Marketing agency in Florida, United States. SEO, paid media and web for home-services, legal and retail clients.
Suncoast Interactive: from 3 of 13 follow-ups captured to 8 of 8
A Florida marketing agency moved its clients, meetings and assets into one shared brain, and the first structured run of its meeting-to-task flow captured 8 of 8 action items. Different meetings, different denominators: a first-run signal, not a controlled comparison.
Dates
Mid-2026 to September 2026
Team
Roughly eight people, owner-led
Responsibility
Delivered from inside the agency by its SEO, paid and AI operator. Attacca provided the method and the shared-vault operating model, not a separate consulting contract.
Reviewer
Jhonn Moreno, operator on this engagement
Client name and numbers approved for publication by the agency owner on 27 September 2026.
What was true before
Client review meetings produced next steps that mostly never became tasks. One review yielded 13 concrete actions across six owners; 3 reached the task board, and none of the four highest-stakes items assigned to the agency’s own principal. The client chased the agency for its own commitments.
One account manager carried 167 open task rows, 40 overdue, before counting Slack and call commitments with no task at all.
A pending-work audit surfaced commitments from October 2024 still unresolved in 2026, including a subscription cancellation that itself became a dropped task.
A departing project manager held the account picture in her head; her exit forced a manual reassignment sweep.
Budget was flowing to things nobody had verified: one search campaign served 3 impressions in 31 days on a $33 daily budget; another spent $809.72 for zero leads at a cost per click roughly eight times the account’s best campaign.
Asset hygiene: a dozen duplicate exports of one client logo; 419 local exports and QA screenshots discarded at a single session close.
“They came from 32 people, 32 designers, went all the way to 18 because they lost a big client.”
The agency owner, on a peer agency
What the work had to respect
Readable by the agency’s own people
The vault is shared with the client team, so every note had to make sense to the agency’s staff, not just to an operator.
The tools stayed
ClickUp, Slack, Google Drive and Apollo. Nothing was replaced. The brain sits beside them and holds what they do not.
No new headcount
The work had to fit the team the agency already had.
What was built
A structured agency brain: one file per client (baseline, canonical facts, channel strategy, house rules), one process map per recurring workflow, one task ledger with receipts.
A meeting-to-task flow: transcript, extracted commitments, dedupe against open tasks, draft, human approval, then a write to ClickUp, leaving a reviewer trail instead of silent gaps.
Canonical facts locked once corrected. A client’s active store count went from an assumed 37 to the verified 33 and was written so it never resurfaced.
House rules derived from real failures, for example a lead definition that excludes raw click and view events after a previous vendor double-reported conversions.
Per-item instructions such as “do not repeat access requests already made” attached to open work, replacing memory with a checkable record.
The meeting-to-task flow. Every step leaves a mark on the reviewer trail, and nothing reaches the task board until a person approves it.
How it was tested, and what changed
The acceptance test was the same one Business Brain sells: does an agent run a useful workflow on the agency’s real files? The first live run of the meeting-to-task flow on a real client call extracted 8 of 8 action items and correctly proposed one dedupe comment instead of a duplicate task.
Follow-through: 3 of 13 captured before; 8 of 8 on the first structured run. Different meetings, different denominators; this is a first-run signal, not a controlled comparison.
One canonical version per asset at every session close; duplicates are discarded with a hash check.
Wrong numbers stop resurfacing once locked as canonical.
Evidence, with denominators and caveats
Measure
Before
After
How measured
Caveat
Meeting commitments captured
3 of 13 actions reached the task board after one client review (six owners)
8 of 8 action items extracted on the first live run, plus one dedupe comment instead of a duplicate task
Action items agreed in the meeting, counted against what reached the board or the draft
Different meetings, different denominators. First-run signal, not a controlled comparison.
Canonical facts
A client’s active store count assumed at 37
Verified at 33 and locked in the client file
Correction written once to the client’s canonical facts
One documented correction, not a count across every fact.
Asset versions
A dozen duplicate exports of one client logo; 419 local exports and QA screenshots discarded at one session close
One canonical version per asset at every session close
Hash check on duplicates at session close
Describes file hygiene, not time saved.
Limitations
Not measured yet: hours saved per week, client retention effect, revenue.
The 3 of 13 and the 8 of 8 come from different meetings with different denominators.
The work was delivered from a seat inside the agency. Attacca’s part was the method and the shared-vault model, not a separate consulting contract.
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.