©2026 ZenABM - All Rights Reserved.
I know many people don't trust AI to "run" their LinkedIn Ads for them, and I'm not delusional: it will never replace a seasoned LinkedIn Ads expert with tons of tacit knowledge and experience across dozens of clients from different industries (well, never say never, but at least not for the time being).
BUT ignoring AI automations for LinkedIn Ads altogether is like deciding to ride a horse instead of driving a car.
They can help you with many tedious tasks that would otherwise fall through the cracks, e.g. finding ads that are decaying, adjusting bids on ad sets that are under or overspending, building retargeting campaigns from page visits for text-ad based website visitor deanonymisation... and now that the ZenABM MCP server can read your account and write to it, it has become dead simple to build them and have Claude run them every week for you.
Below are the 37 workflows and Linkedin Ads agents I would build for ABM on LinkedIn: 3 we already run on our own account, and 34 more drawn from the swipe file of posts by LinkedIn Ads practitioners I have been saving "for later". Later is now 😅
Each one is labelled ✅ Read only (it reports) or ⛔ Agent executes actions for you (it proposes numbered changes and applies only the ones you approve). Every card ends with the prompt to paste into Claude.
If you are new to feeding Claude the right background, start with context engineering for LinkedIn ABM with Claude Code.
You need literally 3 steps upfront, and then you can create them with simple prompts in Claude Cowork:
claude.ai/settings/connectors).https://app.zenabm.com/api/mcp
Ok let's see all of them now!
It works now because of two ingredients: a server that writes, and scheduling with approval. Most LinkedIn Ads "MCPs" are read only, so they answer questions and then you go and click in Campaign Manager. The ZenABM MCP server also creates and edits campaigns, ad sets and ads, changes budgets, bids and statuses, and manages exclusions and audiences, and that is what turns a report into an automation.
ZenABM's MCP server connects your LinkedIn ads to Claude, ChatGPT or Cursor, and it writes as well as reads: you can create and manage campaigns, ad sets and ads straight from Claude. I covered the read side in our guide to the LinkedIn Ads MCP server and the wider idea in AI agents for LinkedIn Ads.
The second ingredient is scheduling. Claude runs a prompt on a schedule, delivers the result to Slack, and waits for your "approve 1, 3" before it touches LinkedIn.
The ideas come from two years of saving posts by Anthony Blatner, Ivars Krutainis, Ali Yildirim, Justin Rowe and Patrick Cumming, each describing a check they run by hand. I built the first three cards this week on our own account, and the screenshots are from my Claude and Slack. The prompts never name MCP tools, only how many are used: you say what to fetch and change, and the server picks the tool.
These decide which ads to pause, replace, promote or refresh, and why.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 5 ZenABM MCP tools.
ZenABM features: company engagements, automated AI reports.
In every ad set, the agent finds ads with a frequency over 1 and the lowest landing page CTR, then proposes PAUSE or REPLACE, numbered, and names the better ad in the same ad set to move budget to. I reply with the numbers and it pauses only those. A lone ad gets REPLACE, never PAUSE, because pausing it would stop the ad set.
Frequency = impressions / members reached. eCTR (effective CTR, the same as landing page CTR) = landing page clicks / impressions. TLA likes and document scrolls do not count as clicks. I use eCTR from here on.
ZenABM's dashboards show eCTR out of the box, so you don't have to pull landing-page clicks from the Campaign Manager engagement report and divide them by impressions yourself.
This was the whole brief:
The first run checked 13 ad sets and 71 ads and flagged 57, each with a reason and a number. Here is the result as it landed:
It lands in Slack every Monday:
Inspired by Anthony Blatner, who monitors the CTR of every ad through Claude, and Philip Ilic, who says most accounts notice fatigue 4 to 5 weeks late.
Every Monday at 09:00, for every live ad set in my ZenABM account, pull the last 7 days of impressions, members reached and landing page clicks per ad. Compute frequency (impressions divided by members reached) and eCTR (landing page clicks divided by impressions). In each ad set, list the ads with frequency over 1 and the lowest eCTR, name the ad in the same ad set with the best eCTR as the one the budget should move to, and propose PAUSE, or REPLACE if it is the only ad in the set. Number every proposal and post the list to the #ads-agents Slack channel. Pause only the ads I approve by number, never the only ad in an ad set, and confirm each pause back to me.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 5 ZenABM MCP tools.
ZenABM features: build and manage campaigns with AI, expert skills.
This is the other half of the fatigue check. It lists ad sets with no new ad in 30 days, drafts 2 replacements on fresh angles using the account's best eCTR ad as the brief, and uploads them as DRAFT.
Thought Leader Ad sets get fresh content weekly once frequency passes 2; image and document sets every 30 days.
Inspired by Ali Yildirim (no new creative in 60 days means moving 1 to 3% of budget to testing), Gabriel Ehrlich (a TLA starts to die after frequency 2) and Patrick Cumming (10 to 20 variants every 2 to 4 weeks).
Every Monday at 09:00, list my live ad sets where the newest ad was created more than 30 days ago (7 days for Thought Leader Ad sets with frequency over 2). For each, take the ad with the best eCTR in the account as the brief, draft 2 new ads on different angles in the same format, upload them to that ad set as DRAFT, and send me the previews. Do not set anything live.
