← Learning
Learning · Microsoft Copilot Studio

Copilot Studio

10-week practical plan to learn Microsoft Copilot Studio inside out — via a real project, not tutorials. Goal: portfolio piece + interview-ready demo. Background: BA career differentiation as AI reshapes the workforce.

● In progress PL-900 — pending Mobile Mentor target
0%
Overall progress

6 phases · ~10 weeks

Start 0% complete Interview ready
0
Phases done
6
Total phases
10
Weeks
2
Demo pieces
The vehicle

The Morning Agent

A Copilot Studio agent that runs Patrick's actual morning routine — live stock prices, news briefs, thesis recall, portfolio summary — delivered into Teams. Real daily use. Every module teaches a different platform capability.

Interview pitch

"I built a personal version of an executive intelligence agent — topic-based flows for structured queries, live external API data via custom connectors, a private knowledge base, and automated delivery into Teams via Power Automate. Here's the same architecture applied to your environment: swap my stock API for your CRM, my KB notes for your SharePoint, and the morning trigger for whatever operational cadence your team needs."

6
Learning path

Phases

01
Not started
Weeks 1–3

Topic-based bot + stock API

Build the shell. Topic-based conversation flows, entities, variables. Connect to Alpha Vantage live price API via custom connector. No generative AI yet — structured and predictable. This is the corporate-safe architecture.

Milestone Ask "what's RKLB at?" → get a live price back. Feels simple. The architecture behind it isn't.
02
Not started
Weeks 3–5

Knowledge base connection

Connect your stock thesis notes (Word docs / markdown) from OneDrive/SharePoint as a knowledge source. Topic-triggered recall — user asks explicitly, bot retrieves and returns. Bridge toward generative without going full AI yet.

Milestone "What's my thesis on RKLB?" returns a useful summary pulled from your own documents.
03
Not started
Weeks 5–6

Generative AI layer

Enable generative answers as a fallback. Topics handle scripted scenarios. Generative handles the rest via the knowledge base. Tune confidence thresholds. Now you can articulate when each mode is appropriate — the sophisticated interview answer.

Milestone Unscripted question gets a sensible answer. Out-of-scope question gets a clean "I can't help with that."
04
Not started
Weeks 6–8

Actions + Power Automate

Agent that does things, not just answers. Trigger flows, send summaries, schedule the morning brief. This is the step from chatbot to agent — and the most impressive demo capability.

Milestone 7:30am, a formatted morning brief lands in Teams without touching anything.
05
Not started
Weeks 8–9

Teams deployment + PL-900

Production-quality deployment. No rough edges. Clean error handling. Then knock out the PL-900 cert while the knowledge is fresh — it's quick, and it's what job specs filter on.

Milestone Clean 10-minute demo. PL-900 passed.
06
Not started
Weeks 9–10

Enterprise demo: IT Helpdesk Agent

Clone the architecture. Strip personal data. Rebuild as a generic IT helpdesk agent — topic-based for common IT requests, knowledge source for policy/FAQ, simulated ticket logging via Power Automate. This is the corporate demo that lands in a room.

Milestone Working IT helpdesk demo you can show anyone, anywhere.
Architecture

Personal → Enterprise mapping

Every personal capability maps 1:1 to a corporate equivalent. This is the bridge you explain in interviews.

Personal: Alpha Vantage stock API
Enterprise: CRM / ERP / custom business system API

Same custom connector pattern. Different data source.

Personal: Stock thesis notes in SharePoint
Enterprise: Policy library, SOPs, product catalogue

Same knowledge source connection. Different documents.

Personal: Morning portfolio brief → Teams
Enterprise: Executive briefing, incident summary, daily ops report

Same Power Automate flow. Different content and cadence.

Personal: Ticker-based conversation flow
Enterprise: IT request flow, HR query, onboarding steps

Same topic-based architecture. Different domain.