The problem this solves
The day ends with an open question: did anything happen today that I need to know about before tomorrow? Answering it honestly means opening the ad accounts, the analytics, and the search data - three logins from a phone, at night, to usually confirm that nothing remarkable occurred. So most owners skip the check most nights, and the practice decays into checking only when something already feels wrong, which means finding problems at day three of their existence rather than day one.
The nightly check also has a depth problem. Even when you do look, a dashboard shows you that CPA rose without saying why. Was it one campaign or all of them, an auction shift or a landing-page issue, a blip or the start of a trend? The "why" requires the kind of cross-account digging nobody does from a phone at 9pm, so the question gets carried overnight as low-grade worry and re-investigated from scratch in the morning.
The digest this replaces both halves with is deliberately tiny and two-way. Five lines with today's numbers, arriving in the chat app already on your phone, answers the "did anything happen" question in fifteen seconds. And because an agent wrote those lines from the source data, the follow-up "why" gets a real diagnosis in the same chat, tonight, instead of becoming tomorrow's first anxiety.
How the mission runs
- "eod" is the whole ritual. Whenever your day ends - 6pm or midnight - you message "eod" to your Telegram bot. The trigger being yours rather than a fixed schedule means the digest covers your actual day, and skipping a night costs nothing. The agent begins assembling the moment the message lands.
- The day is read across three systems. Google Ads and Meta Ads yield today's spend, conversions, and cost efficiency against your recent baseline. Search Console adds ranking movement worth knowing about. The agent compares today to the trailing picture as it goes, because a number without its baseline is trivia rather than information.
- Five lines come back, and only five. Spend, leads, cost per acquisition, ranking moves, and one line of anything unusual - each with an at-a-glance direction against baseline. The discipline is the feature: a digest that stays five lines gets read every night, and anything needing more depth waits behind your follow-up question.
- "Why is CPA up?" gets an actual diagnosis. A follow-up question sends the agent digging: decomposing the movement across Google Ads and Meta Ads, isolating which campaigns drove it, and distinguishing one-day noise from the start of a trend. The answer comes back in the same chat as a diagnosis with the evidence underneath it, phone-readable.
- The chat accumulates the story of your accounts. Thread memory persists night to night, so "is this the same thing as Tuesday" resolves correctly, and the agent flags when tonight's anomaly matches a pattern it has reported before. Digests and follow-ups draw on the same workspace meter as everything else, with deep digs subject to your spend threshold.
The prompt
This is the exact objective the agent receives. Swap the obvious placeholders for your own domain, segment or channel and run it as-is from the console, Slack, or the API.
What comes back
A nightly fifteen-second read on the business, delivered into Telegram on your command: spend, leads, CPA, and ranking movement in five baseline-aware lines. Behind it, an on-demand diagnostic layer - any "why" question gets decomposed and answered in the same chat, tonight. Over weeks, a continuous narrative of your accounts that makes anomalies recognizable as either noise or the second data point of a trend.
Make it yours
- Add a morning counterpart: "sod" returns yesterday's late movement and anything that changed overnight, so the day starts with the same fifteen-second contract.
- Set exception rules: if CPA or spend crosses a bound you define, the agent messages you proactively at the threshold instead of waiting for tonight's "eod".
- Route a copy: after you read the digest, reply "share" and a clean version posts where your team can see it, with your follow-up thread staying private.
Frequently asked questions
Why five lines instead of a proper report?
Because the nightly question is binary: does anything need my attention before tomorrow? Five lines answer it at a glance and survive being read every single night, which a longer report would not. Depth is one reply away when a line raises a question, so brevity costs nothing.
How deep can the "why" follow-ups actually go?
As deep as the connected data supports: campaign-level and ad-set-level decomposition across Google Ads and Meta Ads, timing analysis, and comparison against any period in the account history the sources retain. A dig projected past your spend threshold asks in the chat first, like any gated mission.
Can the agent fix what it diagnoses, like pausing the offending campaign?
Diagnosis and action are separate gates. The digest and follow-ups are read-only against your accounts. If a diagnosis convinces you to act, saying so in the chat composes that change as its own mission with explicit approval, so a late-night reading session can never accidentally modify a live campaign.