⛔ Agent executes actions for you (needs approval). Daily at 09:00. Uses 3 ZenABM MCP tools.
ZenABM features: bidding agent.
Every ad launched in the last 2 to 7 days is scored against its ad set average, with early CTR inflation discounted, and gets a keep, watch or pause call.
After 48 hours and 500+ impressions, discount the ad's CTR by 25%. If it is still under the ad set average, it is probably a below-average ad.
Inspired by Kirill Vdov: you can guess with 70% certainty how a campaign will do after 1 to 2 days of CTR, and early CTR runs 20 to 30% high.
Every morning at 09:00, list ads created 2 to 7 days ago with 500+ impressions. For each, compare its eCTR, discounted by 25%, with the average eCTR of its ad set, and propose keep, watch or pause. Number the pause proposals, send the list to the #ads-agents Slack channel and pause only the ones I approve.
⛔ Agent executes actions for you (needs approval). Last Friday of the month. Uses 3 ZenABM MCP tools.
ZenABM features: build and manage campaigns with AI.
Ads that held strong dwell time and eCTR for a full month get copied into the always-on ad sets. Thought Leader Ads are judged on eCTR, not on engagement.
Promote when dwell time is 10 seconds or more for 30 days, or eCTR is in the top quartile for its format (TLA top quartile is 0.60%+).
Inspired by Valentin Dimitrov (10 seconds of dwell held for a month earns a permanent slot) and Philip Ilic, who showed engagement and landing page CTR rank TLAs very differently.
On the last Friday of each month, pull weekly dwell time and eCTR per ad for the last 4 weeks. List ads with dwell time of 10 seconds or more in every week, or eCTR in the top quartile for their format, and propose duplicating each into the matching always-on ad set. Do it only for the ones I approve.
✅ Read only. 1st of the month. Uses 2 ZenABM MCP tools.
ZenABM features: dashboards, automated AI reports.
One blended CTR hides everything. A TLA at 3% engagement CTR looks amazing next to a 0.5% image ad until you look at landing page clicks, so this report compares every format on two like-for-like metrics:
Our own ZenABM benchmarks for reference: median eCTR is 0.29% for TLAs, 0.42% for single image ads, 0.24% for video and 0.32% for carousel, and the ABM median across formats is CTR 0.69%, CPC $11.04, CPM $78.30.
Inspired by Philip Ilic, who found an unspecialised LLM called a 3% TLA CTR "way above average" when for TLAs it is low.
On the 1st of each month, group last month's ads by format (single image, video, document, carousel, text, spotlight, Thought Leader Ad). For each format show spend, impressions, landing page clicks, eCTR and eCPC, and compare them with the ZenABM benchmark for that format. List the 3 best and 3 worst ads per format on eCTR and eCPC. Post it to the #ads-agents Slack channel.
✅ Read only. Every Monday at 09:00. Uses 3 ZenABM MCP tools.
ZenABM features: company engagements.
A TLA with no link in the post collects likes and comments but zero landing page clicks, so it never shows up in eCTR. This check lists every running TLA without a link, with its edit link and author.
Recommended action: add the landing page link to the organic post, then the sponsored version picks it up.
Every Monday at 09:00, list my running Thought Leader Ads whose sponsored post contains no link, with the ad name, the link to edit it in Campaign Manager, the post author and the author's LinkedIn profile URL if you can find it. Recommend adding the landing page link to each post and, where the author is on our team, draft the one-line message asking them to add it.
✅ Read only. Daily at 08:00. Uses 4 ZenABM MCP tools.
ZenABM features: automated AI reports, dashboards.
Each ad set's last 3 days are compared with its own previous 14 days, and jumps in eCPC or CPM and drops in eCTR are flagged with the likely cause: fatigue, saturation, a bid change or an audience change.
Flag when eCPC is up 25%+ or eCTR is down 30%+ against the ad set's own 14-day baseline, on at least $50 of spend.
Inspired by Ali Yildirim, who treats a 25% jump in cost per result as the fatigue signal, and Jack Cleere, who has Claude spot unusual changes across campaigns.
Every morning at 08:00, for each live ad set with at least $50 spend in the last 3 days, compare eCPC, CPM and eCTR for the last 3 days with the same ad set's previous 14 days. Flag eCPC up 25%+ or eCTR down 30%+, check whether the bid, the budget or the audience changed in that window, and give the most likely cause. Post the flagged ad sets to the #ads-agents Slack channel.
After the ads comes the money: keeping every ad set fed, paced and bid correctly.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: bidding agent, dashboards.
The agent shows every live ad set's daily spend as a share of its daily budget, then proposes a 10% bid cut for overspenders and a 10% bid increase for underspenders. One strong day never counts alone, because LinkedIn may legitimately spend up to 150% of the daily budget, and auto-bid ad sets (for example MAX_REACH) get a budget suggestion instead, since a bid change does nothing there.
Overspending = 110%+ of budget on average, or 130%+ on two or more days. Underspending = under 80% on average.
