TLDR

Cowork's Live Artifacts let marketing and sales teams build self-refreshing dashboards by chatting with Claude. No code. No data team.
The skill moved to brief-clarity, not engineering.
If you can write a brief for a creative team, you can write a prompt for a dashboard.
Get my marketing campaign dashboard prompt.
Vibe subscribers get The Cowork Starter Kit: 10 ready-to-paste prompts, the dashboard-building skill, the connector setup walkthrough.

Last month, I was prepping for a Monday training session with a marketing team at a luxury brand.

Their question for me: "Toni, how can we build dashboards without bothering the data team every time? Can you build them for us?"

I opened Cowork. I typed eight sentences.

12 minutes later, they had a working campaign performance dashboard that pulls from their HubSpot and refreshes itself every time they open it.

I haven't written a line of code in years.

Since, I've spent the past weeks building Live Artifacts with marketing and sales teams for our AI adoption programmes.

These is my own take, beyond all the promises you read on social.

What Live Artifacts are

Quick context for anyone who hasn't opened Claude Cowork yet. Cowork is a feature for the desktop version of Claude. Mac or Windows. Paid plans only (Pro, Max, Team, Enterprise). It’s Claude Code for non tech teams. It can work directly on your files, take action, run agents, without any coding knowledge.

Live Artifacts are persistent HTML dashboards living in a dedicated tab inside Cowork.

Artifacts have been around for a long time in Claude. ai. They are basically HTML assets you can create with a prompt (webpage, dashboards, mini games, tools…). But they are static. You build them once and that’s it. If you want to update the data, you need to rebuild them.

Live Artifacts solve that limitation.

You build them by chatting with Claude. You describe the layout, the data sources, the design. Claude generates the page, wires it to your connectors (HubSpot, Salesforce, Google Sheets, Notion, your local files), and saves it. Every time you reopen the artifact, it re-queries the data and refreshes.

A few things that matter:

You don't write code. You describe what you want in plain English.
The dashboard lives forever, not just inside the chat that made it.
Version history is built in. Broke it? Roll back.
You iterate by chatting. "Add a chart for paid spend." "Change the colors." "Filter by region."

A few things that don't work yet:

you can't share artifacts with teammates (sharing is on the roadmap).
And they live in Claude Desktop, not the browser.

"But I'm not technical"

I know what you're thinking. "The data team has always built these for me. I wouldn't know what to ask for. "

I'm not a developer either. I'm not great with Excel. I am no Power BI wizard.

The skill here is being clear about what decision the dashboard supports, who needs to see what, and where the data lives. That's a marketing skill. Or a sales skill. Not an engineering skill.

If you can write a brief for a creative team, you can write a prompt for Claude Cowork.

The first live dashboard I'd build

Here's the actual prompt I used for the marketing team last month. Paste it into Cowork, swap the specifics for your data, and you have a working campaign performance dashboard in minutes.

Build me a Live Artifact: a weekly marketing campaign performance dashboard. Data sources:
- HubSpot deals export
- Google Analytics weekly export (CSV in the cowork folder)
- Q4 targets spreadsheet (CSV in the Cowork folder) Layout:
- Top row: 4 KPI tiles · Total spend this week · MQLs this week · SQLs this week · Pipeline created this week Each tile shows the number + % change vs last week.
- Middle: trend line chart of MQLs by week, last 8 weeks, broken out by acquisition channel (paid social, paid search, organic, email).
- Bottom: table of campaigns ranked by pipeline created. Columns: campaign name, channel, spend, MQLs, SQLs, pipeline created, cost per pipeline. Sortable. Design:
- Filter at top: This week / Last 4 weeks / This quarter
- Brand color: [your primary]
- Sans-serif fonts, clean spacing
- KPI changes: green up arrow if positive, red down arrow if negative Refresh data every time I open it.

That's the whole thing. Eight bullets. Plain English. No code.

Claude will probably get 70-80% of it right on the first run. First time I tried this, the KPI tiles were stacked vertically instead of in a row, and the trend chart was using the wrong column.

Two follow-up messages fixed both.

"Put the KPI tiles in one horizontal row at the top." "For the trend chart, use the 'channel_name' column from the GA export, not 'source.'"

