Activity
Mon
Wed
Fri
Sun
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
What is this?
Less
More
19 contributions to AI Automation Society
Everyone in here can build an agent now. So what do clients actually pay for?
Meta shipped Muse for Small Business with Shopify and HighLevel connections. OpenAI shipped a managed Agents API. Salesforce is packaging agents by role. If your pitch is "we know how to wire the tools together," that just became a feature. Not in two years. Now. I run an agency and I've been sitting with this. The honest answer is that the integration layer was never the moat, it was just the part clients couldn't do themselves yet. What survived, at least for us: 1. Knowing the client's operation well enough to decide how the agent should behave. Same booking agent, two businesses: one has slots tomorrow, one is booked out three weeks. Same first message, different right answer. That judgment isn't in the box. 2. Follow-up logic that can see what already happened. Did they call? Book elsewhere? Waiting on a quote? Sending another message is easy. Knowing whether to send it is the work. 3. Ownership after launch. Hours change, services change, people leave. Someone has to keep the system matched to the business. That's recurring revenue if you position it right. 4. Being honest about what you don't control. An appointment isn't revenue. The agencies that promise revenue are the ones that get churned. Full breakdown here, including how I'd reprice: https://youtu.be/RWEdQkTFD1s Real question for this group: what's in your offer right now that a platform could ship as a feature in the next 6 months? And what's left after that?
1 like • Oct 19
@Tracy N. thanks tracy!
I almost sold a client something that would’ve made them less money...
Had a weird moment on a sales call last month. Owner wanted us to automate their proposal process. Takes them about 3 hours each. Easy win on paper, cut it to 30 minutes, done. Then I asked what they'd do with the 2.5 hours. Silence. That's when it clicked for me. We've been selling AI automation on "hours saved" and most of the time the hours just disappear. More meetings. More inbox. Or the competitor down the street gets the same tool and now everyone's charging less for the same work. The savings are real. They just don't stay with you unless you decide where they go first. I recorded a full breakdown of this because I think it's the thing most people get wrong before they spend a dollar on AI. The $2,000 to $500 math, where the money actually leaks, and the 5 questions I now ask before we build anything. https://youtu.be/9vu_2ylzQkY Curious though. If AI gave you back 10 hours a week tomorrow, what would you actually do with them? Be honest.
I charge $1,500 for this AI audit. Here’s the whole thing, free.
Nick and I just ran our AI audit live, in person, with a room of business owners. Same process we use before we build anything for a client. We recorded the whole thing. Why it matters: a property management client came to us sure they needed AI for tenant messages. The audit showed the real leak was leads. They were taking 3 to 4 hours to answer and about 30% went cold. We built lead response first. Replies now go out in 18 seconds. If they had skipped the audit, they would have automated the wrong thing. In the video you get: 1. How to pick the ONE task worth automating (it is not the one that takes the most hours) 2. The 3 prompts we run in Claude, in order 3. How to catch Claude when it invents a finding 4. How to turn the audit into a small test you can measure The slides and the workbook with every prompt are linked in the description. Make a copy and follow along with your own business. Or run it on a client and charge for it. Watch here: https://www.youtube.com/watch?v=d46GsClU2FI Run it on one task this week and drop your finding in the comments. I will reply to every one and tell you what I would build first.
2
0
🚨 I Gave OpenAI Dots $100 — Then Added Jev
I wanted to see if these new AI agents could do more than just build cool demos. So I gave OpenAI Dots a $100 budget and told it to start a real business. It chose a $25 website audit, used Codex to build it, and then I added Jev to handle the fast routing decisions inside the workflow. The interesting part wasn’t the build. It was seeing whether the system could actually go from: idea → product → automation → real customer → real money We ended up getting our first genuine $25 customer. This is probably one of my favorite experiments so far because it shows where AI agents are getting insanely good — and where running an actual business is still hard. Would love to know what you guys think I should make the Dot build next 👇 🎥 Watch here: https://www.youtube.com/watch?v=Qa9O7NMYL9Y
Own the Boring Intelligence > Rent the Exceptional Intelligence
Every time your business uses AI, you’re basically renting compute from someone else. For the hard, weird, high-value stuff? That makes sense. But for the same repetitive workflows running thousands or millions of times? Maybe not forever. That’s the idea behind my newest video. The future of AI probably isn’t all-cloud or all-local. It’s hybrid: Own the boring intelligence. Rent the exceptional intelligence. Watch here: https://youtu.be/lKQmIxmde3M?si=abHmgaC_Q_0ixCTU Curious if you agree — where do you think the cutoff is between what a business should own vs rent?
1-10 of 19
Keshav Suki
5
349 points to level up
@keshav-sukirya-2735
CTO @ Reprise AI.

Active 9d ago
Joined Apr 27, 2026
Powered by