Flagship proofThe AI Factory
An enterprise AI operating model built to scale governed delivery.
Architected Centrilogic's AI Factory, a production delivery model designed to move organizations beyond POC and support multiple AI use cases per month at predictable, measurable velocity.
Multiple AI use cases per monthReusable governed architectureExecutive-ready delivery cadence
Leadership decision
What leadership changed
- Defined the AI Factory service model and delivery motion
- Designed the AI COE framework, AI Landing Zone, and AI Agent Factory
- Connected business sponsorship, governance, architecture, and team execution
Architecture choices
How the system was shaped
- Pro-code patterns on Azure OpenAI, Azure AI Foundry, Microsoft Fabric, and OneLake
- Reusable controls for data access, orchestration, and auditability
- Foundations built for parallel enterprise engagements rather than single demos
Business outcomes
What changed in practice
- Created a repeatable path from experimentation to production
- Raised confidence that AI delivery could be governed and measured
- Shifted the conversation from isolated pilots to operating capability
Azure AI FoundryAzure OpenAIGithub CopilotMicrosoft FabricOneLakeGoverned pro-code delivery
Production ClaudeGoverned Claude Skills for Private Equity
Turning 70-page CIMs into signed-off financial summaries, with controls a finance team can defend.
Architected and led the technical engagement for a production Claude skill at a private equity firm. Claude reads a roughly 70-page Confidential Information Memorandum and proposes each line item. The analyst approves, and deterministic code writes the approved figures into the firm's own template. The skill went live 15 days after kickoff. It was built with Claude Code under a skills SDLC Dave defined, then reused for a second private equity firm through configuration alone. This delivery is a core part of the production proof behind Centrilogic achieving the Select partner tier in the Claude Partner Network Services Track.
20 minutes to under 5 per Financial Summary8 analysts live in week oneLive 15 days after kickoff
Leadership decision
What leadership changed
- Defined an SDLC for multi-step Claude skills: Claude-drafted requirements, an agreed design, traceable GitHub issues, CI gates and versioned packages
- Led the delivery hackathon and finish sessions that took the skill from v0.1 to production in about two weeks
- Set "build once, parameterize per client, don't fork," which made the second PE engagement a configuration exercise rather than a rebuild
- Owned the client engagement end to end, from proposal and kickoff through go-live evidence and customer-story sign-off
Architecture choices
How the system was shaped
- Claude proposes, the analyst approves, deterministic code writes. The model never does arithmetic on a financial figure.
- Every figure records its page, its row and unit as printed, a confidence score, and who approved it and when. Anything below a 90% threshold is routed to a human.
- Runs in Claude Cowork inside the client's own Claude Enterprise tenant. Nothing is hosted by Centrilogic.
- About 35,000 lines of tests and 69 architecture decision records within 11 days. Extraction is scored against client-cleared CIMs, with a confident wrong value treated as the hard failure.
Business outcomes
What changed in practice
- Financial Summary build time cut from about 20 minutes of manual effort to under 5, and the analyst is free to do other work while it runs
- 8 analysts and associates running it on live deals in the first week
- Every number answers "which page, who approved it, and when" without a meeting
- Provided a production deployment and the public customer reference that helped Centrilogic reach Claude Partner Network Select tier (Oct 2026)
Claude CodeClaude CoworkAgent SkillsClaude EnterpriseGitHub ActionsPythonArchitecture Decision RecordsClaude Partner Network
Production agentAI-Powered Contact Centre Agent
Operational automation in a regulated customer-service environment.
Delivered a production-grade AI contact-centre agent integrating Genesys Cloud, Salesforce, and Azure OpenAI to eliminate manual after-call work and create auditable CRM records at scale.
Manual after-call work removedAuditable CRM records createdContact-centre workflows accelerated
Leadership decision
What leadership changed
- Shaped the solution around regulatory expectations and operational fit
- Aligned platform, process, and stakeholder decisions across business and technology owners
Architecture choices
How the system was shaped
- Integrated Genesys Cloud, Salesforce, and Azure OpenAI
- Embedded summarization and record generation into production workflows
Business outcomes
What changed in practice
- Automated end-to-end call summarization
- Reduced manual effort for agents
- Improved auditability in a regulated environment
Genesys CloudSalesforceGithub CopilotGoverned workflow integration
Governed knowledgeAgentic AI Knowledge Assistant Avatar
Secure access to sensitive regulatory knowledge at enterprise scale.
Delivered a secure, governed Agentic AI knowledge assistant with role-based access to regulatory content, reducing dependence on SMEs and accelerating compliance workflows.
Role-based access to sensitive contentSME dependence reducedCompliance workflows accelerated
Leadership decision
What leadership changed
- Positioned governance and access control as first-class design requirements
- Connected risk management expectations to user experience and adoption
Architecture choices
How the system was shaped
- Designed governed retrieval and access patterns for regulated information
- Protected content exposure while improving search and response speed
Business outcomes
What changed in practice
- Improved access to regulatory knowledge
- Reduced reliance on manual expert intervention
- Managed compliance risk through governed architecture
Agentic Retrieval-Augmented GenerationRole-based accessGoverned content patterns
Workforce AISales AI Coach and Enterprise Rollout
On-demand AI coaching for the full sales org, shipped in 24 hours, alongside the Claude Enterprise deployment and governance program that made it possible.
