How Ektie's AI Ad Manager handles Google Ads optimisation

Blaze, Ektie’s AI Ad Manager, runs on the same 60–180 minute heartbeat cycle as the rest of the agent team. Every cycle, it reviews active Google Ads campaign performance against the previous period, identifies patterns worth acting on, queues optimisation actions it can execute autonomously, and flags decisions that require human judgment for the Marketing Director (Ava) to review. The human’s job is strategy and approval — the routine optimisation work runs without prompting.

What does Blaze review on each heartbeat cycle?

Performance against benchmarks: click-through rate by ad group, conversion rate by campaign, cost-per-conversion trend, impression share. Search Terms report: new queries that triggered ads in the last cycle — flagging any that appear irrelevant and queuing them as proposed negative keywords for approval. Budget pacing: is each campaign on track to hit its daily budget without exhausting early (which causes ads to stop before the end of the day in high-conversion afternoon hours)? Quality Score changes on high-spend keywords. The review is systematic and comprehensive — every campaign, every cycle.

What does Blaze do autonomously vs what does it escalate?

Autonomous actions (no human approval needed): adding negative keywords from the Search Terms report that clearly match patterns already in the exclusion list, adjusting ad rotation to favour higher-performing ad variations, pausing ad groups with zero conversions over a significant spend threshold. Escalated for human review: bid strategy changes, budget increases or reductions above a defined threshold, changes to audience targeting, new campaign creation, pausing a previously high-performing campaign that has recently declined. The boundary is clear: Blaze acts on the routine and surfaces the consequential.

How does Blaze coordinate with the rest of the agent team?

The Ad Manager doesn’t operate in isolation. When Blaze identifies that outbound (Morgan’s sequences) is generating high interest in a specific message angle — for example, a compliance-first framing that’s generating strong reply rates in healthcare — it shares that signal with the Marketing Manager (Mira), who updates the ad creative brief to test the same angle in paid. When the Marketing Manager publishes a new piece of content, Blaze picks it up for promotion in the next cycle. The three-tier Team Brain is the shared layer — what one agent learns, all agents can access.

What does Blaze escalate to the human founder?

Budget decisions above predefined thresholds. Significant performance changes that require a strategic decision — a campaign that’s consistently outperforming vs one that’s declining despite optimisation. New channel opportunities: if LinkedIn Ads data suggests a new audience segment worth testing, Blaze surfaces the recommendation with the supporting data. And any situation where the available actions are ambiguous — Blaze writes a requires-review note with context, options, and a recommendation, and waits for the human’s decision before acting. The human stays in the strategic seat.