Self-learning ads with guardrails

Ads that learn from real leads, inside a monthly cap you set, with every change shown before it is made.

HumonixACME Field ServicesSample data
Approvals: nothing goes out without a yes. Try Approve or Skip; every decision goes on the ledger. Sample data · ACME Field Services, an invented company.

The problem

Ad accounts drift: budgets go to what gets clicks, not what brings customers, and nobody has time to test.

  1. 01

    What starts it

    Every 6 hours per company, plus a Monday report.

  2. 02Agent

    Hue reads performance (a CSV export or your connection) and the leads and sales that came from it.

  3. 03Agent

    It judges each experiment and spots ads that are wearing out.

  4. 04Agent

    It proposes changes with the reason, and drafts new variants to test.

  5. 05Agent

    Approved changes are queued for a person to set in the ad platform; Hue does not change your ad account.

  6. 06You

    What a person approves

    An owner approves each proposal. An automatic mode is the owner’s choice, only inside the monthly cap and daily limits, with a stop-loss per lead and a kill switch.

Tools it uses

The tool catalog →

What you see in the app

Built in

For the curious: the files that do this, in our codebase. lib/ads-learn.js, lib/ads-learn-core.js, lib/ad-attribution.js, docs/design/app/ADS-LEARNING-LOOP.md

Accurate to what runs today. Where a step waits for Humonix’s review, it says so; nothing on this page is a customer result.

Work email and website. We read your site (robots.txt first), train Hue on it and run a free audit with three quick wins. No card.