Three rounds of iteration. Twelve minutes total.

🔓 Get the full Cowork Starter Kit

The Dashboard-crafting skill + all 10 dashboard prompts (5 marketing + 5 sales) + connector setup walkthrough.

Join Vibe subscribers and get full access.

Subscribe now

Honest verdict

A few things didn't work the first week. Worth knowing before you start.

First dashboard I built had no clear owner. After two weeks, the data feed broke and I didn't notice for four days. Connectors do break. Schemas change upstream. Someone has to validate weekly that the numbers still match the source.

Do not “over-rely” on Claude’s dashboard. Verify the output.

Second build was cluttered. I asked for everything I might want to see. The team opened it once, it was too much.

Now I start every dashboard with one question: "What's the one decision this supports?" If I can't answer in one sentence, I don't build it.

The shift, in three lines

The dashboard divide is over.
Live Artifacts didn't make dashboards better. They just moved them out of the data team's queue.
Build the one your weekly meeting actually needs.

We've trained over 1,000 professionals on AI this past year. The single biggest pattern is non-tech teams thinking:

"I can’t do this, I’m not a [specialized technical job], I need help" when the help is now a 4-bullet prompt away.

AI adoption is learning when this sentence is true, or not.

🔓 Get the full Cowork Starter Kit

The Dashboard-crafting skill + all 10 dashboard prompts (5 marketing + 5 sales) + connector setup walkthrough

Join Vibe subscribers and get full access.

Subscribe now

What's the one report you keep getting asked to build manually that you'd love to never build again?

Reply and tell me. I keep a running list, and the patterns are starting to repeat.

If this is useful, the best thing you can do is send it to whoever on your team just wished out loud they could build a dashboard without bothering the data team. They need this.

The Cowork Starter Kit

For Vibe subscribers, The Cowork Starter Kit gives you:

10 ready-to-paste prompts. 5 marketing dashboards (campaign tracker, content production, competitive monitoring, lead source attribution, SEO opportunity radar). 5 sales dashboards (pipeline cockpit, rep personal board, territory planner, executive revenue brief, deal risk monitor).
The connector setup walkthrough. How to wire HubSpot, Salesforce, Google Sheets, Notion, and local files without breaking things.
The Dashboard Creator Skill. Build perfect dashboard in seconds.

The 10 Dashboard Prompts

Each dashboard prompt is paste-ready. They’re based on my own dashboards.

Swap with your actual data sources, your brand color, and your specifics. Iterate in Cowork. You're done.

Each prompt has a Decision, a Meeting, and a Tip. If you can't fill in the Decision and the Meeting for your version, skip the build.

Marketing dashboards

1. Weekly Campaign Performance Tracker

This is the one I just shared above

✅ Decision: Which campaigns to cut, double down, or test next week. 📆 Meeting: Monday marketing standup.

💡 Tip: The narrative panel makes the difference, team reads three sentences, starts the meeting on substance, not on random, abstract charts.

2. Content Production Tracker

✅ Decision: What's at risk, where the bottleneck is. 📆 Meeting: Friday editorial sync.

Build me a Live Artifact: a content production tracker. Data sources:
- Notion content database (Notion MCP). Fields: title, owner, status, draft due date, publish date, channel
- Google Analytics export for published content (last 30 days) Layout:
- Top: 3 KPI tiles · Pieces shipping this week · Pieces overdue · Pieces published last week + total pageviews
- Middle: kanban columns: Brief → Drafting → Review → Scheduled → Live. Each card: title, owner, due date, channel.
- Bottom: top 10 published pieces from last 30 days, ranked by pageviews. Columns: title, channel, pageviews, top traffic source. Design:
- Cards turn red when overdue, amber when due in 2 days
- Brand color: [your primary] Refresh on open.

💡 Tip: Don't try to replace Notion with this.

Use it as a read view. The team still works in Notion. The artifact turns the meeting from "where are we" into a 2-minute answer.

3. Competitive Monitoring Board

Want to go deeper? At Spicy Advisory, we help startups and scale-ups integrate AI into their product and GTM processes. Explore our AI adoption programs for hands-on workshops and deployment support.

How I use Claude to build Live Dashboard. Without any code.