SAMSON MARKS

Plenty of founders post AI content every day and get no clients from it, because it looks like AI made it. We make our carousels with AI every day. A new one lands in Slack at 9am, in our brand, using our own photos, and it doesn't look like a template. This is the setup: 1. Find what good looks like Save three or four carousels you'd be proud to post, each in a different style. 2. Tell AI to copy the style in your brand Your fonts and colours, with variation inside each style so every post isn't the same layout. 3. Point it at your own photos Give it the folders for founder photos, client logos and testimonials, and say when to use each one. 4. Have it send every draft to Slack Reply with changes the way you would to a designer, and it makes them. 5. Feed it what you've already said One YouTube video gave us five carousels in about 15 minutes. Bad: asking AI for "a carousel about lead generation" and posting whatever comes back. Good: AI working from your examples, your photos and your words, with a person approving every post. The cheapest version of this setup costs about £33 a month. Save three or four carousels you rate into one folder today.

Every new client should be a good day in an agency. For a lot of founders it's the start of a week of chasing. Somebody has to send the onboarding form, book the kickoff, ask for logins twice and tell the team the client is live. Usually that somebody is you. So I wrote up exactly how we set up onboarding so the form does the chasing: 1. Write down every step from "client paid" to "live" 2. Give each step one owner and one short SOP 3. Hand the client the steps they can do themselves, like connecting their domain 4. Make the onboarding form create the tasks, assign them and send the SOP 5. Let AI draft the steps that need no judgment 6. Get a Slack ping when something is ready to approve Bad: a Slack thread where you remind three people what to do for a new client. Good: the client submits the form, the tasks appear with owners and SOPs, and the first thing you do is approve a draft. If you're solo, the owner on most steps is you or AI, and the same system still does the remembering for you. Full breakdown here: teapaw.org/samson_ai/articles

Agency owners are paying for five dashboards and still asking their team for the numbers every Monday. The data is all there. It's spread across the ad account, the CRM, the calendar and the finance tool, so nobody sees the whole picture without building a spreadsheet first. What I'd build instead is one AI layer that sits on top of all of it. 1. Pull the data into one place Ads, leads, booked calls, closes and cash collected. When it lives in five tools, nobody trusts any of them. 2. Put AI on top that can see everything So you can ask a question in plain English instead of opening a dashboard. 3. Ask it the questions you'd normally ask your team "Which ad brought in the calls that closed last month?" "Why did booked calls drop this week?" 4. Act on the answer The next decision gets made the same day, not after someone builds a report. Bad: logging into four tools on Monday morning to work out how last week went. Good: asking one question and getting the answer with the numbers behind it. This works the same if it's just you and a CRM or a team of 20 with a dozen tools. Start with the two tools you check most and connect those first.

My first ever client came from a theme page I ran when I was 15. I'd already failed at dropshipping at 14. Then I started building theme pages on Instagram, TikTok and YouTube Shorts, and got a few of them to between 10k and 30k followers. I wasn't trying to get clients. I was learning editing, what goes viral and why people watch. Then a top UK Amazon FBA founder needed someone to run his social media, and I was the person who obviously knew how. By 16 the agency was doing about £3k a month. The lesson still holds for agencies now: 1. Show the skill in public The work people can see does the selling before any call. 2. Post about the thing you sell I sold social media management, and my proof was my own pages. 3. Let the content bring the buyer to you He found me. I didn't pitch him. The difference now is that AI handles the editing and the repurposing, so the only part you have to do is have the ideas and show up. What's the one skill you sell that you've never shown in public?

You can hire a full-time social media manager for about $50 a month. This is what that looks like in our business: - AI turns what I already said on video into posts - Paper designs them, about $16 a month - Buffer schedules them, $5 a month The whole content stack comes to about 33 quid a month at most. The part AI can't do is the ideas and the taste. That's still my job. So record the raw idea once, then let the stack turn it into a week of posts. What are you paying for content right now?

You've probably already said your next 30 posts out loud. They're sitting in your client calls. Every week you explain the same things on sales calls, onboarding calls and client check-ins. That's your best content, and it's already recorded. So this is how we turn it into posts without coming up with new ideas: 1. Record everything you already do Sales calls, client calls and any long-form video. You're not having to re-come up with ideas. 2. Give the transcripts to AI and ask for the best bits The questions clients keep asking, the explanations that landed, the objections you answered well. 3. Turn each one into a post in your voice Give it your best posts as the reference so it doesn't come out as AI slop. 4. You pick what goes out Your job is the idea and the final call. AI handles the repurposing. Bad: staring at a blank page on Sunday night trying to think of something to post. Good: a list of 30 ideas pulled from calls you had this week. Start with last week's calls and see how many posts are already in there.

