Ordo Ab Chao: Project Pulse
Hackathon Product Brief — Campana & Schott Author: Nicolas Wiggenhauser, PhD Date: April, 2026
An agentic AI assistant that transforms consulting chaos into clarity.
1. Product Concept & Title
Name & Rationale
The chosen title is Ordo Ab Chao: Project Pulse, a Latin phrase meaning "Order from Chaos," paired with a functional subtitle that captures the product's core value. The name carries gravitas appropriate for a consulting environment while signaling exactly what the product does: it imposes structured order on the fragmented chaos of project communication.
2. Full Product Description
What It Is
Project Pulse is a multi-source agentic AI assistant that continuously ingests live project signals from Microsoft Teams, Outlook, SharePoint, meeting transcripts, and uploaded documents. It reads, reconciles, and synthesizes this information into three always-current outputs:
- A Project Log: a chronological, searchable record of all key decisions and events
- An Action Plan: structured deliverables, tasks, owners, and deadlines
- A Timeline Diagram: a visual Gantt-style view of the engagement
Core Capabilities
Scope Drift Detection
When priorities or deliverables change across conversations, the agent flags the delta, documents it automatically, and prompts the consultant to confirm before updating the plan. No more undocumented scope changes that come back to haunt the engagement.
Narrative Coherence
Beyond aggregating data, the agent maintains a unified project narrative, so answering "where do we stand?" takes seconds, not an hour of archaeology across emails and chat threads. This is the core differentiator from generic project management tools.
Dual-View Output
Generates both a consultant-facing internal view (with full context, risks, and open items) and a client-facing view (clean, curated status update ready to share), with no internal notes or sensitive commentary included in the client version.
On-Demand Q&A
A conversational interface allows the consultant to ask "what are our open deliverables?" or "what changed this week?" and receive fast, structured, actionable answers grounded in the actual project record, not hallucinated summaries.
Key Integrations
| Microsoft 365 Core | Document Processing |
|---|---|
| Teams, Outlook, SharePoint, OneDrive: the primary data sources via Microsoft Graph API | SOW PDFs, meeting transcripts, uploaded decks and notes: via PDF parsing and OCR |
| Project Tools | Optional Integrations |
|---|---|
| Microsoft Planner, To Do: for bidirectional task sync | Notion, Jira, Confluence: for clients using non-Microsoft tooling |
Strategic Positioning
The Campana & Schott Moat: Project Pulse is built for consultants, with consulting workflows baked in, SOW awareness, deliverable tracking, scope change documentation, client-facing vs. internal views. This is something generic tools like Microsoft Copilot or Notion AI do not do out of the box. The Microsoft 365 integration angle makes it immediately deployable across Campana & Schott engagements without new infrastructure.
3. Step-by-Step Build Plan
Step 1 — Define the Data Model
Design the schema for Project Log entries, Action Plan items (deliverable, task, owner, due date, status), and Timeline events. This is the backbone everything else writes to and reads from. Key entities: Project, Deliverable, Task, LogEntry, ScopeChange, Source.
Step 2 — Build Ingestion Connectors
Connect to Microsoft 365 (Teams, Outlook, SharePoint) via the Microsoft Graph API. For the hackathon MVP, manual file upload (emails as .eml, chat exports, SOW PDFs) is a valid substitute that demonstrates the full value without requiring live API credentials.
Step 3 — Build the Extraction Agent
Use Claude (claude-sonnet-4-6) with structured prompts to extract decisions, action items, owners, deadlines, and scope changes from raw text. Output structured JSON matching the data model. The extraction prompt should be trained on consulting-specific language patterns.
Step 4 — Build the Reconciliation & Deduplication Layer
When the same task or decision appears in multiple sources, the agent reconciles them into a single canonical entry. Conflicts (e.g. two different deadlines for the same deliverable) surface as alerts requiring consultant confirmation before resolution.
Step 5 — Build the Narrative Engine
A second LLM pass synthesizes the structured data into the unified project narrative: a 3-5 sentence project summary, a bulleted current status section, and a "what changed this week" delta report. This is what makes the product feel intelligent rather than merely organizational.
Step 6 — Build the UI
Three panels: Project Log (filterable chronological table), Action Plan (kanban or structured task list), and Timeline (Gantt chart). Include a conversational chat interface for on-demand Q&A against the project record.
Step 7 — Add Scope Drift Alerts
Compare each new ingestion cycle against the current baseline SOW. Flag deviations, propose a log entry automatically, and ask the consultant to confirm before updating the plan. This creates the audit trail that protects both the consultant and the client.
Step 8 — Generate Client-Facing Export
One-click export of a curated, client-safe status update (PDF or Teams message) with no internal notes, risks, or sensitive commentary included. The export should be formatted to look like a professional consulting deliverable, not a raw data dump.
4. The Task Template
Every task and deliverable generated by Project Pulse uses this canonical template. It combines three proven frameworks: SMART (Specific, Measurable, Achievable, Relevant, Time-bound) for precision; CPQQTRD (Client, Product, Quality, Quantity, Time, Resources, Dependencies) for scope clarity; and CPORT (Context, Problem, Outcome, Rationale, Tasks) for outcome framing.
