If your invoice reminders are automated and payments still come in late, check upstream first. Was a deposit taken. Were terms confirmed. Is there a real consequence for paying late. Fix that conversation, then let automation handle the follow-up.
@georgeclem This only works if the closer actually reads the setter's notes before the call. A lot of teams still do the discovery, then the notes sit in a CRM field nobody opens, and the closer starts from zero anyway.
@georgeclem Careful with this on local trade ads though. A roofer or plumber's page covered in 'you spelled that wrong' comments just looks bad. The CPM drop might be real, but people hire trades on trust, not on who has the most comments.
Telling an agent to "be helpful" and "sound more human" tells it nothing. Break each vague word into something you can actually check: what it says when it doesn't know, when it hands off to a person, how long an answer runs, what it never says.
Most of our own leads come in through Instagram DMs right now, barely any through email or a website form. Trying to work out if that's just our audience or true for coaches generally. Where do most of your inbound messages actually come from?
@georgeclem The part that breaks this is the account manager already has other new clients that same week, so the welcome video slips a day and nobody notices till month 3. Make that first hour a trigger that fires the second the contract is signed, not a task on someone's list.
Most of our own leads come in through Instagram DMs right now, barely any through email or a website form. Trying to work out if that's just our audience or true for coaches generally. Where do most of your inbound messages actually come from?
@georgeclem Same thing happens with support. You don't see which questions repeat until you've logged enough of them. Once you do, it's always the same handful, and that's the first thing worth fixing.
@jordan_ross_8F The tell is the same 5 or 6 tasks eating the week, every week. Founders don't see it until they log their own hours by task for a week. Once it's written down, most of that list doesn't actually need the founder, it just always ends up on the founder's plate.
A task log said a batch of work was still undone, so we redid it from scratch. The tool's own history showed that batch was actually finished days earlier, just never logged. When the sheet and the tool disagree, check the tool first. The sheet is the one likely to be stale.
@georgeclem Same thing happens with support. You don't see which questions repeat until you've logged enough of them. Once you do, it's always the same handful, and that's the first thing worth fixing.
@jordan_ross_8F The part that breaks this fast is the block reason. If it pulls from a status field nobody updates, it just repeats a stale flag every morning and people stop trusting it. Pull from something that gets touched naturally instead, like the last comment timestamp.
Right now our agents hand off to a real person the second someone shows buying interest. We haven't tested letting it try to close instead. If yours could close small deals on its own, would you want that, or do you want a person closing every time?
@jordan_ross_8F Most "build AI ops" pushes try to fix everything at once and stall out. Pick the one question or task eating the most founder time each week, document exactly how you'd answer it, and get just that piece running first. Do that a few times and the ops load actually goes down.
Owners agree replying first usually wins the job, then still take a day to answer. It's not laziness, every message needs fresh thinking before you can reply to it. Write out the five or six questions you get most and keep the answers ready. Replying becomes copy, adjust, send.
@georgeclem Most of what escalates to you is the same handful of questions on repeat. Document those once so answering doesn't have to be you every time. That fixes the bottleneck way faster than another hire does.
Training an agent on old content works well once there are a few hundred messages to learn from. We don't know how it holds up for someone just starting out with almost nothing to feed it. If your content library is thin, would you still trust an agent trained on it?
@markdmei This tracks. Ads don't fix a shaky offer or messy fulfillment, they just pay to expose the cracks faster. Free work and cold email force you to nail delivery before you're spending to find out it's broken. Once that's solid, ads are gas on a fire that's already lit.
Most coaches are already paying for tools that could answer their DMs, they just never got set up right. I train mine on your own content instead of a generic script. Running it now for Brock Johnson's InstaClubHub. DM me if you want to see it.
@GlennNieuwenh Support tickets and DMs beat reviews for this too. Nobody's performing for a star rating there, they're just annoyed or relieved, so the language comes out rawer and usually tests better in the ad.