Built for agencies that have to defend their numbers

The problem we started from

An agency running TikTok Shop for several brands ends up operating five or six tools that do not know about each other: a spreadsheet of creators, an inbox, a shipping tab, a reporting deck, and somebody remembering which sample went where. The work that scales badly is not the strategy — it is the chasing, the joining up, and the reconstruction of what happened when a client asks.

Aegis exists to hold all of that in one place, and then to do the repetitive part of it. Not to replace the person who decides what to do; to remove the six hours a week they spend finding out what happened.

The rule the product is built around

A system that fails loudly is worth more than one that appears to work. Almost every decision in Aegis follows from that. A metric nobody measures shows as unmeasured rather than as zero. A forecast without enough history to support an interval is not produced. An AI answer that cannot be traced back to your own rows is refused with the reason instead of being approximated into something plausible.

It makes for a less impressive demo. It also means that when Aegis shows you a number, you can put it in front of a client without checking it first — which is the only thing that actually matters.

What we will not do

We do not scrape platforms that have not authorised it, bypass rate limits or bot detection, solve CAPTCHAs, or automate accounts nobody gave us permission to touch. Several things customers have asked for sit on the other side of that line, and they stay there. Where a platform does not expose something through its API, we say so and build the manual workflow properly instead of pretending.

We also do not use customer data to train models, and every AI call records that it was made with training opt-out in force — a fact in a log rather than a sentence in a policy.

Where the product is today

Aegis is new. We report status on four levels — implemented, tested, verified against a live third party, and proven in production — and we keep them separate rather than letting a passing test suite imply something has been through a real deployment. If you ask where a particular capability sits, you will get the accurate answer, not the flattering one.