Nothing changes until I reply "approve all", "approve 1, 3" or "skip". Before applying, it re-checks the live bid and confirms each change as old to new.
Here is the task as it runs, with the thresholds and the approval flow:
Inspired by Ivars Krutainis, who calls his colour-coded morning pacing tracker in Slack the most useful AI tool his team has built.
Every Monday at 09:00, for every live ad set, pull daily spend and daily budget for the last 7 days. Show spend as % of daily budget per day with the weekly average. Flag ad sets averaging 110%+ or hitting 130%+ on two or more days as overspending and propose a 10% lower bid; flag ad sets under 80% as underspending and propose a 10% higher bid; list auto-bid ad sets separately with a 10% budget cut as the suggestion. Number every proposed change and post the list to the #ads-agents Slack channel. Apply only the changes I approve by number, re-check the live bid first, and confirm each as old to new.
⛔ Agent executes actions for you (needs approval). Daily at 07:30. Uses 5 ZenABM MCP tools.
ZenABM features: Budget Savers, bidding agent.
LinkedIn turns Audience Expansion and the LinkedIn Audience Network on by default, and Kristof Bardos says those two plus Maximum Delivery can waste about 90% of a first test budget. The agent checks every live ad set for those three, for a "recent" instead of "permanent" location setting, and for missing standard exclusions.
Then it does what a report cannot: it switches on ZenABM's Budget Savers, which keep Audience Expansion and the Audience Network off team-wide for every current and future ad set. New ad sets are covered within an hour, so the tripwire only has to catch Max Delivery and the location setting afterwards.
Inspired by Ivars Krutainis, who built a one-prompt Claude check for exactly this, and Kristof Bardos.
Every day at 07:30, check the settings of all live ad sets for LinkedIn Audience Network on, Audience Expansion on, Maximum Delivery bidding and a non-permanent location setting. Post a numbered list of breaches with the fix for each to the #ads-agents Slack channel. Apply only the fixes I approve. Then check the Budget Savers status in ZenABM and, if either saver is off, propose switching it on for all ad sets so the two network settings stay off for every future ad set too.
⛔ Agent executes actions for you (needs approval). Set once, reviewed monthly. Uses 4 ZenABM MCP tools.
ZenABM features: campaign scheduling.
Justin Rowe runs his campaigns only between 5am and 10pm, because LinkedIn will not do it natively, and says turning off overnight and weekend spend saves 10 to 15% straight away. ZenABM's scheduling does it natively: you paint the hours on a week grid in your audience's timezone and the campaigns switch on and off for you.
One caveat matters: this only works if your audience is geographically homogenous. If one ad set targets the US and Europe together, night in London is the working afternoon in San Francisco and you would pause ads on people who are in the feed. Split the ad sets by region first, or skip this one.
Inspired by Justin Rowe.
List my COLD campaigns with the countries each one targets. For campaigns whose audience sits in one timezone band, propose a weekday 05:00 to 22:00 schedule in that audience's timezone using ZenABM's campaign schedule, and leave warm and retargeting campaigns running all the time. Apply the schedules I approve. On the 1st of each month, compare eCPC for the 30 days before and after and tell me in Slack whether to keep it.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: bidding agent.
Ali Yildirim's audits of 100+ accounts find the same thing: 20+ ads on $100 a day, so each ad gets $4 and nothing exits learning. The agent flags ad sets where the budget is spread too thin and proposes which ads to pause, keeping the best eCTR ones.
Flag when daily budget / active ads is under $25, or when an ad set runs more than 5 active ads.
Inspired by Ali Yildirim: ads per ad set = monthly budget / (days in month x 25).
Every Monday at 09:00, for every live ad set, divide the daily budget by the number of active ads. Flag ad sets under $25 per ad or with more than 5 active ads. Propose which ads to pause, keeping the ones with the best eCTR, number the proposals and pause only the ones I approve.
✅ Read only. 1st of the month, and whenever a new ad set is created. Uses 4 ZenABM MCP tools.
ZenABM features: ABM campaign dashboards, company engagements.
Most people get this one wrong. Patrick Cumming sees it every week he audits an account: a 1M+ audience paired with a $10K budget, so everyone sees one ad, nobody remembers it, and the company concludes LinkedIn Ads don't work. Kirill Vdov's fix is a formula:
Monthly budget needed = audience / 2 / 1,000 x CPM x 6 / 0.75
The "/ 2" is because only about half of any LinkedIn audience is logged in during a month, the "x 6" is a monthly frequency of 6 (the minimum where people stop ignoring you), and "/ 0.75" is a pacing allowance.
A worked example with our numbers: our COLD Q4 TLA ad sets on brand awareness run at a CPM of about $3. For an audience of 50,000 that is 50,000 / 2 / 1,000 x $3 x 6 / 0.75 = $600 a month. An image ad set at a $60 CPM needs $12,000 a month for the same audience. That is 20x the budget, which is why format matters as much as audience size.
The agent runs this for every ad set with its actual CPM, lists those funded below half of what they need, and gives the audience size your budget would fit, in line with Patrick's target of 80%+ penetration and 10+ frequency per quarter. Campaign Manager cannot always tell you the audience size (we hit "There was an error getting target audience size information" twice this week), and ZenABM's audience count can.
