AI strategy & operating models
What should we build, and how will it run?
Vision, guardrails, and multi-year roadmaps built with executives, then the operating model to make delivery real.
Roadmaps in 90 days · hub-and-spoke · use-case discoveryFrancisco Paniagua · AI & Digital Transformation Leader
From pilots to production.
Twenty years building Microsoft-based technology practices across energy, financial services, consulting and SaaS. I work where AI has to scale safely: operating models, governance, and adoption that hold up in regulated environments, with practical agentic AI architectures tied to real work.
AI strategy, adoption, and agentic architecture
Chapter 01
The Shift
Most engagements sit somewhere on the path from we’re experimenting to this is how we work now. The work is turning ambition into a practical operating model, adopted workflows, and production-ready AI.
The experiment
Lots of teams can prove a concept.
The harder question is whether the organization has the roadmap, governance, ownership, and adoption model to make AI useful beyond the demo.
Production
This is where I spend my time.
AI strategy, operating models, enablement programs, and agentic AI architectures that hold up in regulated environments and connect directly to measurable work.
Chapter 02 — Two Milestones
Pilots prove possibility.
Production proves capability.
AI only matters when it changes how work gets done.
Chapter 03
Proof in Numbers
$0M
Saved through enterprise automation in oil & gas
0
Automation & low-code solutions managed in production
0
Modern work users across M365 & Power Platform
0
Citizen developers enabled through governance & training
0
Labor hours automated
0
Years turning emerging tech into reliable capability
Chapter 04
What I Do
Three things I’m brought in to do.
One production outcome.
Build the strategy and operating model, make adoption real, and design agentic AI use cases that are useful, governable, and ready for real enterprise work.
Adoption & enablement
Can people change how they work?
Enablement programs, cohort-based learning, champion models, manager follow-up, and adoption metrics that turn AI from a tool launch into a new way of working.
AI literacy · exec workshops · champion programs · adoption metricsAgentic AI use cases & architecture
Where do agents actually make sense?
Use-case shaping, workflow architecture, evaluation, controls, and agent designs that respect data, permissions, orchestration, and human review.
Agent use cases · Copilot Studio · workflow architecture · evaluation & controls
Chapter 05 — The Method
From zero to an AI-powered organization.
Frame
Assess readiness, business value, permissions, content quality, ownership, and the operating conditions AI needs before it scales.
Govern
Set practical guardrails, governance, risk controls, and decision rights so teams can move without creating unnecessary risk.
Adopt
Run enablement around real work: cohorts, champions, manager reinforcement, usage signals, and habits that stick.
Scale
Identify high-impact workflows, build responsible agents and automations, measure ROI, and expand what works.
The pattern repeats: assess the foundation, make adoption real,
then scale the
workflows that create measurable value.
Chapter 06
Selected Work
From a central bottleneck to a self-sustaining automation practice
Designed an enterprise enablement model that moved delivery from centralized to hub-and-spoke, stood up a Center of Enablement with training, intake, and compliance standards, and put governance and quality gates around a portfolio that grew past 2,500 bots and apps.
$2.4MCost saved
23K+Labor hours automated
2,500+Solutions in production
700+Developers enabled
An agentic underwriting assistant for lender-ready decision materials
Designed the business architecture for an AI agent that turns dense borrower files, appraisal documents, property records, and opportunity context into draft underwriting artifacts. The workflow finds the deal folder, analyzes documents, enriches collateral data, drafts credit memo and appraisal addendum content, and keeps credit decisions with the lending team through human review.
3Decision moments supported
CMCredit memo generation path
AAAppraisal addendum generation path
Chapter 07
Insights
Notes from the field on scaling AI past the pilot — and the occasional panel.
Article · Microsoft Copilot
Careful what you wish for: one month working with Copilot Cowork
What changes when AI stops being a chat window and starts operating across the repetitive, connected work that fills the week.
Read the post →
Article · Agentic AI governance
Agent 365 is now live
Why enterprises need a control plane for observing, governing, and securing agents across Microsoft 365, Copilot Studio, and beyond.
Read on LinkedIn ↗
Article · AI adoption
Becoming Customer Zero
Three actions for making AI adoption real inside the team first — the same workflows and adoption muscle recommended to clients.
Read on LinkedIn ↗
Article · AI readiness
The illusion of AI readiness
Ambition is not the same as readiness — data, governance, and operating discipline still decide whether AI creates value.
Read on LinkedIn ↗
Article · AI & product
Enabling the next generation of product managers
Co-creating an AI Product Management certification course, and what new PMs need to understand about generative AI.
Read on LinkedIn ↗
Event · Platform Calgary
Technical vs. non-technical product managers
Panelist at Product Calgary on a debate that won’t die — and where the line genuinely matters for people building products.
View the panel ↗
Chapter 08
About
Twenty years turning emerging technology into capability organizations can rely on.
I lead enterprise AI and automation work focused on one thing: moving emerging technology out of the experiment phase and into how work actually gets done — across financial services, energy, technology, and consulting.
That has meant leading transitions from centralized delivery to hub-and-spoke enablement, standing up centers of enablement, and partnering with executives to line AI investment up against real business value and real risk — with a strong bias toward delivery in regulated environments.
Before the AI work, I co-founded and ran product and SaaS delivery practices — including full P&L ownership. That grounding in shipping production software for paying customers still shapes every engagement: the demo is the easy part.
- Background
- B.Sc. Computer Science · Software Engineering
- Based in
- Calgary, Alberta
- Sectors
- Energy · Finance · SaaS · Mining · Retail
- Latest
- Team Lead, AI & Modern Workplace · Convverge
- Languages
- English · Spanish (native)
Credentials
AI Transformation LeaderMicrosoft Certified · 2026
Create
Agents in Copilot StudioMicrosoft Applied
Skills · 2026
Security & Compliance for M365 CopilotMicrosoft
Applied Skills · 2026
Copilot Business Value — ProficientMicrosoft GCPS ·
2025
AWS
Solutions ArchitectAmazon Web Services
Machine
Learning FoundationsFor Product Managers ·
Duke University
AI Transformation LeaderMicrosoft Certified · 2026
Create
Agents in Copilot StudioMicrosoft Applied
Skills · 2026Security & Compliance for M365 CopilotMicrosoft
Applied Skills · 2026Copilot Business Value — ProficientMicrosoft GCPS ·
2025AWS
Solutions ArchitectAmazon Web ServicesMachine
Learning FoundationsFor Product Managers ·
Duke University
Microsoft 365 & Copilot·
Copilot Studio·
Azure AI·
Power Platform·
Power Automate·
Blue Prism·
Claude Code·
Codex·
Microsoft 365 & Copilot·
Copilot Studio·
Azure AI·
Power Platform·
Power Automate·
Blue Prism·
Claude Code·
Codex·
Chapter 09 — The one we write together
Building an AI capability, or hiring someone to lead one?
I’m open to leadership-level roles and advisory work in AI and digital transformation — particularly in energy, financial services, and technology.
Prefer LinkedIn?
Connect on LinkedIn ↗