# AI agents

Six agents, each independent, all running on MicksonBrain(TM) with a person in the loop where it counts. Every one of them can be demoed on your own account: https://micksongrowth.com/agents#demo

## Speed-to-lead agent: a reply to every lead in under a minute
Anchor: https://micksongrowth.com/agents#speed-to-lead
For: Home services and private equity
Every form fill, ad lead and missed call gets a reply inside a minute, written to your CRM with its click ID still attached. It works the conversation toward a booked appointment and hands anything off-script to a person with the whole thread already in the CRM.
Flow: lead arrives (Form, ad lead or missed call, with the click ID attached); written to the crm (Straight in. No middleware holding the record hostage); replies in under a minute (And works the conversation toward a booked time); hands off when it should (Emergencies and anything off-script go to a person).
How a client uses it, This website: Fill in any form on this site and you get a personal reply from Drew within a minute, with his calendar link in it, and the booking step opens inline on the thank-you page rather than in an email you have to go and find. That is the agent, running on us before we point it at anyone else.
SMS: opt-in captured at the point of inquiry, messaging inside quiet hours set per market, one-word opt-out that stops it permanently and is written back.

## GTM agent: warm ICP accounts, verified contacts, drafted outreach
Anchor: https://micksongrowth.com/agents#gtm
For: B2B tech and GTM engineering
It watches for accounts warming up on your ads and your site, filters them against your ICP, and suppresses current customers and open deals before anyone is contacted. It finds the buying committee, verifies the contact details, and drafts against the real signal. A person approves every send.
Flow: spot the signal (Accounts warming up on your ads and your website); filter and suppress (Against your ICP, your customer list and your open deals); find the committee (The right people, with contact details verified); draft, then a person sends (Written against the signal, approved by a human).
How a client uses it, An AI data platform, Series A: One ICP definition drives the whole engine: warm account detection, enrichment, the outbound sequence and the write-back, instead of four tools each with its own idea of who the customer is. The daily digest reports what was sent and what replied, so the question "is outbound working" has an answer before the quarter ends.

## Audience agent: LinkedIn audiences that stay in sync with your ICP
Anchor: https://micksongrowth.com/agents#audience
For: B2B tech and private equity
It turns a signal list into a company audience, enriches each account, and keeps the audience in step with the list instead of leaving a one-time upload to go stale. It also reports the other direction: which ICP accounts your ads have actually reached, and which ones they have not.
Flow: read the signals (Ad engagement, site visits and the ICP list in your CRM); enrich the accounts (So a company name becomes something you can target); keep linkedin in step (Additions and removals, on a schedule, not by hand); report the coverage (ICP accounts reached, and the ones still not reached).
How a client uses it, An AI data platform, Series A: Every company their LinkedIn ads reached is pulled weekly and matched against the ICP lists their CRM holds, so the report answers the question the board actually asks: of the named accounts we care about, how many have we reached this quarter, and which ones have we not touched at all. The "not reached" list is the one the team works from.

## Negative keyword agent: wasted search spend caught every night
Anchor: https://micksongrowth.com/agents#negative-keywords
For: Every paid search account we run
It reads the search terms across every account every night, scores them against a taxonomy built per trade, and proposes the negatives rather than quietly applying them. Competitor terms get their own treatment, because blocking a competitor name is a decision rather than a cleanup.
Flow: read every search term (Nightly, across every account, not at the monthly meeting); score against the trade (Wrong trade, wrong intent, wrong job, each handled differently); propose, do not apply (The list goes to a person before anything reaches the account); learn from corrections (A rejected suggestion changes what it proposes next time).
How a client uses it, The home services portfolio: It runs nightly across every home services account we manage and holds its suggestions for approval, with a separate pass for competitor brand terms. Corrections are recorded, so a negative a human rejects once does not come back the following week wearing the same hat.

## Budget agent: overspend flagged the same day
Anchor: https://micksongrowth.com/agents#budget
For: Private equity and multi-brand operators
It takes a daily spend snapshot for every account and channel, compares it against the month’s plan, and forecasts where the month lands at the current pace. When an account is pacing to blow through the number, it raises an emergency alert the same day rather than at the month-end reconciliation.
Flow: snapshot every day (Spend per account and per channel, PPC, LSA, Meta and Bing); pace against the plan (Where the month lands if nothing changes); alert before the overspend (Same day, to a person, with the account named); report the landing (Planned against actual, for every brand, every month).
How a client uses it, A PE-backed home services platform: Pacing runs across every brand in the portfolio at once, so an operating partner sees one list rather than chasing each general manager for a number. Budget emergency alerts fire the day an account starts pacing hot, which is the only time the information is still worth anything.

## Website agent: landing page tests judged on booked revenue
Anchor: https://micksongrowth.com/agents#website
For: Home services and B2B tech
It splits traffic between page versions, tracks each visitor through the funnel, and reports the result per version with spend attributed by visit share. The decision rule is written before the test starts, and the number it is judged on is bookings and revenue rather than clicks.
Flow: split the traffic (Two versions of the page, an even split, one loader); follow the funnel (Every step from first view through to a booked job); score it honestly (Probability to win, with spend split by visit share); call it on the rule (The decision rule is set before the test, not after).
How a client uses it, Top Tier Plumbing: Two versions of their quote funnel run against each other on our own testing host, with a per-version funnel report in their nightly dashboard: unique visitors, start rate, probability to win, cost per lead and the booked revenue behind each version. The test is called on the rule that was written down before it started.

## Demo
https://micksongrowth.com/agents#demo