Inspired by Kirill Vdov and Patrick Cumming.
On the 1st of each month and whenever a new ad set is created, for every live ad set pull the audience size and the actual CPM of the last 30 days. Calculate the monthly budget it needs as audience / 2 / 1000 x CPM x 6 / 0.75 and compare it with the current monthly budget (daily budget x 30). List ad sets funded below 50% of the need, and for each give the audience size that the current budget would reach at frequency 6. Post the table to the #ads-agents Slack channel.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 3 ZenABM MCP tools.
ZenABM features: bidding agent.
Every manual bid is compared with LinkedIn's suggested range, stepped down where the ad set still spends its budget and up only where it under-delivers. These are expert rules, not ZenABM advice.
Start 10 to 20% below the low end of the suggested range. Step down 10% if the ad set spends 95%+ of its budget; step up 10% if it spends under 70%.
Inspired by Maximilian Herczeg (start about 10% below the low end), Patrick Cumming (bid at 25% of LinkedIn's recommendation) and Anthony Blatner, who says automated bid optimisation saved him $31K.
Every Monday at 09:00, for all manual-bid ad sets, fetch LinkedIn's suggested bid range and the last 7 days of spend as % of budget. Show my bid against the range. Propose a 10% lower bid for ad sets spending 95%+ of budget and a 10% higher bid for ad sets under 70%, numbered. Apply only the ones I approve and confirm each as old to new.
✅ Read only. 1st of the month. Uses 4 ZenABM MCP tools.
ZenABM features: dashboards, bidding agent.
eCPC and CPM by objective and bidding type across the account. Ad sets on Website Conversions or Maximum Delivery that pay several times more per landing page click are listed with the cheaper objective to test.
Any ad set paying 2x+ the account median eCPC for its format is flagged.
Inspired by Ali Yildirim, who ran this across 10 accounts with Claude Code, and Tim Davidson's retest, where Website Conversions cost $539 CPM against $139 on Engagement for the same audience.
On the 1st of each month, group last month's spend by ad set objective and bidding type and show eCPC and CPM for each group. List ad sets paying 2x+ the median eCPC for their format, with the objective and bidding type you would switch them to and why. Post it to the #ads-agents Slack channel.
⛔ Agent executes actions for you (needs approval). First Monday of the month. Uses 4 ZenABM MCP tools.
ZenABM features: ABM campaign dashboards, bidding agent.
Ad sets are ranked by ABM-influenced pipeline per $ over 90 days (eCPC where there is no pipeline yet), and the agent proposes moving budget from the bottom 3 to the top 3 while keeping 20% of spend for tests.
Never cut an ad set younger than 30 days. Patrick Cumming killed a podcast-snippet campaign after a month for low CTR, and two months later the viewers had converted and it was his lowest CPA campaign.
Inspired by Liam Collins and Patrick Cumming.
On the first Monday of each month, rank live ad sets older than 30 days by ABM-influenced pipeline per $ over the last 90 days, using eCPC where an ad set has no pipeline. Propose moving budget from the bottom 3 to the top 3 while keeping 20% of total spend on test ad sets. Number the budget changes, post them to the #ads-agents Slack channel and apply only the ones I approve.
✅ Read only. 1st of the month. Uses 4 ZenABM MCP tools.
ZenABM features: account stages, ABM campaign dashboards.
The share of spend on cold, warm and retargeting campaigns against your target split, plus the number that actually judges cold spend: how many new accounts entered the warm stage that month.
Default target: 80% cold, 20% warm. Flag drift of more than 10 points.
Inspired by Patrick Cumming and Justin Rowe, who judges cold spend by how many new warm accounts it creates.
On the 1st of each month, show last month's spend split between campaigns named COLD and WARM against an 80/20 target, and count the companies that entered the warm ABM stage from cold during the month. Flag a drift of more than 10 points and post it to the #ads-agents Slack channel.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: bidding agent, build and manage campaigns with AI.
Anthony Blatner warns you can spend $10K on one Thought Leader Ad while none of your other ads get shown. The agent flags ad sets where one ad takes most of the spend and proposes moving it to its own ad set or switching rotation to even.
Flag when one ad took over 60% of its ad set's spend in 7 days while the others sit under 10% each.
Inspired by Anthony Blatner and Kirill Vdov.
Every Monday at 09:00, for each live ad set with 2+ ads, compute each ad's share of the last 7 days spend. Flag ad sets where one ad is over 60% and the rest under 10% each. For each, propose either moving the dominant ad into a copy of the ad set with its own budget, or switching the ad set's rotation to even. Apply only the ones I approve.
⛔ Agent executes actions for you (needs approval). Scheduled for 1 December and 2 January. Uses 3 ZenABM MCP tools.
ZenABM features: bidding agent, campaign scheduling.
December lead-gen budget gets front-loaded into the first two weeks while calendars are open, BOFU spend is cut over the holiday weeks when CPMs jump, and January restores everything.
Deliver about 80% of December lead-gen budget by 15 December; cut BOFU ad sets by 50% from 20 December; restore on 2 January.
