AI Productivity System · Live system

How I cut dashboard
delivery from
2 weeks to 2 hours

A Fractional CDO's AI productivity system. Six real workflows that replaced hours of manual work with single commands.

6
Workflows
in production
10x
Faster
delivery
50+
Data
engagements

Free · 6 copy-paste prompts · Notion page opens immediately after signup

or explore the system below ↓
$ /morning-brief Calendar: 4 meetings today Gmail: 2 urgent from David Jira: DATA-91 due today Slack: 3 mentions in #data Morning brief saved to Daily/

Data leaders lose
60% of their time to operational work

Writing SQL, prepping for meetings, reconciling numbers between systems, managing tickets. The strategic thinking gets squeezed into whatever time is left. I built a system that automates almost all of it.

Six workflows that run
my entire practice

Each one replaces an hours-long manual task with a single command. All of them run every day in my own setup.

01

The Context Library

"AI without context is ChatGPT. AI with your full project context is a second brain."

Most people talk to AI like they talk to Google: ask a question, get a generic answer. My approach is different. For every project I build a context library: metrics definitions, standards, past decisions, stakeholder preferences. Claude reads all of it before every action.

The core is CLAUDE.md, a 224-line instruction file covering communication style, note-saving rules, connected services, workflows for every task type. This is not a prompt. It's an operating system.

80+ helper scripts 37 memory files 10+ connected services
nick-os/ # Personal Operating System ├── CLAUDE.md # 224 lines of instructions ├── Daily/ # Auto-generated morning briefs ├── agency/ │ └── projects/ │ └── Client/ # Client project context │ ├── Meetings/ # Processed transcripts │ └── INFO.md # Project bible ├── skills/ # Automated workflows │ ├── morning-brief/ │ ├── prep-meeting/ │ └── process-meeting/ ├── docs/ # Plans, playbooks └── templates/ # Reusable formats
┌──────────────┐ Claude Code (Brain) └──────┬───────┘ ┌─────┬───────┼───────┬──────┐ ▼ ▼ ▼ ▼ ▼ Calendar Gmail Slack Fathom Jira Things Notion LinkedIn Obsidian Upwork
02

Morning Intelligence

"A complete picture of your day in 30 seconds instead of 45 minutes."

Every morning I run one command: /morning-brief. Claude queries my calendar, email, Things tasks, Slack mentions, Telegram messages, and Jira tickets in parallel. In 30 seconds I have a structured brief: what's urgent, what's important, what can wait.

Before this: open Gmail, scroll through Slack, check Jira, review Things, look at the calendar. At least 30-45 minutes every morning just to understand what's happening. Now those 45 minutes go straight to deep work.

45 min → 30 sec 6 sources Every morning
# Morning Brief — April 17, 2026 🔴 Urgent Revenue workbook review with George (11:00) 3 unread from David re: Q1 metrics 📅 Today 11:00 George — Revenue Workbook 13:00 LinkedIn Weekly — Dasha 15:00 Boris sync — deploy review Tasks DATA-91: Revenue query analysis (due today) Review dbt PR from Vini 💬 Highlights Slack #data: Chris raised question on trial dates Upwork: 2 new messages from Wirelane
03

Meeting Autopilot

"2-minute prep. 1-minute post-processing. Four systems updated at once."

Before a meeting, Claude gathers context: who's attending, what we discussed last time, open tickets, recent emails from participants. The output is a PREP document with key talking points and questions.

After the meeting, one command triggers a chain: Fathom transcript gets processed into an Obsidian note, action items go to Things, the deal in Pipedrive gets updated with a note, and relevant Jira tickets get comments with decisions made.

When a long Slack thread turns into something actionable, Claude summarizes the discussion and creates a Jira ticket with full context, tagged to the right assignees.

2 min prep 1 min processing 4 systems updated Zero copy-paste
Chain 1: Meeting Prep /prep-meeting "George Revenue" ├── Calendar meeting context ├── Gmail recent threads ├── Obsidian past meeting notes └── Jira open tickets 📄 PREP document saved Chain 2: Post-Meeting /process-meeting Fathom transcript ├── Obsidian (notes) ├── Things (action items) ├── Pipedrive (deal update) └── Jira (ticket comments) Chain 3: Slack → Jira "Summarize the #data thread" 85 messages ├── Summary (decisions, blockers) └── Jira ticket DATA-94 created
04

Dashboards on Demand

"From metric catalog to interactive dashboard in 2 hours. Not 2 weeks."

A client needed a unit economics dashboard. The old way: write SQL queries, test them, configure Looker Studio, wait for it to load. Minimum one week, realistically two.

