At 7:42 a.m., Priya sits at the agency’s kitchen table with forty clients to report on and no dashboards open. The second monitor that used to hold them is no longer part of her Monday.
This is a typical composite, not a reported case study: seven people, forty retainers, an account manager who owns reporting, and a laptop covered in stickers. The timings, account movements and client replies are illustrative, not measured results from one agency. I’m following one Monday to show what changes when the reporting ritual becomes a weekly prompt in Claude, reading Google, Bing and Meta data through groas.
Last year, Priya’s Mondays started with tabs. Forty clients meant forty Looker Studio dashboards, each with its own date range to check, stale connection to investigate or Facebook-versus-Google conversion mismatch to explain. She opened the first three intending to read everything properly. By client nine, she was copying last week’s commentary and changing the numbers.
This Monday, she opens one window.
Before Monday: retiring the dashboard graveyard
The old second monitor held forty tiles named some variation of ClientName_Master_v3_FINAL. Each dashboard had a spend line, a conversions table, a blended CPA widget and a logo in the corner. Someone had spent time lining up those logos. Nobody could remember who.
Looker Studio is free and a default choice for Google Ads and GA4 teams. I understand the appeal. The first dashboard feels like infrastructure. Duplicate it, swap the account, change the logo. A reporting system, apparently.
Then the exceptions accumulate. One client needs a different conversion definition. Another wants the dates changed. Meta and Google count conversions differently, and displaying their numbers beside each other does not settle the disagreement. Priya ends up maintaining the report and explaining its assumptions before she gets to the account itself.
The dashboard is not broken. The Monday is.
One estimate puts report assembly at 2 to 5 hours per client per month. Another illustrates the smaller, familiar version: 30 minutes of manual pulling per client becomes 10 hours across a 20-client agency. Neither figure needs to describe every shop to make the problem recognizable. Repeating a small task forty times gives it a department.
What bothers me is what gets squeezed out. After the date pickers, exports and formatting, the commentary becomes: clicks down, CPC up slightly, monitoring closely. What the deck calls an insight, I call reading the chart aloud. A useful critique of PPC reporting tools makes the same distinction: clients need to know what wasted money, what is carrying the account and what gets tested next.
Priya’s change starts on Friday. She connects the client ad accounts to groas, then adds groas to Claude over MCP at https://mcp.groas.ai/mcp. The initial setup still exists. Each client’s accounts still need connecting. But once that is done, Claude can read the connected accounts through groas rather than wait for her to assemble another export.
She keeps the accounts and drops the forty separate reporting surfaces.
groas Unlimited costs $19 a month for the whole team, with unlimited teammates, ad accounts and websites. Another retainer does not require another paid dashboard slot. Against the Looker workflow, the attraction is not replacing free software with cheaper software. Free is difficult to undercut. It is removing the separate reporting chore between checking an account and doing something about it.
She does not migrate every chart. That would be rebuilding the graveyard with a different entrance.
7:58 a.m.: one prompt, forty rows
At 7:58, Priya types the paragraph she now uses each Monday:
For each client, summarize spend, conversions and CPA versus last week, flag anything that moved more than 15%, and write a three-sentence note in plain English a busy owner can read.
No export first. No tour of forty date pickers. Claude reads the Google, Bing and Meta numbers through groas and produces a table across the client accounts. The broad pattern is familiar from agency MCP reporting workflows: pull the account figures into one conversation, identify the movers and draft the client communication there.
The prompt is reusable. The output still needs reading.
In this Monday’s illustrative run, the table arrives in about four minutes. Thirty-four clients sit in a tight band: spend within a few percent of last week, CPA broadly flat. Six rows are flagged. A plumber in Leeds has spent 22% more with fewer conversions. A SaaS trial account has a 31% higher CPA.
Those are signals, not explanations. The first pass tells Priya where to look. It does not tell her that a competitor caused the problem, that a keyword deserves cutting or that someone should rewrite a landing page before breakfast.
She reads the six flagged notes first. Anything that describes an action not yet taken stays a proposal. The quieter accounts get a shorter check against the figures, then the same compact format: spend, conversions, CPA and a plain-English account of the week. No attempt to manufacture drama because a report has an empty box.
The first prompt sorts attention; it does not finish the diagnosis.
Eleven minutes into the reporting pass, Priya has a queue instead of forty open tabs. The routine notes are ready. The flagged ones wait for an explanation worth sending.
I like that distinction more than the speed. “CPA increased 31%” is a useful alarm. It is not yet a useful client email.
8:20 a.m.: the report becomes account work
By 8:20, Priya has narrowed the six flags to the three accounts needing attention first. She types a second prompt:
For the three worst movers, show me the search terms and ads behind the change and propose fixes.
This is where the old workflow sprawled. Open the account. Pull the search term report. Check the ads. Go back to the dashboard to remember the original question. Start a note somewhere else. Repeat until lunch acquires an optimistic definition.