Inspired by Ali Yildirim. Anthony Blatner saw CPMs jump 20 to 100% over a holiday weekend while BOFU conversions went to zero.
On 1 December, propose a December budget plan for all live ad sets that delivers 80% of the lead-gen budget by 15 December and cuts BOFU ad set budgets by 50% from 20 December, numbered per ad set. On 2 January, propose restoring the original budgets. Apply only what I approve.
With budgets sorted, the question is whether the right accounts actually see the ads, and the wrong ones do not.
⛔ Agent executes actions for you (needs approval). On demand. Uses 6 ZenABM MCP tools.
ZenABM features: build and manage campaigns with AI, exclusions.
One prompt built a separate text ads campaign for people who visited our competitor comparison pages: 2 ad sets of 8 text ads each, website visits objective, $20 a day each, competitors excluded. Everything was created as DRAFT and only set live after review.
The targeting fix was plain English too. I told it to drop the impression caps, the geo and the industry filters and keep only job titles, and the audience went from 860 to about 1,800 members per ad set. Here is the result read back from LinkedIn, with objective, bid, budget and what was included, dropped and excluded:
Inspired by Ivan Falco, who runs competitor conquesting as one of his core ad plays.
Create a new campaign called "Competitor page retargeting" with 2 text ad sets of 8 text ads each, targeting my website retargeting audiences for the competitor comparison pages. Objective website visits, manual CPC, $20 a day per ad set, Audience Network and audience expansion off, keep only job-title targeting, exclude our competitors by employer, and create everything as DRAFT. Show me the audience size per ad set and the settings before I set it live.
✅ Read only. Daily at 10:00. Uses 4 ZenABM MCP tools.
ZenABM features: dashboards.
Active ad sets or ads with 0 impressions yesterday, or 0 clicks on 1,000+ impressions over two days, usually mean a technical break rather than a bidding problem: a rejected ad, an audience under 300, an expired schedule or a payment issue.
This week LinkedIn rejected two of our video ads as "political advertising" (they are about ABM for agencies 🤡). Without an alert, those ads would have sat rejected for days:
Inspired by Ivars Krutainis, whose second alert fires only on 0 impressions or 0 clicks.
Every morning at 10:00, list active ad sets and ads with 0 impressions yesterday, or 0 clicks on 1,000+ impressions over the last two days. For each, check the ad review status, the audience count and the campaign schedule, and tell me the most likely cause in the #ads-agents Slack channel.
✅ Read only. Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: company engagements, ABM campaign dashboards.
For each ABM campaign, the share of target accounts that got impressions, the average impressions per account, and the accounts already over-served.
Cold target: 80%+ of accounts reached and 8 to 12 impressions per account per 30 days. Retargeting: more than 10 impressions in 90 days with no stage change is wasted spend.
Here is the per-account view on our COLD Q4 campaign over the last 30 days: 13,460 companies targeted and 1,739,673 impressions, so 129 per company on average. The distribution is the whole story:
Company | Impressions, last 30 days | vs campaign average (129) | Landing page clicks |
|---|---|---|---|
Pavilion | 7,297 | 57x | 0 |
5,766 | 45x | 14 | |
Clay | 3,255 | 25x | 0 |
RevGenius | 3,241 | 25x | 0 |
Boardy | 2,494 | 19x | 0 |
TikTok | 2,172 | 17x | 0 |
Nova | 2,163 | 17x | 0 |
Stealth AI Startup | 2,147 | 17x | 0 |
Source: ZenABM company engagements, COLD Q4 (Total), 15 September to 15 October 2026.
Three of the top eight are communities we belong to, not customers. That is automation 22's job.
Inspired by Patrick Cumming, who says LinkedIn Ads start working at 80% reach and 10 to 15 frequency every 90 days.
Every Monday at 09:00, for each ABM campaign, pull the companies targeted and the impressions per company for the last 30 days. Show the % of target accounts with any impressions, the average impressions per account, and a table of the 10 accounts with the most impressions with their multiple of the campaign average and their landing page clicks. Flag campaigns under 80% penetration and list accounts over 10 impressions in 90 days with no ABM stage change. Post it to the #ads-agents Slack channel.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 5 ZenABM MCP tools.
ZenABM features: company engagements, exclusions, impression capping.
Tim Davidson found one large company pulling 38% of a campaign's budget, and in the table under 21 Pavilion got 57x the average. The agent finds accounts eating a disproportionate share of impressions and proposes one of two fixes:
Flag any single company over 15% of an ad set's impressions in 7 days, or any account over 20 impressions in 30 days.
ZenABM applies your exclusions across all campaigns, including new ones, so you don't have to re-upload the list each time you launch.
Inspired by Tim Davidson, Philip Ilic and Miles Beard, whose 70/30 split is the mechanism.
Every Monday at 09:00, for each live ad set, list companies over 15% of its impressions in the last 7 days or over 20 impressions in the last 30 days, with their company size and whether they are on our target list. For each, propose exclude (not a target) or split (a target: duplicate the ad set for those accounts at 25% of budget and exclude them from the original at 75%). Number the proposals and apply only the ones I approve.
⛔ Agent executes actions for you (needs approval). 1st of the month. Uses 4 ZenABM MCP tools.
