Adrails Docs

Automations

Workflows that read your ads and store on a schedule or a threshold, report what changed or moved, and post or act on what they find.

An automation is a workflow: one trigger, then steps that read data, filter it, ask an agent about it, and send the result or act on what crosses a line: pause it, resume it or change its budget.

Ask for one in the chat

Say what you want in the message box, in your own words: send me a daily report of what changed on my ad accounts, alert me when a campaign spends 100 today with no purchases, pause any ad with a ROAS under 1 after 50 spent.

  • One card of questions. A recurring report asks what you did not say: which accounts, how often, what time and where it goes, with the usual choices already selected, 09:00 among them. What you said is not asked again, and you can answer in the message box instead.
  • A card to create it. It says when it runs, what it reads and where it goes. Nothing exists until you choose Create automation, and it starts switched off: turn it on in Automations. The agent never switches an automation on or off.
  • The ad account's time zone. The hour you give or choose is read in the time zone of the account it watches.
  • With no channel connected. A rule that posts goes to your own email: Create automation adds an email channel called My email, if your role can add channels.

Change one from the chat

Ask for the change: make that report weekly, only on the Lead Form 01 campaign, move it to 7am, turn off the email. The agent can change the accounts, the campaigns, how often it runs, the hour, the day, the channel and the email.

A card shows each change with what it was, and the rest of the automation stays as it is. Nothing changes until you choose Apply change. When a report's name says how often it runs or what it reads, the name follows, and the card says so.

A report's period follows how often it runs when it is that frequency's own: yesterday for a daily report, the last 2 or 3 days for one every 2 or 3 days, the last 7 days for a weekly one, the previous month for a monthly one. A daily report made weekly then reads the last 7 days, and a Reads line on the card shows the move. A period you chose stays, and so does the one a budget or pause rule reads.

That report means the one this chat created or last changed. In another chat, name it; with several reports saved, the agent lists them and asks which. To add a report rather than change one, say another. Switching on, switching off and deleting stay in Automations.

Change reports

A change report lists what changed on your Meta and Google Ads accounts over a period, from Meta's activity log, Google Ads' change history and Adrails: one line per change with its time in the account's own time zone, the account, the campaign, ad set or ad, the old and new value of a budget or a status, and who made it. What was done in one batch stays one line, such as ad sets renamed together. The copy button under the answer gives it as plain text, ready to paste into a document.

  • Once: ask the agent, for example a report of everything that changed today on Core-4. The Reporter answers in the chat, and the report under the answer holds every line as one CSV and one PDF, with an account column, whatever the number of accounts. When the answer lists fewer lines than the report holds, it says how many are left out.
  • Whose changes: yours, one person's by name, or everyone else's, such as the changes I made yesterday or what Netos did this week. By default everyone's, each line naming who. Yours are those of the person who connected the platform to Adrails; when the log cannot tell, a card asks which name is yours. Google Ads names people by their email address.
  • What it leaves out: Meta billing the account and Meta's review of new or edited ads, which change nothing. Ask for the billing lines to see them. Store sales stay out too, unless you ask what the store sold.
  • On a schedule: every day, every 2 days, every week or every month. Ask for a recurring one, or accept the agent's offer after a report. Each run reads the days since the run before, or the period you stated, writes a new chat under Automations and notifies you. Send a notification by email adds an email with a link to that chat.
  • Which changes: made in Adrails, on the platform in Ads Manager or Google Ads, or both, which is the default.
  • How far: one or more accounts, up to 50 at once, or chosen campaigns with their ad sets and ads.

On Google Ads, the report tells apart what was done in Google Ads, Google Ads Editor, a script, a rule, Google's auto-applied recommendations and Adrails. Google keeps 30 days of change history: a report that reaches further back says what it cannot read.

An ad set deleted before Adrails last synced the account can no longer be tied to its campaign: a report kept to a campaign can miss it, and a report on the whole account shows it.

What moved

A what moved report lists the ads that moved between a period and the one just before it, on your Meta and Google Ads accounts, from their last sync. It gives no reason and no advice: only what the figures show.

