Boutique creative and social-media agency in Colombia, serving about 20 client brands.
Inside Studio: from 0 brand values written down to 46 in one day
A creative agency in Colombia that had already abandoned three tools turned the files it had into 46 written brand values in one day (candidate values sent to the client to confirm, not a finished brand system) and had its first AI-assisted content plan approved within a month, early evidence from month one rather than a measured return.
Dates
Discovery 24 August 2026; first approved plan 27 September 2026
Team
Leadership committee of four; non-technical team
Responsibility
Delivered by Nirbound under its 90-day embedded-operator engagement, from September 2026. Attacca provided the method and the brain model.
Reviewer
Jhonn Moreno, operator on this engagement
Agency name and numbers approved for publication by Inside Studio’s operations lead on 27 September 2026.
What was true before
Tool history: Excel, then Jira (abandoned), then Slack (cancelled on 14 August 2026 for under-use and cost), then a ClickUp trial, then back toward Excel. A partner’s rule for the next attempt: “one chance”.
The traffic lead ran 20 brands from memory, an iPad and a spreadsheet; a brand expecting 20 pieces got 3; an August backlog was still open in September.
The one person reconciling hours went on leave, and the account team lost its numbers with her.
Drive: 6,213 files for one client group, one brand name spelled two ways across all of them, a 36-page brand manual made entirely of images, reports named for the wrong month, the same logo duplicated dozens of times.
Brand data in the brain on day one: zero hex codes, zero typefaces, zero logo paths for ten sub-brands.
Fear, in the production lead’s words: “If I step out of the operation for a month to feed the AI, fine. But if I ask every lead to do it, we get what happened with Slack: nobody used it.”
“I am in trouble because I have no data.”
An account manager, to the operations lead
What the work had to respect
No month of downtime
The operation kept running while the brain was built.
No new tool to learn
The team is creative, not technical. Adoption had to survive without anyone learning a new tool.
More human judgment, not less
A client had just said the agency was “abusing AI” and wanted more human touch. The brain had to make human judgment more visible.
What was built
Discovery and process maps for traffic, creative and management, with quotes and decisions from each session.
An asset registry from Drive with explicit triage rules: dedupe by name and size, drop system junk, export native Google files, then fill the shelves in priority order. Result: “better 600 MB useful than 800 MB of copies” from the same budget.
Brand profiles: one day’s document sweep turned the empty brain into 46 concrete candidate values across five brands, two sourced from real manuals with Pantone, CMYK, RGB and HTML codes, converted into a precise client ask (“confirm these 46 values”) instead of “send us the manual”.
A standing rule in the agency’s own instruction file: instructions before references before examples, always labelled, after a reference image in the centre of a slide was read as a locked creative decision and caused a real production mess.
A content planning pipeline with the committee’s own quality bar built in, after the first AI-assisted grid was rejected live for being generic.
The Drive triage ladder. Files fill the shelves from the top down; duplicates, system junk and wrong-month copies never reach a shelf.
How it was tested, and what changed
The committee rejected the first AI-assisted content plan in the room on 22 September 2026 (“it did not mention a single product”). Five days later the same pipeline produced a November plan for one brand with seven concrete ideas, three carousel formats and a recovered results report the team thought did not exist, and it was approved to move to design.
Brand data: 0 to 46 written values in one day, from files the agency already had. They are candidate values sent to the client to confirm, not a finished brand system.
Asset registry: a usable set from the same storage budget instead of alphabetical duplicates.
First approved AI-assisted deliverable within the first month; humans kept the final call, which is what the client had asked for.
Evidence, with denominators and caveats
Measure
Before
After
How measured
Caveat
Brand values written in the brain
0 hex codes, 0 typefaces, 0 logo paths for ten sub-brands (day one)
46 concrete candidate values across five brands (day two)
Values counted in the brand profiles after one day’s document sweep
Candidates sent to the client to confirm. Two brands sourced from real manuals.
AI-assisted content plan
Rejected live by the committee on 22 September 2026
November plan for one brand approved to move to design on 27 September 2026
Committee decision in session
One brand, one plan. Approval to design, not a published result.
Asset registry
6,213 files for one client group, with duplicates and misnamed reports
A usable set from the same storage budget, filled in triage order
Triage rules applied to the Drive inventory
Describes audit scope and order. No time or cost saving measured.
Limitations
Not measured yet: hours per deliverable, margin per account, client retention.
The engagement is in month one.
The 46 values are candidates the client was asked to confirm, not a finished brand system.
Discuss a comparable workflow
If a workflow in your business keeps losing things between people and tools, start with the Business Brain setup and one workflow you can check.