If your ads keep pulling in owners who can't afford your bigger package, check which buyer that channel actually reaches before you rewrite another ad. Bad: One offer, pushed through every channel, hoping the right buyer turns up. Good: 1. Write down who your best clients are 2. Check where each of them came from 3. Bigger brands: go after them through referrals and partnerships 4. Owners who want growth: that's who paid social finds 5. Shape the offer around the buyer that channel brings you Same agency, different routes in, and an offer that fits each one. Where did your best client this year come from?

If I ran an agency doing under $50k a month, these are the 6 jobs I'd hand to AI first. Every one of them runs inside Systemised right now, so this is what we actually use. 1. Onboarding tasks The onboarding form creates the tasks on the Asana board, assigns an owner, notifies them and attaches the SOP. Nobody has to nudge people along. 2. The client's first website Their onboarding answers go into Lovable and come back as a first build, so the team edits a draft instead of starting from a blank page. 3. Sales call QA After every sales call an agent reviews it and drops a report in Slack, so you know what to fix before the next one. 4. Ad hooks from testimonials Feed AI your client testimonials and pull the hooks straight out of the words buyers already use. 5. Client dashboards Live dashboards pulled from GHL through a private integration, so clients see their numbers without anyone building a report. 6. Repurposing One long-form recording becomes the carousel, the newsletter and the posts. What I'd keep for people is the offer, the ideas and the final call on what goes out. Start with the job that eats the most of your week. Which one would you hand over first?

A new client rocks up on day two asking where things are at. The expectation setting has already failed by then. As I put it on YouTube: have an FAQ page that they enter the moment they pay. Build it from your onboarding call transcripts. Put the month-by-month timeline on it. Put what your best clients do on it. Then the kickoff stays on their business. Full breakdown (FAQ page + the rest of the expectation system): teapaw.org/samson_ai/status/21176…

Agencies selling AI services lose a lot of clients right before the results show up. From what we see with our clients, month one gets the first results, month two builds on them, and month three is where our best clients see their best results. So the client who cancels in month two never sees the month that matters. This is how I'd set up an AI offer so they stay: 1. Sell an outcome they already pay for Booked calls, faster delivery or hours saved. Something they can check without understanding the tech. 2. Show them the timeline on day one Month one, month two, month three, written down, so a quiet week doesn't feel like failure. 3. Report on one number every week If they can't see it moving, they'll assume it isn't. 4. Deliver it the same way every time One packaged outcome. A custom build for every client means chasing integrations all week. Sell the outcome, show the timeline, and report the number every week. If you want us to look over how you package and sell your AI offer, book a strategy session here: calendly.com/d/dvpq-7zz-x8w/…

Agency owners are paying for AI tools and then doing the work again themselves, because the first draft wasn't right. The tool usually isn't the problem. Nobody ever told it what good looks like, and nobody gives it feedback. So this is how we run AI on real client work: 1. Write down what good looks like Your best ad, your best landing page, a report a client loved. If it's well documented, AI can work to it. 2. Give it one job Ad scripts for a new client, say. One job, one set of rules. 3. Have it tell you in Slack when it's done Our setup drafts the work and sends a Slack message so we can approve it. Nobody has to go and check. 4. Give feedback in Slack like you would a team member Say what's off and why. Then update the rules, so it won't need to come back to you for the same revision. Bad: rewriting the draft yourself and moving on. Good: telling it what was wrong, updating the rules, and getting a better draft next time. You keep the taste and the final yes. AI does the drafting. Pick one job this week and write down what good looks like for it.

