Search campaigns your agent can read and draft against

Search is the channel where structure is the strategy. The account tree is wider than any other platform here — keywords and assets sit under ad groups, budgets are shared objects rather than fields, and the same campaign can be three different products depending on its channel type. An agent is only as useful on Search as the model underneath it is honest about that.

A connection is a grant against the Google Ads API, scoped to the accounts you choose. Tempio syncs the tree — campaigns, ad groups, ads, keywords and the assets under them — and the daily numbers behind each, and lands them in the same normalised shape every other platform does: integer minor units with an explicit currency, base counts rather than stored ratios, one canonical id per entity for a change request to point at.

The reads an agent gets from that are the same four regardless of platform, which is the point. Blended KPIs with a per-channel breakdown for a date range. Per-campaign rows highest spend first, each staying in its own currency, with a marker saying whether the order is a ranking or an unconverted list. The structure under any campaign or ad group — the ad groups, ads, keywords and assets — each row carrying the operations that level can actually express. And the account brief: ninety days of context per account, including how concentrated spend is, which campaigns rank on return, which rank worst on cost per acquisition, and which weekdays are cheapest, so the agent has a baseline before it calls a number good or bad.

Ratios are read-time arithmetic over base counts, and a zero denominator yields null rather than zero. On Search that matters more than elsewhere, because a brand campaign and a prospecting campaign in the same account routinely differ by far more than the account average on CPA and neither figure means anything without the other beside it.

Google addresses more levels than any other platform connected here: the campaign, the ad group beneath it, the ad beneath that, and — depending on the campaign's type — keywords and assets as first-class rows of their own. Two facts about that tree shape every change an agent can propose.

The first is where money lives. A budget on Google is a campaign-level object, not a field on an ad group, so an ad group has no budget to change and a request to change one is refused before a reviewer ever sees it. What an ad group carries instead is a bid. The connection reports this: it names the levels a budget attaches to, and on Google that list has exactly one entry.

The second is what an objective is. Google has no objective concept at all in the sense the other platforms use the word — what a campaign carries is a channel type, and the read model maps that onto the objective field so one shape can hold every platform. An agent reading an objective on a Google campaign is reading a translation, and the tool layer says so rather than inventing a taxonomy Google does not have.

Two more differences are worth knowing before anything is drafted. Google supports a dry run, and Tempio uses it: a draft can be validated against the live account before it is queued and again on the way to being applied, because an account can change underneath a draft that has been waiting. And Google's removal is terminal. Removing an entity is not an archive and cannot be undone — not even to re-pause it — which is why a removal is never cleared automatically and why a reversible status change is almost always the better instruction.

What Google withholds is the current day. It restates today's figures for hours, so the tool layer treats data as ready through yesterday and refetches a trailing month on every sync. A page that quotes a number therefore quotes the day it was read, and an agent asked about today says what it can see rather than filling the gap.

A change is a change request before it is a change. One call writes an entire batch in the order it applies, with each entry naming the earlier entry that creates its parent, so a campaign, its ad group and the ad inside it arrive as one reviewable group. The result is the batch, one review link for the whole group, and the authority that cleared it: waiting for a person, or already cleared by a rule the workspace wrote. Every row carries its position in the run and, once something has cleared it, whether that was a person, a key or a policy.

The validation step is where Google's dry run earns its place. A draft is checked against the platform's declared capabilities at the moment it is written, and validated against the account again on the way to being applied, so the common failure — a draft written on Monday against structure that moved on Tuesday — is caught at the gate rather than halfway through a batch. A draft that does fail keeps its payload and can have it replaced whole, which re-validates it and puts the row back to waiting; that is how a stopped batch is resumed rather than rewritten.

Approving is a separate permission from drafting. A key minted for an agent can read the account and fill the review queue without holding the permission that applies anything, and the two are never the same grant.

Google reviews ads and it reviews them per policy, not per account, so an edit to a headline or a final URL can send a live ad back into review on a schedule that belongs to Google. A change can be applied, verified as landed, and still not be serving. The record keeps those apart: what was changed and when, and what the platform then did with it.

Pacing is the other half. A campaign budget is a shared object with a daily amount, and delivery paces against it from the moment an edit lands rather than retroactively across the day already spent. A change made at four in the afternoon is a different instruction from the same change made at nine in the morning, which is why every draft records the state before it and the moment it was applied. When someone asks a fortnight later why cost per click moved, the answer is on the row rather than in somebody's memory.

Applying a change to a live Google Ads account is in private beta, behind the developer token that account-level access requires. What is built and in use is everything before it: the read model, the media plan, the guardrails, and the gate every change waits at — approved by you, or by a rule you wrote.