  • Spending more or spending less: spend moved by 30% or more, and by at least 1% of the account's spend.
  • Dying slowly: spend or revenue down three periods in a row or more, by 20% or more in all.
  • Stopped: an ad that no longer runs, with its status, such as paused, archived or deleted, and its last figures.
  • Low ROAS: under 70% of the account's ROAS.

Each line names the ad with its ad set and its figures before and now. On Google Ads, a Performance Max line is an asset group, since it has no ads. When the period ends in days whose ROAS is still filling in, the report says so.

  • In the chat: ask what moved this week or which ads are dying. The answer is at ad level, over the period you name, or the last 7 days.
  • On a schedule: ask for a recurring one, every day, every 2 or 3 days, every week or every month. A period you state, such as the last 3 days every Monday, is kept over the schedule's own. Each run writes a chat under Automations with the report as a CSV and a PDF. No AI writes it, so each report costs 1 credit.

Start from a template

Automations has two tabs: yours, as a table you can also show as cards, and Templates. Open Templates, or choose New automation. The templates are one grid of cards, ready ones first, each with how often it runs, its credits and whether it is Ready, Needs 1 thing or Needs setup. Click a card to create it: it opens in the editor with what is still missing and the way to add each. On a phone, a tap creates it. Templates read the ad platform where most of your spend was over the last 30 days; switch a source to the other one in the editor. Empty canvas starts from a schedule alone.

Templates arrive filled in for your workspace and start switched off:

  • Accounts: every ad account of the platform where most of the spend is. When your accounts report in different currencies, the template reads the accounts in the currency of the one that spends the most, since one money line cannot hold for two currencies. The dialog names the accounts left out.
  • Lines: set from your accounts' last 30 days rather than one number for everyone. A stop-loss asks to pause at about twice your cost per result, a CPA alert fires at about half again your CPA, a high spend alert at about twice your biggest campaign's usual day, and the weekly ad reports judge the ads that spent about one result's worth. Each template says the line it took and why. Without enough figures, the template keeps its standard line.
  • Channel: the Slack, Discord or Telegram channel you already connected. Without one, the send step connects one in place.
  • Time zone: yours, from your browser.

A template that counts results, such as a stop-loss or a CPA alert, says so when your accounts count no result yet: it would stay silent until a campaign optimizes for a conversion.

What each template needs

Every template lists the data it needs from your workspace, checked against what Adrails already synced, before you create it:

NeedRead fromWhen it is missing
An ad account on the platformYour connected accountsThe template cannot run. Connect Meta or Google Ads in place.
Spend in the period it readsThe last syncThe report is empty, or nothing crosses a threshold.
A counted resultThe ad sets' optimization eventA CPA, results or stop-loss line never passes. Set the event in Ads Manager or as a Google Ads conversion.
Purchase valueThe conversion's valueA ROAS ranking or filter is empty. Send the value with the purchase event.
A channelYour Slack, Discord, Telegram or email channelsThe run cannot post. Connect one from the template.
Shopify, Klaviyo, Google DriveYour integrationsThe source has nothing to read; an agent answers without it.
A Drive folder and an ad set with an adChosen in the editorLaunch new creatives has nothing to launch, or nowhere to put it.
Credits, a paid planYour planWith no credit left, no report, message, launch or agent step runs; pauses and budget cuts still apply. On Preview, agent, launch and ad change steps need a plan.

What is there is listed quietly; what is missing says what it does to the run and how to fix it. Templates ready to run come first, in every filter. One that would have nothing to say yet stays visible and can still be created: the editor lists the same items.

With no ad account yet, connect Meta or Google Ads from Automations itself. The sign-in opens in a window and brings you back to the same page.

The alerts (CPA alert, High spend today, Low ROAS, Low click-through rate, High frequency) post to your chat when the workspace has one, and otherwise email what they caught: the editor asks who receives it. High frequency reads Meta's own frequency; Google Ads reports none.

Start from a campaign or from Analysis

A campaign's or an ad set's Automations tab, and an ad's Performance tab, have New automation: the catalogue opens on the templates that can read that campaign, ad set or ad alone, and the one you pick is created for it, named after it. Schedule report in Analysis opens the reporting templates on the accounts on screen.

