INBOX INSIGHTS: Local AI Migration Part 1, Vibecoding 101 Part 5 (2026-09-02)

INBOX INSIGHTS: Local AI Migration Part 1, Vibecoding 101 Part 5 (2026-09-02) :: View in browser

👉 Take our new course, AI for Writers!

Watch this video on YouTube

Local AI Migration: Part 1

In last week’s newsletter, I talked about the bar we’ve set for ourselves with productivity. That productivity comes at the (figurative) hands of generative AI. While it’s great that we’re getting more done, there is an environmental cost. During that issue, I pledged to document the process, and here we are.

This is the part of the process people skip over, the thinking. Granted, I might tend to “overthink” things, but you get my point.

Step 1:

I started at the literal beginning. I created a place to house all the thinking, analyzing, and supporting documents.

Everything about this effort now lives in one place on my machine and in my LLM. Every session I run picks up where the last one left off because the context is in the project rather than in my head or scattered across six conversations. I recognize the irony of using an LLM to do this work, but this is where my challenge lies. I do my best thinking and planning when I’m talking to myself (my Co-CEO skill). I do that work before I build anything. Something I’m trying to figure out is what that looks like in a different context. You might be asking, “How did you get along before LLMs?” Slowly. Very slowly. My thinking is still the same, but no one was around to ask deeper questions, present different data, or challenge my ideas. This is what is helping my efficiency. I no longer just “sit” on something. I can pull it apart and reassemble it and come to a conclusion much faster. That is one of the larger benefits I’ve seen with LLMs. But I digress. Step one was to get organized, and once that was done I could keep moving.

Step 2:

Before any tools, any research, any opinions, I wrote a user story. A user story, borrowed from software development, is a single sentence with three parts: As a [persona], I [want to], so [that].

Here is where my user story started:

As the CEO of Trust Insights and an environmentally conscious citizen, I want to create and execute a plan to move from cloud-hosted AI to local AI without losing productivity, so that I can reduce my environmental impact and feel good about my AI use.

This will anchor the entire project. Everything I do, every decision I make, needs to trace back to this user story. If it doesn’t, it’s a distraction. A project like this could easily get really large and ambitious. A user story will help me prevent that.

Step 3:

Now that I have folders and a user story, I can start building requirements. If you read Chris’s Data Diaries, you’ve seen the scaffolding: a product requirements document, then a technical specification, then a work plan. The why, the what, the how, in that order. You don’t build anything until all three exist.

When Chris talks about this process, he writes about this for code. I’m not writing code, but the logic holds for anything you’re about to spend money and time on: a mistake on paper is cheaper than a mistake in production. When I encourage you to do your requirements before building anything, this is why. You’ll see through my process just how vital requirements are.

I took his PRD prompt more or less as written, swapped in my user story, and ran it.

Step 4:

I got to play twenty questions with my LLM. This was (and still is) the hardest part of the process so far. It’s the deep thinking. It’s the decisions. It’s where assumptions are challenged, and revelations are made. This is where the real work happens.

The PRD prompt asks the AI to interview you before it drafts anything. What I got back was twenty questions, and most of them I couldn’t answer immediately. A sample of the ones that stopped me:

What does failure look like? Name the condition where you would stop and go back to how you work today.
Which single workflow, if it got noticeably worse, would end this project?
Rank these against each other: output quality, response speed, breadth of capability, privacy, cost, carbon. Which do you give up first?
Does the manufacturing carbon of new hardware count against your goal?
How much ongoing maintenance are you personally willing to own?

I’ve been working through them one at a time for the past week. Some questions are easier than others. Some questions I have to keep coming back to in different sessions. The ranking question took me two days, and I got it wrong the first time.

A week later, I am still working through the questions. I’m close to done with them, but you know me; I want to be thorough, so I’ll run through them again. This is the part of the process that will save you time and money when you start building.

I don’t have a deadline for this project and stated that at the beginning. Despite that declaration, the tool kept finding new ways to ask. What about a target date? How about milestones as a substitute for a target date? Each time I dodged the question.

For what it’s worth, without a hard deadline or milestones, I need to hold myself accountable for moving this project forward. That’s partly why I wanted to document it here, with you.

Where things stand today

I went into this assuming a binary: Cloud AI, bad. Local AI, good. Move everything, feel better.

That’s not what came out.

What came out is that this doesn’t have to be all or nothing, and that treating it as all or nothing was the actual mistake. Some of my work has a real opportunity to run locally. Some of it clearly shouldn’t, and the reasons are practical rather than ideological. A third category turned out not to need AI at all, which I did not see coming. None of that came from research. It came from being asked what I actually wanted, specifically enough that I had to decide. I’ll say it again, this is why you want to take your time with planning.

