Ask Sellforte MCP: What should our agency change this week?
Welcome to the Sellforte MCP Use Case Library. Each post takes one question that marketing teams really ask their AI assistant through Sellforte MCP, which is now in general availability and accessible to all marketers worldwide. It strips out anything customer-specific and walks through what the assistant does with it, step by step. We run every question on our demo account so the numbers you see are illustrative, not anyone's real results. The idea is simple: steal the prompts, try them on your own model, and see what you get back. We hope you find this series inspiring!
-Juha, CEO at Sellforte
Turning this week's recommendations into a brief your agency can act on
If an agency runs your Google or Meta campaigns, the weekly handover is where good analysis often gets lost. You've reviewed the numbers internally, you know what should change, and then someone has to retype it into a message the agency can follow: which campaign, what to change, by how much, and why. This use case is for the performance marketing manager who writes that message every week.
What you ask
Using Sellforte, list this week's recommended bid and budget changes for our Google Ads campaigns, with last week's spend and ROI for each. Turn the five biggest into a short action list for our agency, one line per campaign with the reason.
Show me the draft before you send it anywhere.
What Sellforte MCP does
The assistant pulls two things and puts them side by side.
First, the bid and budget recommendations for every Google Ads campaign. These are a snapshot over a fixed analysis window, on our demo account 2026-10-05 to 2026-11-01. For each campaign it gets the current and recommended bid value, the recommended spend change, the marginal return on the next dollar (miROAS) and the projected weekly effect on incremental sales.
The summary: 10 Google Ads campaigns have recommendations, 7 to increase and 3 to decrease. Daily spend would go from 18 766 to 20 169, with a net expected gain of 8 576 in incremental sales per day.
Second, last week's results from the model, 2026-10-25 to 2026-10-31, campaign by campaign. That's the context your agency needs to see the change isn't coming out of nowhere.
| Campaign | Recommended change | Bid setting | miROAS | Weekly incremental sales impact (USD) | Last week spend (USD) | Last week ROI |
|---|---|---|---|---|---|---|
| YouTube DE | +38.9% budget | Target ROAS 0.8 → 0.64 | 4.49 | +21 519 | – | – |
| Performance Max DE | +7.1% budget | Target ROAS 9.9 → 9.43 | 3.44 | +20 153 | 56 020 | 4.73 |
| YouTube USA | +50% budget | Target ROAS 1.0 → 0.75 | 6.66 | +18 092 | – | – |
| Search Generic USA | +37.9% budget | Target CPA 31 → 50.09 | 4.27 | +5 818 | 2 976 | 4.66 |
| Shopping DE | −34% budget | Target CPA 24 → 17.94 | 2.04 | −4 856 | 6 729 | 2.43 |
Illustrative demo data. Demo account figures are in USD.
Two details are worth pointing out. The YouTube campaigns didn't come back in last week's model pull for Google Ads, so the assistant leaves those cells empty and says so, rather than filling them from somewhere else. And the Shopping DE line is a cut: the negative impact is the incremental sales you give up in exchange for the spend you save. It has the lowest marginal return in the list, so that trade is the point.
The draft message
From those rows the assistant writes something like this, one line per campaign:
- YouTube DE: raise budget about 39% and lower target ROAS from 0.8 to 0.64. The next dollar here still returns about 4.5.
- Performance Max DE: raise budget about 7% and lower target ROAS from 9.9 to 9.43. Our biggest campaign, ROI 4.73 last week.
- YouTube USA: raise budget 50% and lower target ROAS from 1.0 to 0.75. Highest marginal return on the list at 6.66.
- Search Generic USA: raise budget about 38% and move target CPA from 31 to 50.09.
- Shopping DE: cut budget 34% and tighten target CPA from 24 to 17.94. Lowest marginal return on the list.
Then it stops and waits for you.
Where the message goes
Posting to Slack, Teams or email isn't something Sellforte does. It depends on your assistant having its own connector for that tool, used alongside Sellforte MCP. That's why the second prompt matters. A posted message can't be taken back, and the wording is the assistant's, so read it first. In this run we stopped at the draft.
Why it matters
A recommendation is only worth something if someone can actually carry it out. Agencies work from clear instructions: campaign, setting, new value, reason. Handing them that every week, straight from the model, means the bid change that goes live is the one the analysis supports, not a paraphrase of it.
It also keeps the reasoning visible. When the agency sees "lowest marginal return on the list" next to a cut, it can push back with its own context, which is a much better conversation than arguing about platform ROAS.
One caveat: the recommendations are a current snapshot. Run the list close to when you send it, not days before.
Try these follow-ups
- Do the same for Meta Ads, at ad set level.
- Which of last week's actions did the agency already apply, judging by spend this week?
- Rewrite the list in German for our agency in Germany.
- Only include changes worth more than 5 000 a week in incremental sales.
Related
Try it yourself
If you'd like to see your own campaigns in a list like this, book a demo. Already using Sellforte? Read how to connect Sellforte MCP to your LLM client and give it a go.
More use cases are on their way to this library.