Every template

Each template needs a channel to post to and, unless it reads only the store, an ad account. Available says what else it needs: a paid plan for an agent, a launch or a change to your ads, and a connected store or app for what it reads there.

TemplateGroupAvailable
Stop-lossProtect the budgetA paid plan
Spend without resultsProtect the budgetAny plan
CPA alertProtect the budgetAny plan
High spend todayProtect the budgetAny plan
Low ROASProtect the budgetAny plan
Account billing and healthProtect the budgetAny plan
Scale winnersScale what worksA paid plan
Weekly reviewReportingA paid plan
Weekly top adsReportingAny plan
Weekly worst adsReportingAny plan
Daily summaryReportingAny plan
Monthly reportReportingA paid plan
Weekly client reportReportingAny plan
Launch new creatives from DriveCreativesA paid plan and Google Drive
Creative fatigueCreativesA paid plan
Low click-through rateCreativesAny plan
High frequencyCreativesAny plan, Meta only
Yesterday's winnersCreativesAny plan
Best sellersStore and CRMA Shopify store, no ad account
Ads against storeStore and CRMA paid plan and a Shopify store
Email revenueStore and CRMKlaviyo, no ad account
Pause ads of sold-out productsStore and CRMA paid plan and a Shopify store
Low stock on products with spendStore and CRMA Shopify store
Product page healthStore and CRMA Shopify store
Store ROAS guardStore and CRMA paid plan and a Shopify store
Suggest a campaign for new productsStore and CRMA Shopify store, no ad account
Notify me of new leadsStore and CRMA Meta account that can read leads

Build a workflow

Triggers

  • On a schedule: every day, every 2 or 3 days, every week on a chosen day, or every month on a chosen day, at a chosen hour and time zone. Day 31 means the last day of the month.
  • Account health: every hour, from what the last sync found, when an ad account has used 90% of its spending limit or reached it, a payment failed, or the account is restricted, disabled or under review. Each problem is said once while it lasts, and nothing is changed on the account. Spending limits and payments are read on Meta.
  • New lead: every 5 minutes, how many new leads a Meta campaign's, ad set's or ad's instant forms collected, or one instant form's, since the last check: one message per check, however many came, with a link to them on the Leads tab and to the Google Sheet they also go to. Only leads that came after the automation was switched on count. A lead's name, email, phone and answers are never read, kept or posted: Adrails asks Meta for lead ids and times alone. It needs the Meta permission to read leads, as the Leads tab does.
  • Ad threshold: when a campaign, ad set or ad metric reaches or falls to a value, over today, the last 7 days or the last 30 days. It is checked every hour and fires at most once a day for each item that crosses the line. An alert that only posts counts once it went out, so one that failed comes back at the next check. When more items cross the line than the steps below can act on in one run, the others wait for the next check.

A threshold or an ad account step reads all your ad accounts, chosen accounts, chosen campaigns, or, when it reads ad sets or ads, chosen ad sets or chosen ads. A chosen ad set keeps the ads inside it.

Steps

StepWhat it does
Meta Ads, Google AdsRead campaigns, ad sets or ads over yesterday, the last 2, 3, 7, 14 or 30 days, or last month.
IntegrationsRead a connected source, such as your store.
FilterKeep rows where a metric is at least or at most a value.
RankingKeep the top or bottom 1 to 25 rows by a metric.
Adrails AgentAsk the agent, or a specialist, about the rows, with your instruction. It can only read.
Write in a chatAn agent writes its answer in a new chat under Automations and notifies you. Choose the Ad accounts it reads, and Send a notification by email to add an email with a link.
Send to Slack, Send to Discord, Send to TelegramPost to a connected channel or chat, with Approve and Reject for the changes waiting for approval. See Slack, Discord and Telegram.
Send by emailEmail up to 10 addresses, with the report attached as a PDF if you want it.
Save to KnowledgeSave the report above it, an agent's answer or what moved, as a document in Knowledge, in the folder you choose. See Save to Knowledge.
Pause ads, Resume adsPause what is left, or resume what is paused, up to 1 to 50 items per run.
Change budgetChange the daily budget of what is left, up to 1 to 50 budgets per run.

