How DXC Built a GenAI Workforce Agent for 130,000 Consultants Using Lakebase

DXC Technology, a Fortune 500 IT services company with 130,000 consultants across more than 70 countries, needed a faster way to keep workforce skills data up to date. With Databricks and Lakebase, DXC deployed a GenAI-powered competency agent that analyzes signals across enterprise systems and delivers personalized recommendations to 130,000 consultants. Processing time fell 94%, talent matching improved and DXC now has a governed foundation for scaling AI across the business.

Outgrowing fragmented systems for real-time workforce intelligence at scale

For a global IT services company that staffs thousands of client engagements, knowing what your people can do is a competitive requirement. DXC Technology employs roughly 130,000 people across more than 70 countries, and matching the right talent to the right project depends on accurate, current competency data. In practice, that data lived in Workday profiles that consultants updated manually and infrequently, creating a growing gap between what people could actually deliver and what their profiles reflected.

DXC set out to close that gap with an AI agent that could analyze signals from: 

• Workday 

• Project systems 

• Learning records 

• Certifications 

• GitHub activity 

• Job descriptions

Using these signals, the agent could then generate recommendations consultants could review and push back to Workday. The first version, built on an Azure stack, assembled a skills dictionary per consultant through sequential table reads, API calls and Azure Cosmos DB writes. It created a single view, but not at a usable pace. Processing recommendations for 1,000 consultants took approximately two hours, and the storage layer made enterprise-wide adoption impractical.

Scale was only part of the problem. With data ownership split across IT and business teams, engineers spent too much time reconciling sources and data scientists struggled to find trusted inputs.

"People didn't know what data was available or who to go to for help," said Mallikharjuna Prasad, Data and AI Leader at DXC Technology. "By the time we reached the limits of our previous stack in performance, scalability and storage, it was clear we needed a fundamentally different foundation."

DXC needed a unified, governed data platform for LLM workflows, plus an operational data layer that could serve model outputs to live applications without introducing another system.

Lakebase connects GenAI workflows to the consultant experience

DXC rebuilt the competency agent on the Databricks Platform, with Lakebase as the operational core. The team replaced an API-heavy prototype with native services that could ingest enterprise data, govern it through Unity Catalog and run the full GenAI workflow on a single platform.

How DXC built the competency agent on Databricks

The rebuilt competency agent works across four connected steps:

1. Unify enterprise data: The platform first pulls together data from across the enterprise, combining consultant profiles, assignment history, learning records and engineering activity into a single environment through Auto Loader and Lakeflow Jobs. 

2. Identify the most relevant competencies: From there, the system needs to figure out which skills actually matter for each consultant. DXC's workforce spans thousands of roles, and the company tracks more than 4,000 possible competencies. Rather than asking the model to reason over all of them, Vector Search narrows the field from more than 4,000 competencies to roughly 200 per consultant through semantic similarity, grounding every recommendation in pre-approved frameworks and reducing the chance of fabricated suggestions. 

3. Extract signals from unstructured job descriptions: Meanwhile, much of the richest skills data sits locked inside unstructured job descriptions. Document Intelligence including ai_extract and ai_query parse that text into structured skill and certification signals the model can actually use. 

4. Serve governed model outputs at scale: For the model itself, the team chose an open-source Llama-family model and deployed it on Model Serving, which gave them governed, autoscaling inference under Unity Catalog and the flexibility to swap model sizes after evaluating quality and cost trade-offs at 130,000-consultant scale.

Lakebase ties it all together, storing preprocessed data, LLM outputs and operational logs so consultant-facing applications can instantly update and retrieve information. Consultants receive personalized recommendations, validate them through automated forms and see approved changes reflected in Workday in near real time.

"Lakebase bridged the gap between analytical data processing and real-time application needs," Malli said. "It made the system more responsive and much easier to manage."

From static profiles to dynamic workforce intelligence

What once took two hours for 1,000 consultants now completes in about seven minutes, a 94% improvement that lets DXC run the competency agent across 130,000 consultants at production scale.

The quality of recommendations improved alongside speed:

• The average number of competency recommendations per consultant rose from three to eight 

• Vector search eliminated out-of-scope suggestions 

• Nearly one third of consultants actively use the solution to update their Workday profiles 

• Managers now have cleaner data for staffing and planning 

• Consultants receive relevant guidance they can validate for themselves

All models, vector indexes and data tables are registered in Unity Catalog, with inference tables logging every model call and capturing cost, latency and failure metrics surfaced through Databricks dashboards. Built-in evaluation tools score the agent's reasoning during development, and online sampling catches quality issues in production with automated alerts.

DXC has already applied Databricks capabilities to other agentic AI solutions, connecting agents to internal enterprise tools while maintaining centralized governance. With Lakebase serving as the operational bridge between analytics and applications, the company sees a clear path to more advanced AI agents and tighter integration between models and business workflows.

"Our teams used to spend their time reconciling systems. Now they're building intelligent solutions at scale. We see a clear path to more advanced agentic AI, and this platform is central to everything that comes next," said Malli.