The Business Challenge
The business situation the organisation faced before the project began.
Inefficient communication driven by high reliance on individuals rather than data-driven systems. Due to the lack of centralized data, the organization was unable to perform data validation, analysis, or accurate forecasting. This created substantial operational bottlenecks, where significant time and cost were wasted waiting for valid information, ultimately leading to lost business opportunities and decreased operational productivity.
Why AI Was the Right Solution
Why AI was the right approach — and what alternatives were considered.
Prior to Virgo, the organization faced inefficient communication driven by an over-reliance on individuals rather than data-driven systems. The lack of centralized data resulted in:
- An inability to perform effective data validation, analysis, or accurate forecasting.
- Significant operational bottlenecks across HR, Sales, Project, Outsource, and Training teams due to time wasted waiting for valid information.
- Data fragmentation, leading to high rates of manual communication errors.
The scale and variability of the organizational data made deterministic rules engines unworkable. A solution was needed that could reason over unstructured feedback and adapt dynamically to shifting business requirements. By integrating manual workflows—such as timesheet collection and project status updates—into a familiar conversational interface, the solution bridged the gap between complex ERP systems and everyday user needs.
How the Solution Was Delivered
Discovery, design, development, testing and rollout — the journey, not the tooling.
- 1The delivery process focused on organizational alignment and seamless integration:
- 2Discovery & Stakeholder Alignment: Mapping the "2025 Expectations" across all divisions (Sales, HC, Delivery, Finance, etc.) to prioritize high-impact workflows.
- 3Data Audit and Gold-set Creation: Standardizing timesheet collection processes (Jira and Non-Jira) and integrating them with project databases.
- 4Prototype and Shadow-mode Rollout: Implementing a Telegram bot as an intelligent interface between users and the PostgreSQL/pgvector database.
- 5Assisted-mode Launch: Automating reminders (H-3, H-day) and approval notifications for various clients, including Telkomunikasi Indonesia , KSEI, SMBC, Kereta Api Indonesia , OCBC, OSS go id , etc.
- 6Continuous Evaluation: Utilizing RAG AI LLMs for skill normalization and dynamic CV generation based on historical timesheet and project data.
Key Technical & Architecture Decisions
Architecture, model selection, workflow and trade-offs.
- Centralized Communication: Telegram serves as the single source of truth for operational messaging and real-time updates.
- Retrieval-Augmented Approach: Leveraged
pgvectorandPostgreSQLto ensure AI responses are grounded in current, accurate company policy. - Technology Stack: Utilized
n8nfor workflow automation, powered by advanced models (Gemini 2.0 Flash and DeepSeek V3) for high-quality reasoning and processing.
Challenges & How They Were Solved
Obstacles hit along the way and how they were overcome.
- Integration Complexity: Differing workflows across clients (e.g., Telkom vs. KSE) were managed by creating a standardized backend pipeline while maintaining a flexible, user-friendly frontend bot interface.
- Data Quality: Addressed the issue of outdated or inaccurate talent profiles by integrating RAG-based skill normalization, which extracts insights directly from verified project histories rather than static forms.
User Adoption & Change Management
How users responded — training, change management and feedback loops.
Adoption was driven by aligning the bot’s capabilities with specific departmental needs. For instance, the Sales team benefited from lead tracking and budget alerts, while the Finance team gained real-time visibility into project cost realization. By simplifying complex ERP interactions into simple Telegram commands, the friction associated with new software adoption was minimized.
Business Outcomes & Impact
The measurable outcomes from the underlying AI Deployment — with the story behind them.
• Communication & Data Processing Efficiency: Reduced data validation and waiting time by 70% through centralized Telegram integration. • Operational Bottleneck Reduction: Improved cross-departmental response rates and project status tracking by 60% across HR, Sales, Project, Outsource, and Training teams. • Information Accuracy: Reduced manual communication errors and data fragmentation by 80% by establishing a unified ERP data pipeline. • Time-to-Decision: Accelerated business decision-making and lead processing time by 50%, enabling faster response to market opportunities.
Business Outcome Categories
The story behind the numbers
By shifting from fragmented, individual-reliant communication to a centralized Telegram-driven ERP pipeline Virgo fundamentally transformed how our teams operate daily. The 70% reduction in data validation and waiting time eliminated the frustrating administrative drag that previously stalled cross-departmental workflows, When the HR, Sales, Project, Outsource, and Training teams saw a 60% boost in response rates and project tracking, it fostered a cultural shift from reactive firefighting to proactive execution. Furthermore, cutting manual communication errors and data fragmentation by 80% meant that leadership and operational managers finally trusted a single source of truth for decision-making. This clarity directly accelerated our time-to-decision and lead processing by 50%, enabling the organization to respond swiftly to market opportunities while redirecting precious employee hours away from tedious coordination toward higher-value business growth.
Lessons Learned
What surprised the team, what worked well, and what would be done differently.
- Alignment First: Technical automation is only as good as the business requirements it serves; mapping departmental expectations is the prerequisite for success.
- Trust is Earned in Silence: Start with subtle, helpful automations (such as H-3 reminders) to build user trust before transitioning to more complex, autonomous decision-making agents.
Future Opportunities
Where this solution could go next.
- Multi-agent Handoff: Scaling to include autonomous agents that can automatically trigger targeted coaching plans for employees.
- Closed-loop Learning: Feeding outcomes from coaching and project results back into the system to enhance future talent profiling and skill-to-opportunity mapping.
- Comprehensive Workflow Expansion: Extending the bot to handle internal loan SOPs, legal PKS reviews, and automated internship application processes.
Professional Reflections
Guidance for another builder tackling a similar problem.
"Invest disproportionately in the evaluation set before touching the prompts." The success of this project was rooted in the deep stakeholder alignment conducted early on. By documenting and addressing specific departmental "Expectations for 2025," the team ensured that the AI deployment delivered immediate strategic value rather than just technical novelty.