Case study · Creative studio, ~12 people · Embedded ops + AI enablement · Active engagement

Raw Creative: from tribal knowledge to systems, then AI on top

A studio running on talent and trust, but not on systems. This is what it took to change that, and where the numbers stand right now, each one labeled by source.

The before

Raw Creative (also operating as BLD) is a creative studio of roughly 12 people producing design, branding, and campaign work across multiple industries. When the engagement started in January 2026, almost nothing was written down. Processes weren't documented, tasks lived in DMs and individual memory, and financial operations were spread across Bill.com, Divvy, and Expensify with no central source of truth. Onboarding was improvised. Payroll took two hours per cycle because timecards were chased down at the last minute.

"98% our company operates in me, Jason, Breton, Sasha, and Lacey's head." Mike Blea, Raw Creative · engagement kickoff

The team knew they needed systems. They had tried before and never been able to make it stick.

Phase one: the operational backbone

You can't layer AI on top of chaos, so we didn't. The first phase installed the boring, load-bearing stuff:

  • Slack as the communication hub. Channels for ops discovery, invoices, onboarding, and project work. Email-to-Slack forwarding. Knowledge moved from DMs that died with the conversation into a persistent, searchable space.
  • Asana as the source of truth for tasks. Onboarding boards, project templates, migration tracking, and a Slack integration with notification settings tuned so people would keep using it.
  • Financial consolidation into QuickBooks. Migration off Bill.com, Divvy, and Expensify. Vendor and subscription mapping, centralized expense tracking, chart of accounts alignment.
  • 1Password shared vaults. No more siloed access to vendor portals, cards, and banking.
  • Meeting intelligence. Transcription and recall across meetings, so "wait, what did we decide?" has a searchable answer.
  • Documented SOPs and onboarding boards, plus a Claude masterclass for the team as the first step toward embedded AI.
"We have gotten more done on systems in the past three months than we have in the past 10 years." Mike Blea, Raw Creative

Phase two: the AI layer

After a pause, the engagement re-engaged to expand into full AI infrastructure across every department. That work is live right now:

  • Claude workspace architecture. Ten workspace buckets spanning project delivery, sales, marketing, operations, people, finance, and fabrication, with three interaction modes (chat, co-work, design) and a shared Google Drive as the portable knowledge layer. The studio's history became something its AI can draw on.
  • Tool consolidation. Individual paid ChatGPT accounts canceled and consolidated onto a single Claude org account, with a free-tier tool kept for one specific use case. Fewer subscriptions, one shared context instead of five private ones.
  • A custom estimating engine, in active build. A two-part tool built around how this studio already scopes work: a voice-driven interview intake connected to a calculator wired to Airtable, pre-populating estimates from the interview and pulling historical comparables from past projects. It has been tested live against a real project scope and it's still in tuning, which is exactly what I'd want a prospective client to know. Purpose-built AI takes iteration, and the team's trust in it is being earned session by session, not assumed.
  • Cadence redesign. Daily 15-minute standups, a Thursday dashboard rhythm, and a shift toward async updates generated from the Slack, email, and meeting record the backbone now captures.

The numbers, labeled

Embedded ops work gets felt long before it can be proved, so every number below carries a label showing exactly where it came from. That's the Receipts Framework. Two of these are placeholder estimates pending client confirmation, and they're labeled that way on purpose. One has no anchor yet, so it stays an open gap instead of a made-up number.

Verified · before Estimate · after

Payroll: 2 hrs → roughly 30–45 min per cycle

The two-hour baseline is verified, named by the client at kickoff. The after number is a Danee-observed estimate based on the timecard rebuild and weekly missed-punch monitoring now in place.

ESTIMATE: confirming exact number with the Raw Creative team this week.
Verified · direction Estimate · savings

Roughly $80–120/month in avoided redundant AI spend

The consolidation itself is verified: individual ChatGPT subscriptions canceled, one Claude org account in their place. The dollar figure is a Danee-observed estimate based on typical individual-plan pricing across an estimated 4–5 canceled seats.

ESTIMATE: confirming exact seat count and prior cost this week.
Not yet measured

Onboarding time

New SOPs and onboarding boards are live, but no before/after baseline exists yet, and I won't invent one. Currently gathering a real baseline with the team.

Open gap, shown as a gap until a real baseline exists.

The bigger shift

Communication moved from DMs to persistent channels. Tasks moved from heads to Asana. Money moved from four disconnected tools to QuickBooks. Meetings moved from "what did we decide?" to a searchable record. The team can now delegate without sitting next to someone while they do the work, which was previously impossible. And the AI layer is being built on top of a foundation that can hold it.

"It has been super helpful… happening in Slack… insanely helpful for the architecture side." Mike Blea, Raw Creative · on the Slack-centered architecture
See the offers behind this work How the numbers get labeled