Ferranti: Multi-Agent Knowledge Platform for Streamlining Operations & Knowledge Access

Ferranti is a leading software provider in the energy and utilities sector. Their flagship product, MECOMS 365, renowned for its global reach and flexibility, provides utilities worldwide with tailored solutions. Given the platform’s flexibility and global reach, MECOMS 365 offers comprehensive and highly detailed documentation to support implementation scenarios across utilities worldwide. Hence, swift and precise responses to customer requirements and complex implementation scenarios are fundamental to operational success and sustained business growth.

However, the necessary knowledge for the implementation of MECOMS 365 was not efficiently accessible. Technical and functional consultants—especially junior team members—spend up to one-third of their time searching MECOMS 365’s extensive manual for detailed implementation guidelines, significantly delaying project timelines.

To achieve its strategic vision of becoming a scalable knowledge organisation, Ferranti required an advanced, AI-powered knowledge management solution focused on optimising the documentation search experience, enabling rapid and accurate information retrieval, and facilitating efficient knowledge creation and utilisation.

To address Ferranti’s knowledge access challenges, Faktion developed an advanced Multi-Agent Knowledge Platform built upon its proprietary evaluation-driven development framework. This evaluation framework forms the core foundation, ensuring continuous, measurable improvement through systematic testing, rigorous human and model evaluations, comprehensive feedback loops, and controlled experimentation.

On top of this robust evaluation framework, Faktion implemented several critical, interconnected components, uniquely tailored to Ferranti’s specific operational needs:

• AI-Driven Chat Assistant: An intelligent conversational interface dynamically interacts with Ferranti consultants and partners, rapidly retrieving precise, context-aware documentation. This assistant proactively clarifies ambiguous queries, dramatically reducing search-related delays and increasing self-sufficiency among junior consultants and external implementation partners.

• Deep Research Mode: A specialised retrieval mode employing advanced reasoning to handle complex, highly technical, or nuanced queries. This mode systematically plans and performs deeper searches, retrieving comprehensive, detailed responses essential for accurate and efficient project implementation.

• Data Processing and Knowledge Base Enrichment: A fully automated pipeline transforms raw, unstructured MECOMS documentation into structured, searchable intelligence. By converting HTML and JSON files into enriched markdown format—complete with semantic embeddings, intelligently generated metadata tags (including navigation data, related documents, summaries, and keywords)—this pipeline significantly improves content discoverability and retrieval speed.‍

Multi-Agent Architecture:

A suite of specialised AI agents collaboratively monitor, analyse, and optimise knowledge retrieval and content quality. This multi-agent approach includes:

• A Feedback Analyser Agent that systematically categorises user feedback, enabling precise identification of usability improvements.
• A Query Pattern Detector Agent which classifies user queries and identifies complex patterns, proactively surfacing emerging informational needs.
• A Retrieval Quality Monitor Agent continuously evaluating relevance, precision, and recall of retrieved documentation, ensuring consistently accurate responses.
• An Insight Generator Agent synthesises insights from multiple agents, visualises performance data clearly, and enabling targeted improvements.
• A Recommendation Agent translating insights into actionable, prioritised enhancement strategies, ensuring continuous, validated improvements.

These interlinked components, underpinned by the rigorous evaluation-driven methodology, collectively deliver a highly reliable, continuously improving knowledge-management system tailored explicitly to Ferranti’s complex operational environment. The Multi-Agent Knowledge Platform ensures Ferranti’s teams and partners rapidly access accurate, verified information, enabling enhanced productivity, reduced errors, accelerated onboarding, and improved scalability.

Faktion implemented a structured and iterative approach to develop Ferranti’s AI solution, emphasising rigorous evaluation, continuous user validation, and agile deployment methodologies.

1. Scoping & Analysis

Faktion first conducted a comprehensive strategy track through interactive workshops involving Ferranti’s stakeholders, identifying clear business goals, success metrics, and the most impactful AI use cases. This strategic foundation ensured alignment with Ferranti’s broader vision while precisely addressing operational needs for efficient knowledge management.

2. Prototyping

Based on insights gathered during scoping, Faktion built an initial working prototype of the AI-driven knowledge assistant using limited, controlled data sets. Early testing and iterative feedback loops enabled rapid identification of improvements, usability enhancements, and verification of feasibility.

3. Operationalisation

After successful validation of the prototype, Faktion fully integrated the knowledge-management system into Ferranti’s production environment, leveraging Azure AI, Azure Cognitive Search, and advanced retrieval-augmented generation (RAG) methodologies. Real-time data integration ensured immediate and measurable impact on Ferranti’s knowledge workflows. Continuous monitoring, structured user feedback cycles (leveraging tracing tools such as Langsmith and Langfuse), and iterative improvement cycles ensure ongoing enhancement of system performance, user experience, and scalability.

4. Productisation

Following initial deployment, the system has been methodically scaled across multiple Ferranti teams, use cases, and processes. Future plans include seamless integration with Microsoft Teams, migration to Ferranti’s Azure environment, expanded content coverage (Flowlogic, test cases, and release notes), and regular exploration of state-of-the-art language models to further optimise system capabilities

Since deploying the AI-driven knowledge management solution, Ferranti has realised measurable productivity improvements, accelerated project timelines, and substantially enhanced user satisfaction:

• Context precision: 88% accuracy achieved, enabling highly relevant content retrieval.
• Citation accuracy: 94% reliability ensures increased trust in system-generated responses.
• Context recall: Achieved a 75% recall rate, significantly reducing information gaps.
• User satisfaction scores: Increased notably from approximately 68% initially to consistently above 78%, demonstrating clear improvement across iterative enhancements.
• Reduced negative feedback: Dropped dramatically from 33% to just 17% between feedback iterations, confirming notable usability and quality improvements.

The system’s structured approach to knowledge retrieval has markedly decreased consultants’ reliance on senior personnel, accelerated the onboarding process, and increased Ferranti’s overall scalability and operational agility. By optimising knowledge access through intelligent automation, Ferranti has significantly enhanced its competitive positioning and readiness for future growth in the Energy & Utility sector.