Behind every fast-growing CPG brand is a back office held together by manual fixes and caffeine.
Systems that don’t talk, data that gets lost, and someone (usually the founder) who ends up being the human API.
Marcos Brisson, co-founder and CEO of Kaizntree, has a fix: agentic AI that automates the tedious tasks, keeps operations airtight, and gives founders back hours in their day.
00:16 - Death by a thousand edge cases
01:41 - Start with one painful workflow
02:43 - Keep a human-in-the-loop
03:54 - The 15-hour rule
05:21 - When AI ops starts paying off
06:14 - Why Zapier isn’t aren’t enough for CPG
07:32 - What CPG ops looks like in 5 years
Text answers below have been edited for length and clarity. Full answers are in video. (We highly recommend watching!)
It’s always an edge case that creeps up and eats people’s time.
We start by talking with operations team to intimately understand the tech stack they’re using, how the data is flowing, and what they’re currently doing to fill in the gaps.
Because there’s always an integration between, for example, your SPS commerce and your NetSuite account. But at the same time, some of that data has to go into Google Sheets or maybe there’s an edge case that doesn’t go through the integration that’s being done manually. And that edge case is taking someone 20 hours a week. it’s always the edge cases that that sort of creep up and eat people’s time.
Someone’s job is meant to be one thing. And then this workflow, this edge case, has increasingly become what their job is. And they hate it. They want to do more strategy, supplier negotiation, or whatever it is. But instead, they’re spending more time moving data across platforms because it has to be done. That’s a big part of what we’re coming in and automating.
Start with one painful workflow.
For any kind of AI automation, start with the simplest thing that still saves time. Say you want to process your distributor reports. Start with one distributor and one kind of report. It typically takes a week to make sure everything’s flowing smoothly from the email into your Google Sheets, for example.
From there, you can scale to capture all your distributor reports. Then layer more levels of complexity and more workflows on top of that. Like now let’s do the POs, the bill of lading, the shipment updates... there’s always a ton of things that need to be automated.
The big difference with us is that we’re actually validating all of the data that’s being processed. So we do have a human in the loop.
It’s not worth automating anything that takes less than 15 hours a month.
As a rule of thumb, I’d say it’s not worth automating anything that takes less than 15 hours a month. The sweet spot is anything over 20 hours, that’s what you really what you want to be looking at. It only makes sense when the time savings are significant.
The value of AI in operations is really connecting the dots. Kintsugi is a Japanese art of repairing broken pottery with lacquered glue. We kind of view ourselves as the glue that fills the gaps between the tools so they’re fully connected.
Some tasks might be really annoying, but aren’t worth automating.
It’s a monthly expense. There’s the cost of development when the systems are set up, then the monthly maintenance and credit consumption expense, which is tied to the complexity of the workflow.
So for it to make sense, the time savings have to be lined up. Some tasks are not worth automating. They might be really annoying, but only happen once a month and take three hours to do it.
The brands that we’re seeing adopt AI automation are doing over 10 million a year. Simply because beforehand, the priorities are not in workflow optimization but more in revenue growth. Once you hit a certain stage that we see brands start to care about process optimization.
One of the case studies on our website is for a brand called Actual Veggies. We automate all of their PO processing, making sure whatever they’re shipping out matches what’s been ordered. We’re saving them over 30 hours a month.
Not only does that time go toward higher value work, we’ve eliminated human error. Even if there’s only one mistake per a thousand purchase orders, it can still create issues. There’s a value in knowing that that’s completely eliminated.
We’re automating the parts they don’t enjoy.
One misconception is AI is replacing people’s work. What I’ve seen across every single brand is it’s not replacing people’s work, but it’s actually unlocking them to do what they enjoy most in their job. We’re automating the parts they don’t enjoy.
Another one is brands think Zapier or another no-code AI tool is ready to automate CPG workflows. It’s actually very risky because there’s no human in the loop. The onus is still on you to check there’s no mistakes in the data moved across different platforms.
There’s a lot of potential for things to go wrong. Whether it’s us or anyone else that you can find, you want to be setting up these automations with someone who knows what they’re doing.
Give it a go over a short, defined period of time and see what’s the difference.
I would definitely recommend to go out there and try using ChatGPT. It’s not an AI agent, so you won’t be able to connect your different systems. But you can already create your own GPT where you feed it some data, ask any questions, and it can generate reports for you.
The brands we work with always start with one simple workflow and their experiment. Okay if we automate this thing, first of all, does it work? And then second, how valuable is it?
To anyone who’s interested, just give it a go over a short, defined period of time and then compare that against someone doing the work and see what’s the difference. The real unlock for operations comes when you can plug in different AI tools across your operations stack.