Canonical Task Record
| Field | Description |
|---|---|
| Context (CPORT) | What situation or project phase does this task belong to? |
| Problem / Need (CPORT) | What gap or risk does this task address? |
| Outcome (SMART + CPORT) | What specific, measurable result does completing this task produce? |
| Rationale (CPORT) | Why does this matter to the client and the engagement? |
| Client / Recipient (CPQQTRD) | Who is the internal or external recipient of this deliverable? |
| Owner (SMART) | Single named person accountable for delivery |
| Quality Standard (CPQQTRD) | What does 'done well' look like? How will quality be verified? |
| Quantity / Scope (CPQQTRD) | How many? How large? What is explicitly in and out of scope? |
| Due Date (SMART + CPQQTRD) | Hard deadline, or the milestone this task must precede |
| Energy Level Required | Deep Focus / Collaborative / Admin |
| Resources (CPQQTRD) | Tools, data, budget, or access needed to complete this task |
| Dependencies (CPQQTRD) | What must be true or done before this task can start? |
| Sub-Tasks (CPORT) | 1. [owner, micro-deadline] / 2. [owner, micro-deadline] / 3. [owner, micro-deadline] |
| Source (Auto-filled) | Email / Teams / SOW / Meeting: with timestamp and link |
| Status | Not Started / In Progress / Complete / Blocked / Deferred |
| Priority | Critical / High / Medium / Low — based on timeline and dependency analysis |
Energy Management Layer
Each task is tagged by cognitive load: Deep Focus (strategy, writing, analysis; schedule in morning blocks of 90+ minutes), Collaborative (workshops, reviews, client calls; schedule mid-day when energy is social), or Admin (formatting, logging, emails; schedule at low-energy times like early afternoon). The weekly planner groups tasks by energy type alongside deadline, so consultants work with their natural rhythm rather than against it. This draws from principles in Cal Newport's Deep Work and Tony Schwartz's energy management research.
5. Primary User Profile
The Campana & Schott Consultant
Mid-to-senior level consultant in the management and technology track. Runs 1-3 concurrent client engagements with meaningful complexity: multiple stakeholders, multi-month timelines, evolving scope.
| Context & Environment | Core Goals |
|---|---|
| Spends 30-60 minutes each Monday morning piecing together a coherent picture of the week from emails, Teams threads, decks, and notes scattered across multiple tools. Rarely has a single source of truth. Operates under time pressure with high client expectations. | Deliver high-quality strategic advice. Maintain client trust through consistent, proactive communication. Avoid the embarrassment of missed deliverables or undocumented scope changes. Spend time on strategy, not information archaeology. |
| Key Pain Points | Gains from Project Pulse |
|---|---|
| Scope drift goes unrecorded. Status updates take too long to assemble. A client asking 'where are we?' requires 20 minutes of prep before answering honestly. Skills are wasted on administrative reconstruction rather than strategic thinking. | Monday morning planning takes 5 minutes. Client status updates are one click. Scope changes are logged in real time with full audit trail. Strategic thinking replaces information archaeology. Client confidence increases. |
Secondary User: The Client Stakeholder
A senior manager or director at the client organization who wants visibility into the engagement without scheduling a call. They benefit from the client-facing view: a clean, jargon-free status update that answers "what's happening, what's coming, and what do I need to decide?", delivered proactively rather than on request.
User Journey (Monday Morning)
- Consultant opens Project Pulse dashboard (2 min)
- New signals from the past week have been automatically ingested and processed (0 min — runs in background)
- Consultant reviews the delta: what changed, what's new, any scope drift alerts (3 min)
- Confirms or edits the auto-generated log entries and task updates (2 min)
- Exports client status update with one click — ready to send (1 min)
- The week starts with complete clarity on priorities, owners, and deadlines (total: 8 minutes vs. 60 before)
6. Hackathon MVP
MVP Scope: One engagement, one consultant, three outputs. The MVP proves the core loop: ingest → extract → synthesise → display. Keep integrations minimal; manual file upload is fine for the hackathon. The demo should show a compelling before (chaos) and after (clarity) that any consultant in the room will immediately recognise from their own Monday morning experience.
Input: Manual Multi-Source Upload
Upload a mix of real or realistic project artefacts: 2-3 email threads (.eml or pasted text), a Teams chat export, a Statement of Work PDF, and a set of meeting notes. This simulates the real chaos without requiring live API integrations: the value is in the output, not the plumbing.
Output 1: Project Log
A chronological table of decisions, key moments, and scope changes extracted from all sources. Each entry shows: date, event type (Decision / Action / Risk / Scope Change), one-sentence summary, and source document with link. Filterable by date, type, and source. Auto-generated, human-editable.
Output 2: Action Plan
A structured task list using the SMART + CPQQTRD + CPORT template. Each task is auto-populated with title, owner, due date, energy tag (Deep Focus / Collaborative / Admin), dependencies, and source. The consultant can edit any field before the plan is finalised. Exportable as a formatted table.
Output 3: Timeline Diagram
A Gantt-style visual showing deliverables and milestones across the engagement horizon, auto-generated from the Action Plan with no manual input required. Color-coded by status (not started / in progress / complete / blocked). Exportable as an image or embedded in the client status update.
Bonus Feature: On-Demand Q&A
A simple chat interface where the consultant can ask "what are the open items?" or "what changed last week?" and receive an answer grounded in the actual project record. This is the most impressive demo moment; the live Q&A interaction makes the intelligence of the system viscerally clear to the audience.
Recommended Tech Stack
| AI Layer | Frontend |
|---|---|
| Claude API (claude-sonnet-4-6) for extraction, reconciliation, narrative generation, and Q&A. Structured JSON output enforced via system prompt. | React + Next.js for the three-panel dashboard. Mermaid.js or D3.js for the Gantt timeline. Tailwind for styling. |
| Data Layer | Ingestion |
|---|---|
| Supabase (PostgreSQL) or SQLite for the project data model. Simple REST API connecting frontend to AI processing layer. | PDF.js for SOW parsing. Email parser for .eml files. Plain text processing for chat exports and meeting notes. |