CASE STUDY
Scaling consulting expertise through governed AI workflows.
CLIENT
EY
PRACTICE
FOOTPRINT
FOCUS
THE CONUNDRUM
The CIO’s conundrum: a world of moving targets
EY’s Strategy and Transactions practice needed a way to apply AI to complex consulting work without losing control of its expertise, costs or quality. As models changed, the firm also needed the flexibility to choose the right capabilities for each task without building its approach around a single provider.
For teams working across M&A and private equity, the challenge was practical: translate consulting methods into workflows that could be shared, evaluated and governed. The system had to support enterprise security and reliability while giving subject-matter experts a way to define how the work should be done.
IT STARTS WITH A PROMPT
An M&A team asks Capitol to assess a company and attaches the available financials, filings, and deal materials.
MINUTES LATER
The relevant workflows are complete: company analysis, competitive context, and decision-ready outputs built from the underlying sources.
THE BET
Why Capitol AI?
EY built its approach around an orchestration layer that could work across multiple models and providers. Capitol AI provided an environment where EY experts could encode their processes, select models at each step, and evaluate the resulting workflows and outputs.
The partnership brought together EY’s consulting expertise and Capitol AI’s product and engineering team, led by cofounders Shaun Modi and Tom Hallaran. Its practical focus was to make complex workflows usable, governable and repeatable as the underlying AI ecosystem evolved.
That meant visibility into AI usage and cost, support for long-running, multi-step workflows, and native evaluations for quality control. EY’s operating knowledge could remain consistent while model choices changed.
EY, ON SHIPPING WITH CAPITOL’S FORWARD DEPLOYED ENGINEERING
BEFORE → AFTER
Six problems that stalled enterprise AI, and how they were addressed
SILOED AI DEVELOPMENT
Distributed teams build overlapping tools and duplicate work.
COLLABORATIVE AGENT CONSTRUCTION
Subject-matter experts build in a shared environment, so EY expertise can be encoded once and reused across teams.
DEPENDENCE ON ONE PROVIDER
Model performance, cost and availability can change.
INDEPENDENT MODEL SELECTION
EY can choose and automate model selection for individual workflow steps, reducing dependence on any one provider.
EXPERTISE THAT IS HARD TO SCALE
Consulting methods are difficult to translate into consistent digital workflows.
REPEATABLE, AUDITABLE WORKFLOWS
Capitol AI’s Composer lets experts define the logic behind Commercial Due Diligence, Research Synthesis and other deliverables.
LIMITED VISIBILITY INTO COST
A workflow can contain tasks with very different cost and performance needs.
GRANULAR COST CONTROL
Models can be matched to each step, with more capable models reserved for tasks that require them.
UNCLEAR OUTPUT QUALITY
Complex agentic work requires checks before results can be trusted.
BUILT-IN EVALUATIONS
Quantitative checks, data transformations and quality assessments help teams inspect workflow performance and outputs.
AUDITABILITY AND DATA CONTROLS
Retrievable logs record agent actions and decisions. Zero data retention is a founding principle of the platform.
COMMERCIAL DUE DILIGENCE
RESEARCH SYNTHESIS
CURRENT STATE OPERATING MODEL
COMPETITIVE SET BENCHMARKING
EY-GRADE DECKS & DATA VISUALIZATIONS
100s OF WORKFLOWS, ENCODED ONCE, SHARED GLOBALLY
UNDER THE HOOD
The Current State Operating Model: decades of consulting rigor, encoded as a canonical, governed workflow.
THE TAKEAWAY
Expertise that can scale
The Capitol AI–EY partnership gives EY a way to make its consulting expertise reusable across teams while retaining control over workflow logic, quality and model choice. As AI capabilities evolve, that operating knowledge remains the foundation for the work.