The way I’m thinking about Cyclops / my ads agent is pretty simple: Intentwise i
By Brett Bohannon · May 19, 2026 · Curated by George's Blog
The way I’m thinking about Cyclops / my ads agent is pretty simple:
Intentwise is the data layer. Helm is the logic and harness layer.....
Cyclops is the agent that turns the signal into something usable.
That distinction matters.
Pulling ad data is not enough.
The agent needs to understand what the data means, how to route the question, what checks to run, and how to package the answer in a way an operator can actually use.
That is where the harness comes in.
For Cyclops, the reporting structure is being built around a few core pieces:
➔ Signal: What changed in the account?
➔ Diagnosis: Why might it have changed?
➔ Context: Is this a campaign issue, keyword issue, budget issue, placement issue, or conversion issue?
➔ Action: What should the operator consider doing next?
➔ Watchlist: What needs to be monitored after the action?
➔ Playbook: What workflow should this trigger?
That is the layer I think is missing when people talk about “chatting with your data.”
The data source matters.
Intentwise gives strong ad signals.
But the structure around the signal matters just as much.
Without logic, the agent guesses.
Without a harness, the agent drifts.
Without a playbook, the output becomes another report someone still has to interpret.
Helm is being built to solve that middle layer.
The layer between raw data and actual execution.
Cyclops is the ads version of that idea.
Intentwise provides the signal.
Helm applies the logic, guardrails, and workflow.
Cyclops packages the output into a brief and playbook so the operator can move faster.
That is where I think this gets powerful.
Not dashboards vs agents.
Data + logic + harness + playbooks.
That is the stack.