The morning tax
Every morning, someone senior assembles the business from a patchwork of tabs
Most consultancies run their business from a patchwork of tabs: HubSpot for pipeline, Notion for tasks, Gmail for flagged emails, Power BI for revenue. Every morning, someone senior spends the first hour of their day pulling those fragments together in their head and deciding what actually matters today.
That is not a knowledge problem. It is a tooling problem — and it just got a practical fix.
What changed
Live artifacts that stay connected to your data
Until recently, asking Claude to build a dashboard produced a one-shot output — a chart, a script, a table you copied somewhere else and never opened again. Useful, but disposable.
Anthropic’s co-work environment now ships what it calls live artifacts. These outputs stay connected to your data sources — HubSpot, Gmail, Notion, Outlook, BigQuery, and anything else that exposes an MCP connection. The artifact lives in its own tab, refreshes when you reopen it, and keeps a full version history so you can roll back if something changes.
The artifact you build today is the dashboard you open tomorrow, and the day after.
For a 30-person consultancy, that compounds quickly. Firms in that headcount band typically spend £15,000 to £30,000 a year on tools and lose five to ten hours a week of senior time to scattered operational data. Live artifacts directly attack both.
Dashboard one
Business Pulse: is my month on target?
The business pulse dashboard is a single page of live tiles — revenue month-to-date, leads, outreach activity, pipeline value, and cash position — answering “Is my month on target?” in under five seconds.
To build it, open a new chat in co-work, describe what you want as a live artifact, specify which tiles you need, where the data lives, and how often it should refresh. Within a few minutes Claude generates the artifact and requests permission to connect to your data sources. From there you can:
- Click into any tile for a detailed breakdown.
- Copy a contextual prompt straight into the chat to interrogate the numbers further.
- Export the current view as a PDF to attach to a client or board email.
Brand styling takes a second prompt — paste in your brand guide or point Claude at your skill file and it applies colours, logo, and typography to the next version. Every revision is saved, so switching between versions is straightforward.
This replaces the monthly slide deck someone on your team spends half a day assembling, and the Power BI report that only refreshes overnight. Revenue, pipeline, and cash are live, in one place, readable in seconds.
Dashboard two
Attention: what needs me today?
The attention dashboard answers a different question: “What needs me today?”
It has six sections pulling from across the stack — overdue tasks from HubSpot and Notion, meetings in the next seven days from HubSpot and Outlook, deals with high close probability this week, emails flagged for follow-up from Gmail and Outlook, and tasks due in the next seven days from HubSpot.
Every item carries two action buttons: Open jumps straight to the source record in whichever system owns it, and Discuss generates a contextual prompt with all the relevant data already loaded, so you can ask Claude a question about that specific deal, task, or email without pasting in any background. A summary row at the top gives a count per category at a glance, items are prioritised automatically below it, and a refresh button pulls the latest state from every connected system.
This replaces the morning routine of logging into five systems, tabbing between them, and spending mental energy deciding what matters before you have done any actual work. The context is ready. You act on it rather than assembling it.
Where it sits
The Workflow layer of the KWA stack
Both dashboards operate at the Workflow layer of the Knowledge – Workflow – Agents (KWA) framework. They assume your data sources are already connected and structured (the Knowledge layer). What they do is surface that knowledge in a prioritised, actionable format without requiring you to write a query, open a report, or brief a person.
The natural next step for firms with these dashboards running is the Agents layer: Claude not just showing you what needs attention, but drafting the follow-up email, logging the activity in HubSpot, or flagging the at-risk deal to the delivery lead. The dashboards create the connective tissue that makes that possible.
If you are working through the 5 Steps for AI Leadership, this sits squarely in the Activate and Amplify steps — moving beyond individual tool experiments toward something the whole leadership team uses every day. When that pattern lands for one person, it spreads naturally because the output is visible and the value is immediate.
Before you build
Four limitations worth knowing
Live artifacts are in preview, which means a few things you should plan around.
- Co-work only. The base Claude free plan does not include this yet.
- Connector coverage varies. Outlook needs the Microsoft 365 MCP wired in separately if you want to write back to it, not just read from it, and BigQuery needs some initial setup before it works as a data warehouse.
- Authentication is per user. Rolling this out across a ten-person leadership team means each person authenticating individually; a full firm deployment needs a more structured approach, likely with IT involvement.
- It is still glitchy. Minor rendering issues and occasional unexpected behaviour are common. Given the pace Anthropic is shipping at, most of this will resolve quickly, but reliability is not there yet — don’t build anything client-facing on it.
Two rules
Keep it internal, and be deliberate about the data
Rule one: keep it internal for now. These dashboards are for your team’s operational use. Do not put a live artifact in front of a client until the reliability is demonstrably consistent.
Rule two: only connect data you can afford to expose. The artifact pulls live data through Claude’s language model. If a data source contains information that cannot be shared with an LLM — certain client records, regulated data, commercially sensitive deals — do not connect it. Be deliberate about what goes in.
Where to start
Build the attention dashboard first
If you are building from scratch, start with the attention dashboard. HubSpot, Notion, Gmail, and Outlook are connectors most consultancies already have, the data is already scattered across them, and the operational payoff is visible from day one. Once that pattern is working, the business pulse follows naturally — you already have the connected stack, you are just surfacing a different view of it.
The shape of both tools is deliberate: strategic context (where the business is heading) plus tactical priority (what needs you today), with direct links back into Claude so you can act on items without switching context. That combination is what makes it a working tool rather than a report. Generic AI tools produce outputs you paste somewhere else and never look at again. Live artifacts are different because they stay connected, stay current, and stay where you work — the shift from one-shot output to operational infrastructure.