Meeting cockpit · Tuesday, August 18, 2026
Steve·Atómica

36 days running on their own. And one honest finding that changes where we point this.

Three client agents delivered 237 reports without a single failed run while you were away. This cockpit shows what ran, what your team told us, and the measured reason we are moving Steve from paid media into creative.

Program Steve · Atómica · multi client agents
Accounts live UCM · Werba · Cetaphil
Period 13/07 → 17/08/2026
In the room Verónica · Marielena · Analicia · Juan
36
days running, no gap
237
reports delivered to Slack
3
accounts with own agent
0
failed runs
2
client risks flagged early
460h
creative hours/month, untouched
1 · What ran
2 · Feedback
3 · The finding
4 · Where it goes
5 · Decisions 3
6 · Agenda
Measured, not claimed

What ran while you were out.

Every number here comes from your own server and your own Slack. Nothing is projected.

100%
Uptime
36 of 36 days, delivered at 08:30
3/3
Agents alive
one dedicated channel per account
30
Asana accounts
connected, strictly read only
2
Risks flagged
accounts going quiet, caught early
The engine is solid. Five weeks, zero downtime, and it is already surfacing things nobody had time to notice: two of your accounts have gone quiet with the client. Today is about pointing that same engine at the work that eats your team's week.
Live from your server, right now

The three agents, this minute.

reading…
Daily control tower Each agent's house The 360 report Full cockpit
Shipped

What exists today, and works.

The agentslive
  • ✓One agent per account. Each one knows whose account it is, no mixing.
  • ✓Daily report at 08:30 with real Meta and Google numbers from Dashbo.
  • ✓Dedicated server for Atómica, isolated, running 24/7.
  • ✓Atómica's own Claude account, used only here.
Safety and controllocked
  • ✓Read only on Asana and Gmail. Steve can never write, move or delete. Enforced in code.
  • ✓Approval by react. Nothing reaches a client without a ✅ from your team.
  • ✓Every output is a draft, labelled as such.
Intelligencelive
  • ✓Client cooling radar. Flags accounts going quiet. It already caught two.
  • ✓Asana across 30 accounts, so Steve sees what is due and what is late.
  • ✓It teaches itself. Drop a brand book in a channel and it reads it into that client's memory.
Visibilitylive
  • ✓Daily control tower with every account at a glance.
  • ✓A live house per agent, showing exactly what it is connected to.
  • ✓The 360 report, your book of 29 accounts.
The most useful message of the month

Your team told us where it hurts. And they were right.

On August 13, in the Cetaphil channel, Marielena wrote one line that is worth more than any report we delivered.

“I am not sure the whole paid section makes much sense, since we do not do any of it.”

Marielena · #steve-cetaphil · August 13, 2026
She spotted it before we did. We built the agents around paid media reporting, because that is where the data was easiest to reach. She pointed out that on her accounts, that is not the work. So we went and checked it against your own hours sheet.
What we did with it

We measured it instead of defending it.

Her observation

“We do not do paid here”

One line in a client channel, from the person closest to the account.

august 13
→
We checked

Against your hours sheet

All 29 accounts, function by function. Not our opinion, your own numbers.

august 17
→
The result

She was right, and it is bigger

It is not only Cetaphil. The next tab shows what the numbers say about the whole agency.

changes the plan
This is how the partnership should work. Your team sees something from the inside, we measure it from the outside, and the plan changes. That loop is worth more than any single deliverable.
Measured against your hours sheet

We aimed Steve at 23% of your agency.

Atómica runs roughly 987 hours a month across 29 accounts. This is where they actually go.

Creative, content and community460 h · 47%
Account management and direction296 h · 30%
Paid media231 h · 23%
where your hours are where we pointed Steve source: Atómica hours sheet, 29 accounts
On the exact three accounts we picked

Two of the three do not run paid at all.

