I'm Building an AI Project Manager That Handles 80% of My Admin - Part 2

Getting back hours of time with AI in my workflow

Sep 15, 2025

Last week I told you about building an AI Project Manager that handles the boring 80% of admin work. The theory resonated with a lot of you.

But concepts are cheap. Implementation is where the magic happens.

So today, I'm waling you through how it works. The story starts with a document and ends with a fully-managed project. No switching between apps. No copy-pasting. No human project tracking.

Let’s get into it.

From Vague Requirements to Project Setup in Minutes

Here’s a mock project that illustrates the time-saving power of Claude connected to your SaaS tools via MCP.

A brief specification for an AI project arrives in your inbox following initial client discussions. In the old world, this would trigger a cascade of manual admin:

Read through the document, extract key requirements
Create a Confluence page to brief the team
Set up an Asana project with all the discovery tasks
Check if there were any related emails I'd missed
Schedule team meetings, create calendar reminders
Chase down the RFP deadline (somehow missing from the document)

Easily 2-3 hours of setup work before we even started thinking about the actual proposal.

With my AI Project Manager? 12 minutes from document to project setup.

Let me show you how.

Confluence Documentation in 60 Seconds

The specification document is pretty dense with technical detail, even if it’s only a few pages. It contains technical requirements, success metrics, integration specs, and evaluation criteria. The kind of document that requires careful analysis and clear communication to the team.

I simply told Claude to write up a Confluence page for distribution to the team.

What happened:

Claude analysed the entire RFP document
Extracted all key requirements and constraints
Created a structured Confluence page with sections for Executive Summary, Technical Requirements, Success Metrics, and Next Steps
Added strategic assessment and risk factors
Included specific action items for the team
Posted it directly to our Software Development space

Time saved: 1-2 hours of reading, analysis, and documentation work compressed into 60 seconds.

Project Setup in 5 Minutes

Next challenge: translating this opportunity into an actionable project structure.

The specification had complex requirements spanning technical assessment, competitive analysis, partnership strategy, and cost modeling.

I asked Claude: "Now create a new Asana project for this opportunity and also create the initial to do tasks according to the information we have."

What Claude created:

A properly named project: "ACME Manufacturing - AI Knowledge Management System RFP"
Comprehensive tasks covering the entire response cycle
Realistic timelines working backward from a October 15th deadline
Detailed task descriptions with specific deliverables
Priority sequencing (urgent deadline confirmation first)
A project status update summarizing the opportunity

The tasks weren't generic. They were specific to this RFP:

"Evaluate integration experience with Wonderware MES, IBM Maximo, SAP S/4HANA"
"Research competitors: IBM, Microsoft, GE Digital, Siemens"
"Design hybrid cloud architecture with NIST compliance"

Each task included detailed action items. The competitive analysis task listed specific companies to research. The technical assessment task broke down AI/ML capabilities, security compliance, and manufacturing system integration experience.

This wasn't just task creation—it was strategic project planning.

Time saved: 3-4 hours of project scoping and task breakdown work done in 5 minutes.

The Missing Email Investigation

When we’re setting up a project, it’s always important to check your comms. What information do we have lurking in the our email that’s been going back and forth?

I ran a quick check.

Claude searched my Gmail systematically:

First for any emails from ACME Manufacturing
Then for recent RFP-related emails
Finally for emails from “Chris"

Found it: An email from Chris sent this morning about the "ACME Project" with a deadline for September 18th—exactly when the technical assessment task was due.

This wasn't just email search. It was investigative work that connected the dots between the RFP opportunity, missing timeline information, and internal coordination.

Time saved: 30 minutes of inbox archaeology done in 2 minutes.

The Immediate Re-Plan

This is where it got interesting. Claude found a critical email from this morning that I'd missed: the proposal deadline was September 18th - just 3 days away, not the extended timeline I'd assumed.

Claude immediately:

Updated all Asana task priorities
Changed due dates to reflect the urgent deadline
Added 🚨 URGENT flags to critical tasks
Created a final submission checklist
Posted a red status update to the project

In one search, Claude identified a business-critical deadline that would have been disastrous to miss. It then intelligently reprioritised the entire project structure.

We then drafted an email response.

The Compound Effect

Here's what makes this powerful: it's not just about individual time savings. It's about momentum and quality.

In the old world, by the time I'd manually set up the project structure, written the team brief, and found the relevant emails, half my morning was gone.

The team would wait until tomorrow's meeting to get oriented. The proposal response would start with a day of lag.

With the AI Project Manager:

12 minutes after receiving the specification, we had a complete project setup
The team could immediately see priorities and task assignments
We identified the critical missing deadline information upfront
The Confluence brief provided instant context for anyone joining the effort

This isn't just faster—it's qualitatively better.

The AI caught nuances in the RFP that I might have skimmed over in manual processing. It structured the work more comprehensively than I would have in a rushed setup.

What's Actually Happening Here

Three things make this possible that weren't available two years ago:

1. Document Intelligence Claude can read, analyse, and extract strategic insights from complex documents. Not just keyword matching—actual comprehension of business requirements, technical constraints, and competitive positioning.

2. Multi-System Orchestration Through MCP connectors, Claude operates across Confluence, Asana, and Gmail as naturally as you'd switch between browser tabs. It's not screen scraping or API hacking—it's native integration.

3. Contextual Decision Making The AI understands project management concepts: what makes a good task breakdown, how to sequence dependent activities, why deadline confirmation is critical for proposal responses.

Combined, these capabilities create something that feels less like automation and more like delegation.

The Bigger Pattern

What you're seeing here applies far beyond RFP responses:

Sales Intelligence: "Parse this customer requirements document and set up a deal progression project in HubSpot."

Product Intelligence: "Analyse user feedback from Zendesk and create feature request tickets in Jira with priority scoring."

Marketing Intelligence: "Review competitor blog posts and create content response calendar in Asana with draft briefs."

Same pattern: AI brain + existing tools + structured instructions = hours of admin work compressed into minutes.

The Reality Check

This isn't magic. There are constraints:

You still need to guide the AI with clear instructions
Complex decisions require human judgment
System integrations occasionally need troubleshooting
You can't delegate strategic thinking—only execution

But for the 80% of admin work that's mechanical? This changes everything.

For now, consider this: How much of your weekly admin could be delegated to an AI that never gets tired, never loses context, and works at machine speed?

The answer might surprise you.

Until the next one,

Chris

Upgrade to Paid