Now the investigation stays in the same conversation. Claude pulls the account detail through groas. For the plumber, two broad match terms are picking up boiler-cover queries. For the SaaS account, one phrase match keyword is consuming 40% of spend on a comparison query that has produced no trials.
That gives Priya something concrete to inspect. It does not make every suggested fix correct.
She reads the proposals and edits one. For the plumber, she approves adding the boiler-cover terms as negatives. For the SaaS account, the proposal is to separate the comparison query into its own ad group with a lower bid and use a landing page that addresses the comparison directly. She checks that proposal against what she knows about the business before authorizing the account changes.
Then she tells Claude to apply the approved changes through groas. Each change is logged with its reason, and any change can be undone in one step.
The useful handoff is from evidence to an approved, logged action.
That log matters to me more than shaving another minute off the prompt. A dashboard records what happened to the numbers. A change log with reasons records why someone touched the account. Two weeks later, when the client asks why a keyword moved, the next person should not have to reconstruct Priya’s Monday from Slack fragments.
The three diagnoses and fixes take twenty-six minutes in this composite. That is a scene clock, not a promised benchmark. The meaningful difference is the work she does not repeat between tools: read the account, investigate the movement, approve the change, preserve the reason.
The multi-account workflow earns its keep there. Cross-account monitoring sounds tidy in a document. At the kitchen table, its value is more ordinary: Priya does not lose the thread while switching from reporting to execution.
She returns to the client notes. Now she can replace “investigating” with what she actually found and changed.
What reaches the client’s phone
The first visible difference is an absence. No PDF. No link to a dashboard that needs a password reset. No screenshot decorated with red and green arrows.
Just a short note with spend, conversions and CPA against last week, followed by the explanation the numbers have earned. For accounts that need action, it says what wasted money, what remains useful and what happens next. For accounts that do not, it says so without dressing a quiet week as a strategic breakthrough.
In the composite, the plumber replies nine minutes later: understood, keep me posted on the boiler terms. The SaaS founder asks one question about the comparison page. Neither asks where the dashboard has gone.
I would not turn those imagined replies into a claim about response rates. They illustrate the narrower change: the client no longer has to open a second thing to find the point of the first thing.
But Priya is not finished with either note.
Claude has identified a keyword problem in the SaaS account. It does not know that the founder changed the pricing page on Friday and hid the monthly plan below a demo form. That fact is not in the ad-platform figures. Priya has it because she asks every client channel the same question on Friday afternoon: anything changing on your side next week?
Launches. Stock-outs. Sales-call coverage. Price tests. The unglamorous context that makes an apparently tidy explanation incomplete.
The plumber’s note needs a fourth sentence too: call handling was short-staffed on Thursday, so missed calls explain part of the conversion dip alongside the irrelevant boiler queries. That line takes longer to write than copying the CPA. It also gives the owner something the CPA cannot.
Priya keeps the line only someone who knows the business can add.
Her approval happens before the fixes, not after them. Her context check happens before the email leaves, not when a client replies with the missing fact. The point of saving assembly time is to leave room for those checks, not to send an unfinished explanation faster.
That is how she uses groas as an execution layer across the client accounts. Claude reads and proposes. groas supplies the connected data and carries out approved changes. Priya decides whether the explanation fits the business.
She sends the notes only when those pieces agree.
9:14 a.m.: two charts survive
Two clients still get something different. Priya mentions them before anyone starts calling this the end of dashboards.
One ecommerce owner wants margin by product category in a custom chart she still builds once a month. One B2B founder wants a 90-day pipeline view. Neither question is answered by a week-over-week CPA note, however neatly Claude writes it.
She keeps those reports. They have a reader and a purpose.
What disappears is the assumption that every retainer needs the same permanent collection of charts. The weekly digest catches movement and starts the investigation. It does not settle an attribution argument or replace a board deck. Trying to make it do everything would leave Priya maintaining another elaborate reporting system, only this time with a prompt at the front.
The template taped to her laptop lid is short enough to read without opening anything:
- Monday, 8 a.m.: summarize every client’s spend, conversions and CPA against last week; flag movements above 15%; draft three plain-English sentences.
- For flagged accounts: inspect the search terms and ads behind the movement; propose a fix; review before approving.
- Before sending: add the business context, check the approved changes and read the log.
- Friday, 4 p.m.: ask every client channel what changes next week.
Forty clients, one reporting owner, roughly ninety minutes in this constructed morning, including the human lines. Not ninety minutes to manage forty accounts for a week. The custom reports remain. So do the client conversations, the awkward decisions and the work the numbers cannot explain by themselves.
What she has removed is the ritual of rebuilding the same view forty times before any of that work can start.
At 9:14, the kitchen table is just a table again. Laptop shut. Stickers facing up. Forty inboxes hold short notes, three accounts have approved fixes in the log, and two custom reports are still owed to the two clients who use them.
Priya pours a second coffee. The second monitor stays dark.