Account
Creative + content
Paid media
UCM
50 h
15 h
Werba
33 h
0 h
Cetaphil
8 h
0 h
What we got rightkeep
  • The engineMulti tenant, one agent per account, zero failures in 36 days.
  • The safety modelRead only, human approval, drafts. Nothing to rebuild.
  • The connectionsDashbo, Asana across 30 accounts, Gmail, Drive, Slack.
  • The radarClient cooling flags. Already useful, already catching things.
What we got wrongfix now
  • The targetWe aimed at paid media, including on accounts that do not run paid at all.
  • The orderWe started where the data was easiest to reach, not where the work is heaviest.
  • The check-inWe should have asked your team what to point it at after week one, not week five.
  • The gap460 creative hours a month, and Steve was not touching any of them.
This is a targeting fix, not a rebuild. The expensive part, the engine, the connections, the safety, is done and paid for. What changes is what we point it at.
The move

Same engine. Pointed into the 460 hours.

Same channels, same approval by react, same read only guarantees. We change the work it does.

Step 1 · this week

Kill the noise

Paid report off on Werba and Cetaphil. Each channel keeps only what that account actually does. Marielena's note, closed.

2 hours of work
→
Step 2 · next 2 weeks

One real video

Steve edits a real client video. Your team approves it with a ✅. The first approval ever used, on work that matters.

the proof
→
Step 3 · 30 days

Hours back

We measure hours returned to the team on creative. That number, not our opinion, decides how far we scale.

the metric
What Steve does once it is pointed at creative

Six capabilities, three already running.

🎬 Video editing

Long form into reels and shorts, auto captions, hooks and cuts on brand. Your team goes from editing from zero to approving a cut.

the big one

🎨 Design and image

Carousels, statics and variations that respect each brand book, which Steve already reads. Santiago moves from producing to directing.

next

✍️ Content and copy

Captions, calendars and copy in each client's voice. First draft from Steve, the team edits instead of facing a blank page.

next

🚩 Client cooling radar

Already live and already useful. It flagged Werba quiet for 149 days and UCM for 59. That is a churn warning, not a report.

running today

📋 Asana, read only

30 accounts connected. Steve sees what is due, late or piling up, without touching a single task.

running today

📊 Paid, where it exists

We keep the daily report on UCM, which has 15 real hours of paid a month. We turn it off everywhere it does not apply.

fix this week
Three decisions. Each one unblocks something.

What we need to agree today.

01

The pivot

Do we point Steve at creative, starting with video?
The data says yes: 460 hours a month against 231 in paid. We need your call to start, because it changes what the team reviews every week.
blocks steps 1 and 2
02

Who owns it

Which person approves Steve's creative output?
Creative needs one named person per account holding the ✅, so drafts move the same day instead of waiting for consensus. Our suggestion: Marielena, since she is the one who spotted where to aim this.
blocks the first delivery
03

The model

Do we move to a tighter monthly, aimed at creative?
A narrower scope, focused where your hours actually are, at a lower monthly than the current one. We settle the open invoice and start the new cycle clean.
blocks september
The proposal

USD 2,000 per month, aimed at the 460 hours.

Down from 3,300, with a tighter and more honest scope: creative production first, paid only where it exists, and a monthly number we both watch, which is hours returned to your team. The open invoice gets settled and September starts clean.

What you get for it: the three agents running, the video and design layer, the client radar, the control tower, and me in Slack same day. What we drop: reporting on work you do not do.

How this meeting runs

30 minutes. Ending on a decision.

Short, honest and ending on a decision, not on a follow up email.

0 → 5min
What ran
36 days, 237 deliveries, zero failures. Open the live block so she sees the three agents answering in real time.
Juan
5 → 11min
What your team told us
Read Marielena's message out loud and give her the credit. Frame it as the partnership working: they saw it, we measured it.
Juan
11 → 17min
The numbers behind it
460 hours a month in creative against 231 in paid. Werba and Cetaphil at zero paid. This is the turning point of the conversation.
Juan → Marielena
17 → 25min
The pivot to creative
Video, design, content. The three steps with dates. Ask directly which client and which video we use as the first proof.
everyone
25 → 30min
The three decisions
Pivot, owner of the ✅, and the model. Leave with the three answered, or with a named date for each.
Verónica
The one rule for this meeting: we lead with her team's insight, not with our deliverables. A vendor reports activity. A partner brings a measured reason to change the plan, and a number to judge it by.
Atómica × Hive Technology · live data from your own server · read only on Asana and Gmail · August 2026