365 Intranet for deployment
Deliverables Requested
- A conversational AI agent that could handle questions about loan programs in natural language
- An admin dashboard with role-based controls to manage users and knowledge-base content
- A usage analytics dashboard to track performance, accuracy, and volume of agent interactions
SOLUTION DELIVERY
Building and Launching Telenok in Six Structured Phases
Flatworld.ai followed a phased approach that allowed rapid prototyping, real-time feedback incorporation, and smooth deployment on the client's existing infrastructure.
1Requirement Gathering & Planning
The team aligned with key stakeholders to define business goals, capture technical constraints, and establish the deployment environment. Clear timelines and role ownership were established at this stage.
2Design & Prototyping
A functional prototype of the AI assistant — customized on top of the Telenok framework was developed. This included setting up intents for common loan queries and drafting sample email templates for escalation.
3Development & Integration
Telenok was trained on the customer's program guidelines. To enable fast and context-aware responses, Flatworld.ai integrated vector search technology for retrieving unstructured documents. The chatbot widget was customized for compatibility with Wix and WordPress and embedded on the client's internal portal.
4Model Fine-tuning & Optimization
Flatworld's team ran prompt tuning exercises and tested Telenok with real queries from Loan Officers. The model was repeatedly refined to improve its handling of tabular data like rate charts and program matrices.
5Testing & UAT
Internal QA was followed by user acceptance testing with the client's team. Feedback was collected around response accuracy, response tone, and coverage of edge cases.
6Final Rollout & Training
After going live, Flatworld.ai provided user training sessions and shared quick-start guides. The team monitored live usage for two weeks to ensure a smooth transition and performance stability.
BEHIND THE SCENES
Key Challenges Tackled by the Flatworld.ai Team
While the project stayed on track in terms of timeline, several technical and operational complexities had to be addressed by the Flatworld team:
Tabular data handling
Much of the client's program documentation included tables with variables. Translating these into accurate, conversational AI responses required additional model tuning and structured prompt design.
Frequent program updates
The client regularly updated their program rules, which meant the AI knowledge base had to be updated on a rolling basis. Flatworld implemented a controlled admin interface to manage and publish changes in real-time.
Compatibility with existing infrastructure
Ensuring Telenok worked smoothly with the client's Microsoft 365 environment and their intranet CMS (Wix/WordPress mix) required careful widget customization and cross-platform testing.
- Telenok AI Chatbot: Flatworld’s proprietary LLM-powered assistant
- Vector Search Engine: For accurate content retrieval from unstructured documents
- Microsoft 365: Internal intranet used for deployment
- Role-based admin dashboard: in human intervention through automation.
- Analytics console: For usage trends, query accuracy, and adoption metrics
RESULTS ACHIEVED
3× Conversion, 80% Faster Responses, and Happier Borrowers
With post-deployment, the client reported significant gains in operational efficiency and sales productivity.
- 80% faster response time for common borrower queries
- 40% improvement in borrower satisfaction scores, driven by clearer, faster communication
- 3× increase in lead conversion attributed to faster turnaround on program-related questions
- Reduced internal support workload, allowing senior underwriters to focus on exception cases
Client Impact:
Real-time Access to the Right Answers
With Telenok live on their internal portal, Loan Officers ask detailed program questions and get instant, accurate responses. This replaced manual searches through static PDFs and reduced reliance on underwriting teams. Loan Officers are now better equipped to respond on the spot, improving speed, consistency, and borrower confidence during every conversation.
Takeaway
Turning Static Lending Rules into Instant, Usable Knowledge
Flatworld.ai helped the client convert complex program documentation into a live, searchable knowledge assistant. This shift not only improved operational speed but also reduced compliance risks and allowed the business to scale support without increasing headcount.
Contact Us 
Email 
Chat Case Studies
Mortgage firm cut response time by 80% and tripled conversions using Telenok AI for program queries.
Learn More
Mortgage lender automated document indexing with MSuite, achieving 80% time savings and 99% classification accuracy.
Learn More