A workflow has at most 40 steps.

Act on ads: automatic or reviewed

  • Ask for approval: each change waits in Priorities on Home for 24 hours, with the figures that led to it, until someone approves it there, or with Approve in the Slack, Discord or Telegram channel a send step posts to.
  • Apply automatically: each change is applied at once, after Adrails checks on the platform that nothing changed since it was read.
  • Resume automatically resumes at once only what that same automation paused, such as an ad it paused when its product sold out. Anything else paused, by a person or another automation, waits for approval like Ask for approval.

Launch ads creates its new ads paused, in both modes: an automation never switches on what it launched. Its card says Launched paused: switch them on from Campaigns, and a person does, from Campaigns.

Every change is logged with the figures that led to it, such as ROAS 3.42x (at least 3.00x), Oct 5 to Oct 11, and can be undone. Changing ads on Meta needs the ads_management permission.

Change budget

SettingWhat it does
ChangeMultiply by a factor (0.01 to 100, so x2 or x10), raise or lower by a share, raise or lower by an amount, set to an amount, or reallocate to the winners.
HighestThe daily budget it never raises past. Required before it raises budgets automatically.
LowestThe daily budget it never lowers past.
Then waitNo wait, 24 hours, 3 days or 7 days before the same budget is changed again.
  • Where the budget is: an ad has no budget of its own. Its ad set's daily budget changes, or its campaign's when the campaign holds the budget. On Google Ads, the campaign's. When several ads of one ad set reach the step, the budget changes once.
  • Daily budgets only: a lifetime budget is left as it is, and the run says so.
  • Your bounds, not the workspace's share: the step goes as far as you set, past the workspace's limit on one budget change. The workspace's highest daily budget still holds.

Reallocate to the winners

Reallocate to the winners moves budget between what you selected instead of adding any. The rows that reach the step, for example the ads at a ROAS of 3 or more, are the winners; the other selected rows are not. Their budgets, on their ad sets or campaigns, are added up, then:

  • the winners share the share you set, 80% by default, in equal parts;
  • the others share what is left in equal parts;
  • the total stays the same. With four ad sets at 25 and two winners, the winners go to 40 and the others to 10.

When every selected budget wins, they share the whole equally; when none does, nothing moves. Lowest keeps a budget from dropping under it and Highest keeps a winner from passing it, the difference going to the other side. A budget that cannot move now, a lifetime budget or one changed too recently, stays out of the total. Lowered budgets go first, and the whole reallocation fits the step's limit per run or none of it moves.

The Scale winners template watches ad sets whose ROAS over the last 7 days reaches 3 on real spend, and asks to double their daily budget.

Send by email

The email says chooses what each run emails. What the step above caught lists the campaigns, ad sets or ads that reached the step, with their figures, as a chat message does: the alerts send this. The period's report sends the client report over the run's period.

Tick Attach the report as a PDF and each run attaches a new client report over the run's period. The Weekly client report template ticks it for you. Tick Send nothing when no row is left to skip empty emails.

When you set a value on a threshold or a filter of Meta Ads or Google Ads figures, the step shows your accounts' recent figure beside it, such as your CPA over the last 30 days, and a line to take in one click.

Save to Knowledge

Save to Knowledge keeps each run's report as a document in Knowledge, under Documents, named after the automation and the day. Put it under an Adrails Agent step, a what moved source or a send step, and choose its Folder, or All documents for the top of Documents. A folder deleted since saves at the top. The step replaces the agent's former Keep each report in Knowledge switch: an automation saved with that switch shows the step right under its agent or source, in the folder it named, and works as before.

  • Order with a send. A send step below Save to Knowledge posts the same report with a link to the document. A send above it links to the chat, as before.
  • Preview run shows Would save on the step, as a send shows Would post, and saves nothing. After a real run, the step says Saved.
  • A run whose document could not be saved says so; its report stays in its chat. See Documents.

