- Where it gets stuck
- The Looker Studio dashboard, fed by Supermetrics or another connector, is already up to date. What still takes hours for each retainer is the part the client actually reads: why the cost per lead jumped, what was tested and what you plan for next month.
- What the workflow does
- A commentary draft is prepared from the data behind your Looker Studio report (Supermetrics, GA4 or a CSV export), last month’s report, the change log your team keeps or the Google Ads change history or Meta Ads activity history, and the tasks in your project tool: what moved, which changes (new creative, budget shift, paused campaign) line up with it, and what the account lead still has to explain. When the figures do not explain a change, the draft leaves it as a question instead of guessing, and the account lead writes the plan and approves the report before it is sent.
AI for marketing agencies
The client’s feedback came in by email, two Slack threads and a Figma comment, and the delivery date has not moved.
Every written comment, plus your call notes if you add them, lands in one list against the brief, with likely scope creep flagged as a change request for the account lead to confirm. First test: a closed project’s feedback thread, with no access to your inboxes or ad accounts.
AI for marketing agencies drafts the work around the creative (report commentary, feedback rounds, pitches), and your account lead signs off each piece before a client sees it, so the same account team can run more retainers.
Today
- Monthly report for a retainer clientDashboard up to date, commentary and next month’s plan still written by handHours per retainer, every month
- Feedback on the campaignEmail, two Slack threads and a Figma comment, some contradictingWrong version reworked
- Pitch next TuesdayBest case studies sit in the founder’s old decksRebuilt the night before
With the workflow
- Commentary draftCost per lead up 38% from 12 March, in line with the budget shift to prospecting logged on 11 March. Open: did lead quality drop with the new form?Account lead adds the plan
- Change listEvery comment in one list, contradictions flagged, out-of-scope requests markedAccount lead confirms with client
- Pitch outlineApproved case studies and results, each linked to its sourceFounder decides
Three moments when a marketing agency still holds the account together by hand
All three repeat on every retainer, every month, and you or an account lead signs off.
The report commentary eats the first days of the month
Client feedback arrives through four channels
- Where it gets stuck
- Comments come by email, on Slack, in Figma and during calls, sometimes contradicting each other. The team reworks the wrong version, and nobody checks whether the request is inside the retainer, which is where the margin goes.
- What the workflow does
- Feedback from the channels you connect, plus your call notes if they are written, is gathered into one change list against the brief and the retainer’s scope of work, with contradictions flagged and requests that look out of scope listed as possible change requests for the account lead to confirm. The account lead settles the open points with the client, and the team works from one list.
Each pitch starts from a blank deck
- Where it gets stuck
- The right case study, the result the client approved for publication and the way the founder frames the problem all sit in old decks and inboxes. The founder ends up rebuilding the pitch the night before.
- What the workflow does
- A pitch outline is prepared from the brief and the agency’s approved material: comparable work, open questions about the brief and only the results cleared for publication. The founder sets the strategy, the fee and the recommendation.
Which flow should you start with?
Three questions, about a minute. The answer points to one of the moments above, or to a smaller step when a workflow would be too much.
Answer the three questions to see where to start.
This is an exercise: nothing is sent when you use this selector.
Play a quarter in a firm like yours
The Profit Game puts you at the head of a fictional marketing agency: pricing, hiring and AI decisions, with 100 simulated futures before each one. Free, about 6 minutes.
Play The Profit GameDo we have to change our project tool?
No. The workflow reads the tools your agency already uses, prepares a commentary, a list or an outline and puts it back where your team works, once an account lead has approved it.
- A month closes or feedback arrivesEnd of month, an email, a Slack thread or a Figma comment.
- The account’s figures and files are readThe data behind your reports (Supermetrics, GA4 or CSV exports), the brief and the project, within the access you allow.
- A draft is preparedCommentary, change list or pitch outline, each point linked to its source.
- An account lead decidesThe client sees only what the account lead has signed off, and corrections are kept for the next draft.
- Back into your toolsThe report goes to Google Slides and the tasks to Asana or Monday.
- Gmail
- Slack
- Asana
- Figma
- Looker Studio
- GA4
- Google Ads
- Meta Ads
Three conditions before anything acts
Your existing tools
The workflow runs inside the tools you already pay for, with the access rights they already have, so the team keeps its usual screens.
A named person approves
Every draft waits for a named person. Nothing goes out in the agency’s name without that approval, and exceptions go back to a human.
Every step leaves a trace
Sources, drafts, corrections and approvals are recorded, so you can see why an output was prepared and improve the next one.
Looking for a specific product?
PitstopThe shared place for briefs, approved case studies and publishable results, so the team finds them without asking the founder.Discover Pitstop →Managed AI agent serviceWe build, run and maintain one defined workflow, such as the monthly report commentary, with named approvers, a written scope and handover notes.See the managed service →
No agency client yet: our only published case is an executive search firm
AI outreach case study: Asia-Connect. Documented outreach at Asia-Connect over three months. One client and one workflow; results depend on context and execution.
“In the last three months, the agent helped me to book more than 90 qualified meetings with executives I would never have had time to reach, and I approved every single message before it went out. It works while I am in meetings.”
There, the partner approves every message before it is sent. With an agency we would start with the test described at the bottom of this page, on a closed project or on one client’s report, with no access to any ad account. Read the case study
Questions agency founders ask before a call
Which agencies is this for?
Owner-led marketing, digital, creative and communications agencies on retainers, where the founder and a few account leads still write the report commentary, chase the feedback and build the pitches. The workflows cover that work around the account. Strategy, creative direction and publication approval stay with your team and your client.
Our team already uses ChatGPT. What is different?
Your team uses ChatGPT to write. A workflow starts from an event, such as the end of the month or a new comment, reads the figures and documents you allowed, prepares the draft in the agency’s format and waits for an account lead’s approval, with a trace of every correction. Nobody pastes client figures into a chat window, and the draft arrives already in your report template.
Do you need access to our clients’ ad accounts?
Not for the test, which runs on an export. For a live workflow, you choose between read-only access through the connections the platforms already offer and scheduled exports, and the choice is written down per client before anything is built.
Where does our clients’ data go?
Before anything is built, a data processing agreement is signed and you receive the written list of every system that will read client accounts: where it is hosted, which model provider processes the text, and whether that provider keeps it or may train on it. A client’s data is only used for that client, and the only examples reused across accounts are the results a client approved for publication. Each deployment is isolated from every other AI Jungle customer.
What should we start with?
Usually the change list from a closed project: one thread, no client figures, and an account lead can judge it in minutes. Agencies whose month-start is eaten by reporting start with the commentary. Use the selector on this page, then bring one example to a 30-minute call.
Will it replace an account manager?
No. The aim is to run more retainers with the account team you have, and to give account managers back the hours they spend writing the same commentary and reconciling comments.
How much does it cost?
Prices are published on the AI Jungle pricing page and depend on the scope. A paid scoping study of your work and tools comes first (our Leverage Assessment), its fee is credited toward the build if you go ahead, and it can conclude that a smaller change is enough.
In 30 minutes, we pick the project or the report to test together.
You then send either the feedback thread of a closed project, or one client’s last two dashboard exports with last month’s report and, if you have it, the change-history export, under a confidentiality agreement, with the client’s name and brand removed if you prefer and no access to any ad account. We return the change list or the commentary draft. By the end of the call you know what to send and when you get the result back, and the test terms are confirmed in writing before you send anything.