Now I give Claude the project's metric catalog (definitions, formulas, data sources) and describe what to visualize. Claude generates the SQL, builds an interactive HTML dashboard with Chart.js, and I verify the numbers. The HTML version is faster, more customizable, and requires no BI tool license.

2 weeks → 2 hours No BI license Fully customizable
> Build a unit economics dashboard using docs/metric-catalog.md Data: BigQuery Include: CAC, LTV, churn by cohort, MRR waterfall Claude generates: ├── 12 SQL queries (validated) ├── Interactive HTML + Chart.js ├── Cohort analysis tables └── Export to PDF
05

Data Reconciliation

"Found missing credit notes in 10 minutes. Manually, that's a full day."

In any business, data lives across multiple systems. The CRM says one thing, accounting says another, the warehouse has a third version. Traditional reconciliation means Excel files, manual cross-checks, and a lot of frustration.

Claude connects to APIs of both systems in parallel, compares records, identifies discrepancies, and groups them by type. But it goes beyond auditing: when a payment arrives in Zoho Books, I ask Claude to find the matching deal in Pipedrive and update its status. No Zapier needed, no custom webhooks. Just: "a payment came in from client X, update Pipedrive."

1 day → 10 min No Zapier Any API
Example 1: Audit > Compare invoices: Zoho vs BigQuery Checking 2,847 records... ✓ Matched: 2,831 (99.4%) ⚠ Mismatched: 11 (amounts > $1) ✗ Missing in BQ: 5 credit notes Root cause: webhook missed 5 records
Example 2: Payment → CRM > Payment from Wirelane in Zoho. Update Pipedrive. Zoho INV-2026-0892 paid ($4,500) Pipedrive Deal "Wirelane Q2" Stage: "Won" Note added with invoice ref
06

Infrastructure as Conversation

"Deploy Metabase on a VPS. I work on something else while Claude handles it."

When a client needs Metabase or another tool on a server, I don't spend time on SSH and manual configuration. Credentials are stored in a local .env file and never sent to the AI directly. Claude connects to the server, installs dependencies, deploys the service, and sets up SSL. I work on other things in the meantime.

45 min → 8 min Zero manual SSH Parallel work
> Deploy Metabase on staging VPS. Credentials from .env. Docker + nginx + SSL. [1/6] Connecting via SSH... [2/6] Installing Docker... [3/6] Pulling Metabase image... [4/6] Configuring nginx proxy... [5/6] Setting up SSL cert... [6/6] Health check: mb.client.com Done in 8 minutes. You were working on something else.
🔒 Credentials stay in local .env. Claude uses variable names ($VPS_IP), never the actual values. Nothing sensitive leaves your machine.

An AI-native
fractional CDO

I'm Nick Valiotti, Fractional CDO with 16+ years in data. This system runs my entire consulting practice, from morning planning to client delivery. Every workflow on this page is something I use daily across multiple client engagements.

LinkedIn Profile →
16+
Years in data
10x
Faster delivery with AI workflows
50+
Data engagements

Frequently asked

Is this secure? Does my data leave my machine?
All workflows run locally on your machine through Claude Code. API credentials are stored in a local .env file, and Claude uses variable references rather than actual values. No data passes through any third-party server beyond Claude's own API, the same as any Claude Code session.
How much does this cost to run?
Claude Code requires a Claude Pro or Max subscription, roughly $20 to $100 per month depending on usage. The productivity gains from even one automated workflow typically cover the subscription within the first week.
Do I need to be a developer to set this up?
You don't need to write code, but you need technical literacy: comfort with the terminal, understanding of APIs, and familiarity with your data stack. If you manage a data team, you likely have enough background to run this system yourself.
Can this work for my company's stack?
The system is tool-agnostic. It works with Jira, Notion, Slack, Gmail, BigQuery, Postgres, Metabase, Looker, Pipedrive, Zoho, and dozens of other services. If there's an API, Claude can connect to it.
How long does it take to set up?
A basic setup with morning brief and meeting autopilot takes a few hours. Adding more workflows (dashboards, reconciliation, deployment) typically takes a week of iteration as you refine the context files for your specific tools and conventions.
What if I want help building this for my team?
That's exactly the kind of thing we do at Valiotti Data. Fractional CDO engagements often include setting up AI workflows like these for clients' teams, so the system scales beyond a single person. Reach out via the contact form below.

Get the six prompts
in one Notion page

Copy, paste, adapt. The exact prompts I run every week for Context Library, Morning Intelligence, Meeting Autopilot, Dashboards, Reconciliation, and Infrastructure deploys. Link opens the moment you sign up.

Free · No spam · Unsubscribe anytime

Want this system working
for your data team?

20-minute call. We'll look at your current workflows and figure out where AI can help most.

Book a Discovery Call →