Home / Industries / Marketing
Industry · Marketing

AI agents for marketing automation.

Marketing is the one vertical where Mirai360 AI is also the customer: the lead sourcing, enrichment, drafting, and attribution described on this page is the exact system we run on our own outbound. Every message it produces still leaves by a person's hand — the code that writes the draft has no way to send one.

How every deployment is framed

The discipline here matches every other Mirai360 AI vertical: name the workflow, define what the agent does, define what a person approves before anything reaches a real inbox or a real phone, and say plainly what would get measured once a wave actually runs. What is different in marketing is how that boundary is enforced. Elsewhere it is a rule a person follows. Here, the draft-generation code contains no SMTP client, no Gmail API, and no other send-capable library at all — there is no code path from "draft written" to "message sent." A person has to be the one who sends it.

The four capabilities

Lead sourcing

Workflow
Building a targeted outreach list for one industry and area, without buying a list from a broker.
What the agent does
Runs Google Places text searches by keyword and area, paginates each query, and deduplicates every result — first by phone number, then by normalized company name — so the same business never appears twice under two different spellings.
What the human approves
A person reviews and qualifies the raw list against the target buyer profile before any lead moves on to enrichment.
Measured outcome
How many unique leads a given keyword-and-area combination actually returns, so a person can see list coverage before deciding whether to expand the search.

Lead enrichment

Workflow
Turning a raw lead into someone with a known industry and a way to reach them.
What the agent does
Classifies each lead into an industry vertical from a keyword match against its name, category, and website, then looks for a contact email on the lead's own website — and its /contact or /about page if one is easily found — checking robots.txt and pausing between requests before every fetch, so one unreachable or blocking site only skips that lead and never stops the run.
What the human approves
A person reviews the classified, enriched list — vertical assignment and any discovered email — before a single lead enters drafting.
Measured outcome
How many leads resolved to a known vertical and how many returned a usable email, so a person can judge whether the list is enriched enough to draft against, before any draft exists.

Industry-specific draft generation

Workflow
Turning an enriched lead into a message worth a person's time to send.
What the agent does
Writes one drafted message per lead, from a single template matched to that lead's vertical, in English, and saves it as a file. Nothing is queued, scheduled, or transmitted.
What the human approves
Everything, and not by choice — by construction. The code has no send-capable import of any kind, so there is no path from a written draft to a sent message. A person reads, edits, and sends each one by hand.
Measured outcome
Nothing yet. Each draft sits in a file with an index mapping it to its lead and its touch_id, until a person sends it — at which point it leaves this system's view entirely.

Closed-loop attribution

Workflow
Knowing which outbound touch, if any, led to a site visit or a booked call.
What the agent does
Mints one touch_id per drafted message and threads it through the link in that draft. If a recipient clicks through, the site's own tracking script reads the touch_id from the URL, caches it for that session on a first-touch-wins basis — kept independent of the separate UTM layer already used for other campaigns — and attaches it to every event fired afterward, including a booked call. This part is live in production today.
What the human approves
Nothing here needs approval; it only tags events the site already generates. A person still decides what happens after a booked call.
Measured outcome
Once a wave actually goes out, this is the layer that would connect a specific touch to a specific outcome. No wave has gone out yet, so there is nothing to report — the mechanism is live; the results are not.

What end-to-end automation creates

Run together, these four capabilities form one pipeline: a raw list becomes a classified, contact-checked list, becomes a set of vertical-specific drafts, becomes — once a person sends them — a set of touch_ids that can be followed from a clicked link to a site visit to, if it happens, a booked call. No outbound wave has run yet. There is no reply rate, no lead count converted, and no booked-call figure to report here; this page describes the pipeline, not a result. Mirai360 AI runs this same pipeline on its own outreach, which is why this is the one industry page on this site written from direct operating experience rather than a stated intent.

One 30-minute call.

We look at your pipeline, find the one queue worth starting with, and tell you honestly whether it clears the bar.

Book your free 30-minute call