ZenABM features: company engagements, build and manage campaigns with AI.
The mirror image of 22. Target accounts with no paid impressions in 90 days get their own ad set, with a different objective or a higher bid, as long as there are enough of them to clear LinkedIn's 300-member minimum.
Inspired by Tim Davidson, who lifted account reach 28% in 7 days this way.
On the 1st of each month, list target accounts with 0 impressions in the last 90 days. If a matched audience of them would have 300+ members, propose a new ad set for them with the engagement objective and manual CPC 20% above the current bid, as DRAFT, and show me the audience size before creating it.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: company engagements, exclusions.
Kirill Vdov says LinkedIn will mismatch 10 to 20% of an uploaded company list. This check compares the companies LinkedIn actually serves with your target list and lists the mismatches: wrong companies matched, non-ICP enterprises draining budget, and target accounts missing.
It needs your target account list as a CSV uploaded to Claude (company name and LinkedIn page URL, the same file you uploaded to Campaign Manager), because it compares served companies against your list. Upload it once in the project and the scheduled task reuses it.
Inspired by Kirill Vdov.
Every Monday at 09:00, read the target account list CSV in this project (columns: company name, LinkedIn page URL). For each ABM ad set, pull the companies served in the last 7 days with their impressions. List served companies not on the CSV with 1%+ of impressions, and CSV companies with no impressions at all. Propose exclusions for the served non-targets, numbered, and apply only the ones I approve.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 5 ZenABM MCP tools.
ZenABM features: company engagements, job title engagement, exclusions.
ZenABM shows, per ad set, which companies got your impressions (with size and industry) and which job titles saw and engaged. The agent compares both with your ICP and proposes exclusions for the leaks: companies under your minimum size, excluded industries, competitors and agencies, and titles outside your personas.
ZenABM groups ad engagement by job title, so you can check whether your spend actually reached the buying committee.
Flag companies with spend that are under 11 employees, in an excluded industry, a competitor or an agency. Flag titles outside the persona list with 5%+ of impressions.
Two warnings before the agent excludes anything:
Inspired by Philip Ilic (11 to 200 employee targeting still serves 1 to 10 employee firms unless excluded), Anthony Blatner (an exclusion across 10+ campaigns in one sentence) and Justin Rowe, whose skill scans 227 demographic segments for wasted budget.
Every Monday at 09:00, for each live ad set, pull the companies that got spend last week with their employee count and industry, and the job titles that got impressions. List companies under 11 employees, in an excluded industry, competitors and agencies, and job titles outside my persona list with 5%+ of impressions. Mark founder, board member, advisor and owner titles as "review, do not exclude". Propose company and title exclusions across all campaigns, numbered, and apply only the ones I approve.
Once the right accounts engage, someone has to call them.
⛔ Agent executes actions for you (needs approval). Daily at 08:30. Uses 4 ZenABM MCP tools plus your CRM MCP and Slack.
ZenABM features: account stages, intent signals, CRM sync.
When a target account moves into a higher ABM stage, the agent looks up the account owner in your CRM through the HubSpot or Salesforce MCP (an easy lookup once the company is synced), sends that owner a Slack message with what the account engaged with and a suggested next step, and creates a CRM task.
Tier 1 accounts get their own message. Everyone else goes in a daily digest to #sales.
Inspired by Ivan Falco, who routes every signal to the account owner automatically.
Every morning at 08:30, list companies that entered a higher ABM stage yesterday with the ads and pages they engaged with. For each, find the company owner in HubSpot. Send one Slack DM per Tier 1 account to its owner with the engagement summary and a suggested next step, create a HubSpot task for the owner, and post a digest of the remaining accounts to #sales. Show me the messages before sending the first time; after that, send automatically.
✅ Read only. Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: intent signals, CRM sync.
These are accounts that engaged with your ads and visited high-intent pages but have no contact or deal in the CRM. Justin Rowe's first question for any ads MCP, because that is an outbound list of active shoppers.
Inspired by Justin Rowe.
Every Monday at 09:00, list companies that engaged with our ads in the last 30 days and sit in a warm or hot ABM stage but have no deal in the CRM and no activity logged in 30 days. Include their top intent signal and the pages they visited. Post it to #sales as an outbound list.
✅ Read only. Every Monday at 09:00. Uses 3 ZenABM MCP tools.
ZenABM features: deal attribution dashboard, intent signals.
Open deals whose company engaged with your ads for weeks and then stopped. Sudden silence after weeks of engagement is an early warning the AE should hear before the deal stalls.
Flag open deals with 3+ engaged weeks in the last 8 and 0 engagement in the last 14 days.
Inspired by Justin Rowe.
Every Monday at 09:00, for every open deal, pull the company's weekly ad engagement for the last 8 weeks. List deals whose company engaged in 3 or more of those weeks but not in the last 14 days, with the deal owner and the last ad they engaged with. Post it to #sales.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses 6 ZenABM MCP tools.
ZenABM features: account stages, CRM sync, exclusions, build and manage campaigns with AI.