Try it, then switch it on

A new automation opens on Ready to turn on: the data it needs, step by step, with what is missing and its fix, and each step on the canvas marks what it lacks. It can be turned on as long as nothing blocks the run; data that only leaves it with less to say is said, not enforced. The panel also gives one sentence on what it will do and when, for example Every Monday at 09:00 (Europe/Paris), Adrails reads the last 7 days of ads on Meta Ads and posts the top 10 ads by ROAS that spent at least €20 to #ads-reports, its credit estimate and its first run. Anything still missing is listed under it, each line opening the step to fix. Review and turn on, beside the name, opens it again at any time.

  • Preview run runs the workflow on real data without posting, changing or saving anything. The panel opens on the steps and the run plays down the canvas, step by step, each one showing what it found: the rows, the message, the ads it would act on and each budget it would change, from and to, and the document it would save. A step that fails is named on the status line.
  • Run now runs it for real, once, as a scheduled run would: what already fired today, or once, is left out. When it changes ads by itself, it first lists the changes a preview read and asks you to confirm.
  • Turn on schedules it, saving your changes first if there are any. The confirmation says when it runs next, and the automations list shows the next run of each one that is on. The person who turns it on becomes responsible for it; if they leave the workspace, it keeps running as the workspace owner, who is told. Its failures and reports reach the workspace's owners and administrators too.

Runs shows the last 10 runs and what each step did.

The list

Add filter narrows the list by status, trigger, ad platform, ad account, where it sends, last run, credits, readiness and agent step. Columns chooses what shows (steps, next run, last run, credits, readiness) and keeps your choice.

Credits

Each automation run that produces a report, a message or an action costs at least 1 credit, more when the AI writes or analyses; a check that finds nothing is free. A threshold checks your accounts every hour for free, and costs a credit only when it posts or acts.

  • A report or a message: 1 credit a run. A report of what moved is written by Adrails from the synced figures, 1 credit per report; a sent message, an email or a report filed in a chat count the same.
  • A change on an account: 1 credit a run that applies or proposes a pause, a resume or a budget change, however many it makes.
  • Launch ads: 1 credit a run that launches, the AI that writes the copy included.
  • Agent: 1 credit for each started $0.10 of its AI, at least 1; most runs use 1, a long analysis 2.

A run is counted once: 1 credit when no AI worked in it, what its AI cost otherwise, never 1 on top of the AI. A retry of the same run is never counted twice. Each automation's AI steps use at most 100 credits a month; past that, they wait for the next month and its other steps keep running. Each run is listed in the usage log in Settings > Plan as Automation run.

With no credits left, a report, a message, a launch or an agent step does not run: its step says No credits left, and the person who switched the automation on gets one email a month. A pause or a budget cut an automation applies still applies with no credit left, while a resume or a budget raise waits for a credit: a safety rule never stops protecting because credits ran out. It is counted when a credit remains, and free otherwise.

The editor and the automations list show the estimate next to the schedule, for example About 1 credit each day, 30 a month. It is an estimate: a heavy run can use more. See The agent and credits.

Watch for

  • A role needs the Automations permission to change automations. Owners, Administrators and Members have it; a custom role has it if you gave it.
  • Only the owner and Administrators can choose Pause automatically, Resume automatically, Apply automatically or Launch automatically, and only they can save, run or turn on an automation that has such a step. A Member, or a custom role, builds automations that ask for approval; the automatic choice shows why it is unavailable.
  • A person limited to some ad accounts sees and changes the automations on those accounts only. An automation that reads every account stays hidden from them.
  • A workspace has up to 25 automations, and up to 5 threshold automations switched on at once.
  • If the person who switched an automation on no longer has the permissions it needs, it switches itself off.
  • A workspace's scheduled runs go one after another, so reports set for the same hour can start a few minutes apart.
  • After 2 scheduled failures in a row, that person gets an email. After 5, the automation switches itself off.
  • With no credits left, an AI step, a report or a message waits rather than fails, and so does an AI step past its automation's 100 credits. The automation stays on and runs again at its next time once there are credits, and this counts toward neither limit above. With no credits left, that person gets one email a month.
  • On Preview, a workspace keeps one automation, without agent steps or ad changes.
  • A channel is tested when you connect it: Adrails posts a short message there. Email recipients are never sent a test: use Preview run to see the email.

On this page