The client who cancels in week six never saw month three. Month one was doing what month one is meant to do. They just didn't know that, because nobody showed them the shape of the work. As I said in a recent video: month one gets you the initial results, month two gets you further, and "month three is where most people actually explode and our best clients see their best results." So put that timeline in front of them on day one: Month 1: funnel live, first ads, first lead form submissions Month 2: the system starts compounding Month 3: where the best results land Agree one number before any work starts, write down where it is today, and keep that same number in front of them every week (including the flat weeks). Book the end-of-month-one check-in yourself, before they ask. I broke the full expectation system down here (timeline, FAQ page, what your best clients do, weekly updates, end-of-month-one check-in): teapaw.org/samson_ai/status/21176…

An objection handling doc full of one-line answers is useless to a new closer. Bad: "Too expensive: focus on value." Nobody can say that out loud on a call. Good: the exact words from your own call recordings. - The objection, in the prospect's words - What your side said next, word for word - What the prospect said back - Whether that call closed or was lost Feed AI every recorded call, the ones you closed and the ones you lost. Tell it to quote the transcripts and not summarise them, because it will try to tidy the quotes into advice. Then a new closer reads real conversations with the turning point marked, before they touch a lead. The full step-by-step is in yesterday's article: teapaw.org/samson_ai/status/21158… Which objection comes up most on your calls right now?

If I had to get an agency's delivery off the founder in 30 days, this is exactly what I'd do: Week 1: Map every step Write down every step between "client paid" and "campaign live". Every one, including the five-minute jobs you do without thinking. Week 2: Put an SOP and an owner next to each step Any time the owner can be AI, make it AI. That's usually account setup, research and first drafts. The low-leverage tasks go first. Week 3: Make the onboarding form the trigger When the form comes in, the account gets created, the folders get made and the client gets invited without anyone touching it. Nobody should be spending time on work that needs no judgment. Week 4: Move yourself to checking only You stop building and only review what goes to clients. Once the team's work matches yours, you step back from that too. Bad: "I'll hand delivery over once things calm down." Good: "Every step has an SOP and a named owner, and AI owns the ones that don't need a person." That's how our client delivery went from a couple of weeks to a couple of hours. Map the process this week, and put a name next to every step.

Every new closer you hire learns the job on leads you paid for. They get a few role-plays and a script, then spend their first month working out your objections live on booked calls. And unless you sit in on every call, you have no idea what's being said on them. So at Systemised we don't rely on sitting in. We have a sales QA agent connected to Fireflies. At the end of every single sales call, we get a report in Slack. George and I read it, and if something on the call needs fixing, we come back on it the same day. Any AI you can connect to your call recorder can do this (Fathom works too), and you don't need to be technical to set it up. This is what the report covers: 1. Who the prospect was, and whether they were actually qualified If unqualified people keep reaching the closer, that's a marketing or pre-call problem, and coaching the closer won't fix it. 2. The outcome Closed, follow-up booked or lost. One line each, so you can scan a week of calls in a minute. 3. Every objection, and how it was handled Checked against your objection map, which is the doc with the exact words that closed each objection on past calls. If the closer agreed to a follow-up the moment someone said "I need to talk to my business partner", you'll see it. 4. Anything the closer promised This is the line to watch. A closer under pressure can say yes to results, timelines or extras that delivery can't do. That turns into a refund or a churned client three months later, and the report catches it the same day. 5. Where the call followed the call map, and where it drifted Which questions were asked, in what order, and which ones got skipped. 6. One thing to do differently on the next call Just one, so the closer actually acts on it. Keep the whole report short. If it takes longer than a couple of minutes to read, nobody will be reading it by week two. Two things make it work: - A person reads every report. AI writes it, and you or your head of sales judge it. - Once a week, give AI all the reports and ask what's changed: objections that aren't in the training doc yet, lines that closed deals, where closers keep drifting. A person approves the changes, so the doc gets better with every call you run. If your marketing and pre-call process are on point, the closer is really there to push things over the line. The QA report is how you know they're doing that without sitting on the call yourself. I wrote up the full system, from pulling every recorded call (closed and lost) to the training doc a new closer studies before they touch a lead: teapaw.org/samson_ai/status/21158…

Before you automate a step in your client delivery, check whether the client can just do it themselves. Connecting their domain to GHL is the classic example. Your team chases the client for login details, waits on them, then does a job the client could have done in their own time. Send a short SOP with screenshots during onboarding and let them do it. The client won't really care if it's a couple of steps, as long as you give them the actual process to follow. Automating a step still means building it, maintaining it and fixing it when an integration breaks. So take out the steps that don't need your team first. This is the order I'd go through for every step in delivery: 1. Hand it to the client if they can do it from a short SOP 2. Anything that needs your team gets an SOP and one owner 3. Where that owner can be AI, make it AI, with a person checking the output What's one step your team still does for clients that they could do themselves?