I do not recommend excluding customers, open deals or closed-won accounts from your ads. I recommend moving them: a company with an open deal should stop seeing cold ads and start seeing case studies and testimonials aimed at its whole buying group, and a customer should see expansion and product content, not "book a demo".
So each week the agent adds companies whose deals opened, closed-won or churned to the exclusions of the COLD campaigns, and moves them to WARM or to the deal-open or customer campaign you defined. ZenABM will soon be able to move companies between stage campaigns on its own; until then the agent proposes the moves and you approve. ZenABM moves every account through the ABM stages you define, from ad engagement and CRM data, so the stage is computed rather than updated by hand.
The deal-open audience has to stay above 300 members, otherwise the account stays where it is.
Inspired by Ivan Falco, who runs testimonial and case-study ads at the full buying group once an opportunity opens.
Every Monday at 09:00, list companies whose deals opened, closed-won or churned in the last 7 days. For each, propose: add to the exclusions of all COLD campaigns, and add to the matched audience of the WARM campaign (open deals: the "Deal open" ad set; customers: the "Customer" ad set). Check the target audience stays over 300 members and tell me if it does not. Number the moves and apply only the ones I approve.
⛔ Agent executes actions for you (needs approval). Daily at 09:00. Uses 3 ZenABM MCP tools plus a webhook.
ZenABM features: webhooks, intent signals, company engagements.
Companies that engaged with your Thought Leader Ads get sent to Clay (or HeyReach, or Instantly) with the post they engaged with, so outreach can come from the same exec whose post they saw.
Can the agent build the webhook? Half of it. Through the MCP it creates the signal in ZenABM ("engaged with a TLA, on the target list, not sent in 30 days"), but you add the webhook endpoint once in ZenABM's webhook settings, pointed at your Clay table. After that every matching company is pushed automatically, no agent needed.
Inspired by Ali Yildirim, who pushes TLA engagers through Clay into HeyReach or Instantly.
Create a signal in ZenABM called "TLA engagers" for target-list companies that engaged with any Thought Leader Ad in the last day, excluding companies that matched in the last 30 days. Then tell me the webhook URL field to fill in ZenABM so each matching company is pushed to our Clay table with the company, the post and the engagement type. Until the webhook is live, every morning at 09:00 post the day's matching companies to #sales.
Finally, proving what the ads did and reading it in one place.
✅ Read only. 1st of the month. Uses 3 ZenABM MCP tools.
ZenABM features: deal attribution dashboard, multi-channel revenue attribution.
For every ABM-influenced deal closed last month, this shows which campaign and ad reached the account first and how many ad touches happened before the deal opened. Most marketers guess at first touch, while the deal timeline has it.
ZenABM's deal attribution shows the ad touchpoints behind every deal, before it opened and before it closed, so you can see whether ads accelerated it.
Inspired by Justin Rowe.
On the 1st of each month, list ABM-influenced deals closed last month. For each, read the company timeline and give the first campaign and ad that reached the account, the date, and the number of ad touches between that first impression and the deal opening. Group the deals by first-touch campaign and post it to the #ads-agents Slack channel.
✅ Read only. Every Monday at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: automated AI reports, ABM campaign dashboards.
A short Monday report that leads with ABM-influenced pipeline and pipeline per $, then accounts that moved stage, spend against plan, and the three things to do this week. Only ABM-influenced demos, deals and pipeline count.
Ali Yildirim loads his Claude projects with historical audits and SOPs so the agent analyses instead of just pulling data. Do the same: put your campaign plan and your last three decisions in the project.
Inspired by Ali Yildirim.
Every Monday at 09:00, post to the #ads-agents Slack channel: ABM-influenced pipeline and pipeline per $ for last week against the previous week, companies that moved to a higher ABM stage, spend against the monthly plan, and the 3 most important actions for this week based on the campaign plan in this project.
✅ Read only. 2nd of the month. Uses 4 ZenABM MCP tools.
ZenABM features: job title engagement, ABM campaign dashboards.
Performance clustered by persona, seniority, industry and company size: who sees the ads, who clicks, who becomes pipeline, and where the budget is wasted.
Flag segments with 5%+ of spend and below-median pipeline per $.
Inspired by Justin Rowe.
On the 2nd of each month, pull last month's impressions, landing page clicks, spend and pipeline by job title, seniority, industry and company size. Group the job titles into my personas, show eCTR, accounts engaged and pipeline per $ for each segment, and flag segments with 5%+ of spend and below-median pipeline per $. Post it to the #ads-agents Slack channel.
✅ Read only. Daily at 18:00. Uses 4 ZenABM MCP tools.
ZenABM features: dashboards.
A snapshot of every ad set's budget, bid, status and targeting each evening, and a post of what changed since yesterday. A bad edit (yours, a colleague's or an agent's) gets caught the same day.
Highlight budget or bid moves over 20%.
Inspired by Liam Collins, who lists change-log tracking among the places AI genuinely adds value today.
Every day at 18:00, pull budget, bid, status and targeting for every ad set and save them as today's snapshot in the "Change log" sheet in this project. Compare with yesterday's snapshot and post every change to the #ads-agents Slack channel, highlighting budget or bid moves over 20%.
✅ Read only. Weekdays at 09:00. Uses 4 ZenABM MCP tools.
