Transforming capital markets with an agentic enterprise knowledge AI platform

Summary

A global capital markets organization with more than 9,000 employees set out to transform how its teams access and act on enterprise knowledge. Its data was spread across more than 20 systems, including Salesforce, Marketo, Confluence, Jira, GitLab, Workday, Zendesk, and ServiceNow. That fragmentation limited visibility into lead quality, internal and external engagement history, pipeline health, project management needs, HR information, and more.

Marlabs delivered an agentic enterprise knowledge AI platform that unifies this data and lets employees work with it through a conversational AI agent that answers questions, generates reports, and searches documents. Marlabs' proprietary PromptRouter® routes each request to the most cost-effective model while keeping AI use within corporate guidelines. The platform has delivered an estimated $14M in value, and users find what they need in an average of 1.79 prompts.

Challenge

Internal teams relied on more than 20 disconnected systems to manage their day-to-day work, which left them with fragmented visibility across the organization. Critical signals sat in separate platforms without context, and pipeline quality was difficult to assess. Before employees could move forward on a task, they spent significant time researching data across tools, which slowed their work and reduced overall effectiveness.

The client needed its teams to reach trusted, contextual insight quickly enough to support fast, informed decisions. That required a single, governed layer of enterprise knowledge and a natural way for employees to ask questions of it.

Solution

Marlabs designed and implemented an agentic enterprise knowledge AI platform that lets users query, synthesize, and interpret company data through a conversational AI agent. The platform turns enterprise data into actionable intelligence that fits into day-to-day operations. Marlabs delivered it in four phases.

Enterprise Data Foundation

Marlabs unified the client's company data into a single, governed knowledge layer. The team built integrations with the client's core systems and normalized data from each source so that signals aligned across platforms. It then designed a knowledge model that gives the AI agent a consistent, trustworthy view of the enterprise.

Intelligence and Prioritization

The team added a generative AI context engine, an AI scoring model, and predictive analytics to evaluate context across systems and signals. AI-driven journey scoring helps teams prioritize, and the platform identifies the key positive and negative signals behind each score. Contextual summaries are generated at query time so that every answer reflects the most current data.

Conversational AI Agent Experience

Marlabs delivered a conversational AI agent that lets employees reach enterprise knowledge in natural language. The team implemented PromptRouter® to match each request with the most cost-effective model, and it built query orchestration and real-time retrieval of prioritized insights. Every response is explainable and grounded in enterprise data so that employees can see why the agent answered as it did.

Secure Deployment and Enablement

Marlabs deployed the platform in a secure, cloud-based environment on AWS and rolled it out through a phased adoption approach. The team put enterprise access controls and auditability in place and began with a pilot deployment for revenue teams. Training, feedback loops, and continuous optimization supported adoption as the platform expanded across the organization.

Results

The platform transformed how teams access and apply enterprise data by replacing manual research with AI-driven insight, and it has delivered an estimated $14 million in value. From July 2025 to April 2026, the platform drew 50% to 150% more engaged users than GitHub, Microsoft 365, and Zoom AI tools, which lowered the organization's overall AI cost.

Accuracy keeps that engagement high. Users find the results they need in an average of 1.79 prompts, which reduces task friction and repeat queries. Behind each answer, PromptRouter® orchestrates requests across 66 large language models (LLMs) and matches every request to the most cost-effective model. The platform has processed 48.7 billion tokens to date.

The platform now hosts 7,431 agents, including 1,609 active agents led by the general chat, code review, and invoice processing agents. Adoption has spread across the enterprise: 5,994 of 10,852 employees, about 55% of the workforce, have adopted the platform and submitted nearly 900,000 prompts. With a unified knowledge layer across more than 20 systems, teams gain clearer insight into organizational health across departments, and the client has a scalable foundation for future AI capabilities.

Impact

The platform delivered measurable value, accuracy, and cost efficiency across the organization:

• Delivered an estimated $14 million in value from lower AI cost

• 50% to 150% more engaged users than other enterprise AI tools

• Results found in an average of 1.79 prompts

• Cost-optimized orchestration across 66 LLMs, with 48.7 billion tokens processed

• 7,431 agents on the platform, including 1,609 active agents

• Enterprise-wide adoption by about 55% of the workforce, with nearly a million prompts submitted