When the CentriLogic sales team struggled to hold AI conversations with clients, Dave didn't schedule training. He shipped an AI coach by the next morning. A 4-mode Copilot Agent grounded in the company's own case study library equipped 25+ sellers to navigate AI objections, match proof points to client industries, and run pre-call briefs with coached discovery questions. In parallel, he championed Claude Enterprise through Executive approval, authored the company's AI Acceptable Use Policy, and completed the Claude Partner Network certification path, the first step toward CentriLogic's Select partner tier, building the internal AI operating model CentriLogic sells to clients.
AI sales coach deployed to 25+ sellers in under 24 hoursClaude Enterprise: pilot to 100+ seats, governed by a formal AI AUPInternal AI adoption model mirrors the pattern CentriLogic architects for clients
Leadership decision
What leadership changed
- Identified the sales team's AI knowledge gap and chose to build a coaching tool rather than schedule training. On-demand, rep-specific, and built from real CentriLogic proof points.
- Built the Claude Enterprise business case, secured executive approval, and tied the decision to the company's commitment to operating as a frontier AI firm.
- Authored the company's AI Acceptable Use Policy to govern the deployment, applying the same governance pattern used in client AI COE engagements.
- Completed the Claude Partner Network certification path (4 required courses plus 2 additional), then helped carry CentriLogic from certification to the Select partner tier in the Claude Partner Network Services Track (Oct 2026).
- Applied the same AI operating discipline to CentriLogic's internal tools that the firm architects for external clients.
- This work earned CentriLogic's Frontier Award, recognizing practical, high-impact AI adoption across the firm.
Architecture choices
How the system was shaped
- 4-mode coaching framework surfaced as Copilot suggested prompts: Quick Lookup, Quiz Me, Pre-Call Coach, and Role-Play / Objection Drill
- Persistent per-rep fluency file, a markdown memory layer tracking each rep's progress across 18 AI topics and all 6 core objection types. Pattern carried over from a personal coaching agent Dave built for himself.
- Knowledge base grounded in CentriLogic's AI Factory materials, Sales Content, and Client case study materials
- Case study matching framework covering 6 client industry profiles with matched proof points and outcomes
- Built in Claude Cowork, ported to Copilot Agent Builder for org-wide deployment on the M365 surface sellers already use
- Claude Enterprise: grew from a pilot of 10 technical early adopters to company-wide adoption, with 100+ seats in use and growing
Business outcomes
What changed in practice
- Sales team able to handle 6 core AI objection types (budget, readiness, ROI, Microsoft vs. Claude, differentiation, and industry fit) with coached, case-study-backed responses
- Reps generate structured pre-call briefs with matched case studies, discovery questions, and anticipated objections in seconds rather than hours
- Monday need identified, Tuesday in use. A 24-hour deployment cycle that demonstrates the same model CentriLogic delivers for clients.
- Claude Enterprise deployed with executive approval and a formal governance policy, now adopted across most of the company (100+ seats). A governed internal platform, not shadow IT.
- Internal AI operating model mirrors the client pattern: identify the gap, deploy with structure, govern it, enable the team, and iterate.
Copilot Agent BuilderClaude EnterpriseMicrosoft CopilotAI AUPKnowledge groundingPrompt engineeringFluency memory systemAI governanceClaude Partner Network
Commercial accelerationRFP Bid/No-Bid and Response Automation
Internal agents that increase response speed without losing judgment.
Built internal Copilot agents that evaluate bid opportunities against real capabilities and draft compliant responses from an evolving knowledge base.
50-70% faster response cyclesStronger bid disciplineReduced effort for proposal teamsBetter quality and brand tone
Leadership decision
What leadership changed
- Applied product thinking to internal commercial operations
- Connected capability reality, proposal quality, and executive decision-making
Architecture choices
How the system was shaped
- Knowledge-backed automation across historical RFP content and response assets
- Draft generation paired with gap analysis and recommendation logic
Business outcomes
What changed in practice
- Accelerated response velocity
- Improved consistency and compliance in outputs
- Created leverage for presales and solution teams
Copilot agentsKnowledge indexingGap analysisResponse & SOW generation
Executive strategyEnterprise IT Roadmaps and Cloud Transformation
Architecture strategy that gives large organizations a direction they can execute.
Led IT strategy engagements for large enterprise clients, delivering roadmaps from current-state assessment through target architecture across cloud, infrastructure, and AI transformation.
Business-aligned target-state directionMulti-year transformation sequencingExecutive confidence in major change programs
Leadership decision
What leadership changed
- Worked directly with account and executive leadership on strategic planning
- Translated broad technical choices into business and program direction
Architecture choices
How the system was shaped
- Defined target architectures across Azure, AWS, and hybrid environments
- Mapped tactical migration steps from current state to target state
Business outcomes
What changed in practice
- Enabled large-scale cloud and infrastructure transformation
- Linked roadmaps to business priorities and measurable value
- Supported major renewals, extensions, and strategic decision-making
AzureAWSHybrid architectureRoadmap designExecutive planning