ZenABM features: Agency MCP server, dashboards.
One Slack post covering every client workspace: spend against budget, pacing, zero-delivery issues, settings breaches and the clients that need attention today, colour-coded green, amber and red. ZenABM's Agency MCP server reaches every client workspace from one connection, so agency reporting doesn't mean one login per client.
Inspired by Ali Yildirim and Ivars Krutainis.
Every weekday at 09:00, go through every client workspace in the Agency MCP. For each, pull yesterday's spend against daily budget, the pacing for the month, ad sets with 0 impressions, and the Budget Savers status. Post one Slack message to #clients with one line per client coloured green, amber or red, and the reason for every amber and red.
✅ Read only. When a client account is connected. Uses 5 ZenABM MCP tools.
ZenABM features: Agency MCP server, expert skills.
A settings audit, budget fit (automation 12), format benchmarks (automation 5), targeting leaks (automation 25) and creative age (automation 2) on the last 90 days, written up as a one-page briefing with numbered recommendations for the strategist.
Inspired by Ali Yildirim: a strategist picking up a new account asks Claude for a briefing instead of spending a day in old Slack threads.
Run a full audit of the newly connected client workspace on the last 90 days: delivery settings, budget vs audience size per ad set, eCTR and eCPC by format against the ZenABM benchmarks, company and job title leaks, and ad sets with no new creative in 30 days. Write a one-page briefing with numbered recommendations, each with the number behind it.
⛔ Agent executes actions for you (needs approval). Every Monday at 09:00. Uses the same ZenABM MCP tools as the checks it runs.
ZenABM features: MCP server and skills, automated AI reports.
Once you run five or ten of these, you want one message, not ten. The obvious design, one agent that reads what the others posted to Slack, has two problems: not every output goes to Slack, and scheduled agents cannot talk to each other, so an approval sent to the digest never reaches the agent that proposed the change.
So this one reads nothing. It runs the weekly checks itself: the prompts for 1, 2, 6, 8, 11, 13, 17, 21, 22, 24, 25, 27, 28, 29 and 32 sit in a "Monday checklist" document in the same Claude project, and it works through them in order against live ZenABM data in one session. Then it merges the findings into one digest: numbers first, then one numbered list of every proposed change, deduplicated (the fatigue check and the budget-hog guard often flag the same ad), ordered by spend involved, with the check that produced it.
It applies the approved changes itself. Your "approve 1, 4, 7" is applied directly, after re-checking the live value, and confirmed as old to new. The daily checks (3, 7, 9, 20, 26, 30, 34) and the monthly ones keep running on their own, because waiting for Monday would defeat them.
Every Monday at 09:00, open the "Monday checklist" document in this project and run every check in it, in order, against my ZenABM account: 1, 2, 6, 8, 11, 13, 17, 21, 22, 24, 25, 27, 28, 29 and 32. Do not post the individual results anywhere. Merge them into one digest: first the numbers (spend vs plan, pipeline per $, accounts that moved stage, penetration per campaign), then one numbered list of every proposed change across all checks, deduplicated, ordered by the spend involved, with the check that produced it and the exact change (ad set or ad, current value, proposed value). Send the digest to my Slack DM and wait for my reply. For each number I approve, re-check the live value in ZenABM, apply the change through the ZenABM MCP, and confirm it back to me as old to new. Skip anything I do not approve.
Mostly the same things the experts behind these automations warn about, and each has a rule:
Unlike most LinkedIn Ads MCPs, which only read, ZenABM's also writes: with your permission it creates and edits campaigns, ad sets and ads, changes budgets, bids and statuses, manages exclusions and matched audiences, and sets schedules. That is why the ⛔ automations can apply what you approve instead of telling you what to click.
In this order, because each one feeds the next:
I will post the results of each one as it goes live, scoreboard style.
With the ZenABM MCP server, both. It reads performance, settings and company engagement, and it writes campaigns, ad sets, ads, budgets, bids, statuses, exclusions, matched audiences and schedules. Every automation marked ⛔ applies a change only after you approve it by number, and the ✅ ones never change anything.
No. The prompts describe what to fetch and what to change in plain English, and the MCP server picks the right tool. I only give the number of tools per automation so you know how much is happening behind each one.
Through Claude's scheduled tasks. Every weekly agent runs on Monday at 09:00 and posts to Slack, and the ⛔ ones wait for your reply before applying anything. Or let the agent of agents (37) run all the weekly checks in one session, send you one numbered list, and apply what you approve.
The settings tripwire audit (9) with Budget Savers switched on. It is the cheapest to build, and Audience Expansion plus the Audience Network are where most wasted LinkedIn budget goes.
It can, which is why it proposes and you approve. Since LinkedIn's March 2026 fuzzy-matching change, a small company can be matched to a large one with a similar name, so treat "too small" companies as accounts to look at, and never auto-exclude founder, board member or advisor titles.
So, should you let an agent run your LinkedIn Ads? Yes, with a number to approve in front of every change, like a hand on the wheel instead of a horse running free. Start with one of the ✅ ones this week, add a ⛔ one when you trust the numbers, and tell me which one you built first 😉