There’s a version of the 5P Framework by Trust Insights™ lesson buried in here. I had a Platform answer before I had a Purpose I could measure. Platform is fourth in that framework for a reason, and I skipped straight to it because buying something feels like progress. It happens to the best of us, and it is a good reminder to slow down and think things through. If I had started to move everything to a local environment, I would be doing it wrong (for me).

I’m one week into this. I haven’t installed anything, bought anything, or moved a single workflow. I haven’t even written the PRD yet. What I have is a user story and twenty (mostly) answered questions. I should be able to start that in the next day or so.

That’s not slow. That’s deliberate. It’s pragmatic. It’s what I try to convey to all of you every time I talk about requirements first.

If you’re planning something similar, try this first. Write your user story in one sentence. Then ask yourself the failure question: what would make you stop? If you can’t answer it, you’re not ready to start.

What are you planning right now that you haven’t written a user story for?

Reply to this email or join the conversation in our Free Slack community, Analytics for Marketers!

– Katie Robbert, CEO

Weekly One Question Poll 🐶

Click or tap just one answer below — there’s no form, this is the entire thing. One tap and you’re done, with our thanks.

Does your company have a separate budget line for AI in 2027?

Do you have a colleague or friend who needs this newsletter? Send them this link to help them get their own copy:

https://www.trustinsights.ai/newsletter

In this episode of In-Ear Insights, the Trust Insights podcast, Katie and Chris discuss how you will overcome creative blocks while navigating information overload. You’ll discover why AI keeps offering generic advice and how you can bypass those predictable suggestions. You’ll learn to pull unconventional strategies from unrelated fields and apply them to your marketing plans. You’ll identify two distinct thinking patterns that unlock fresh content ideas when your usual topics feel stale.

Watch/listen to this episode of In-Ear Insights here »

Last time on So What? The Marketing Analytics and Insights Livestream, we looked at setting up Claude Marketplaces. Catch the episode replay here!

This week on So What? we’re examining how different AI models write. Are you following our YouTube channel? If not, click/tap here to follow us!

Here’s some of our content from recent days that you might have missed. If you read something and enjoy it, please share it with a friend or colleague!

Take your skills to the next level with our premium courses.

Get skilled up with an assortment of our free, on-demand classes.

You could be reaching 37,000+ marketers, analysts, data scientists, and executives directly with your ad. Want to learn more? Reach out and contact us.

In this week’s Data Diaries, the last of a five part series on “vibe coding” basics for non-coders. Katie and I have talked extensively about vibe coding in the past on the podcast, but we’ve never really dug into the basics of it. So in this series, completely unsurprisingly modeled on the 5P Framework by Trust Insights™, I’ll cover those basics.

Measuring Vibe Coding Performance

This week, we close out our series on vibe coding with performance. In the 5P Framework by Trust Insights™, performance is the bookend of purpose – we start with why we’re doing something, and we end with whether we did the thing or not. In software development, there are some clear performance measures we can use to measure whether we did the thing or not.

One of the reasons AI has doubled down so hard on coding as a primary use case is precisely because software development has very, very short feedback loops. When you deploy a marketing campaign, it might be 30-90 days before you know whether the campaign worked or not. When you deploy code, you know in seconds whether the code is correct or not, and that short feedback loop means AI can quickly adjust.

There are four categories of performance measures we use in coding:

Unit tests. A unit test is a test of an individual file or chunk of code. Does it do what it says it’s supposed to do according to the plan? If you have a function like this – function(add){ 2 + 2 }, it had better return 4. Anything else, and the unit test should fail – the code didn’t do what it’s supposed to do.

Integration tests. Sometimes called end to end or E2E tests, these tests measure how well individual pieces of code interact with each other and hand off results from one to the next. In software, a piece of code can do exactly what’s intended and not hand off its results properly, breaking things down the line. E2E testing measures how well our code functions overall as a coherent unit.

Smoke tests. Even the best integration testing doesn’t measure the production as a whole. A smoke test – from hardware engineering – tests the entire system exactly as a user would use it. Amusingly, a smoke test comes from when an engineer would plug in a new piece of hardware and see if it starts smoking.

User acceptance testing. Once software exits development, it doesn’t go straight into production. The users that we named in the People section have to bless it by using it in a controlled environment to make sure it does what they expect it to do. If it doesn’t, it goes back to the lab for re-work.

Where most folks go wrong in vibe coding is one of two places: first, AI can only do tests 1 and 2. They can’t effectively do a smoke test because you the human has to watch and see that the software actually turns on the way it’s supposed to. And they can’t do user acceptance testing at ALL because they’re not the user.

Second, AI trained on the public internet, and that means it’s adopted human-level standards for testing. It will often write that the best practice for tests is 80% passing, which is what the average human software engineer and project manager will agree to because you hit diminishing returns after a while if you manage your developers to 100% passing tests and it frustrates them, burns them out.

AI has no such limitations. It’s why we declare in our process and especially in our coding standards from last week that 100% test coverage and 100% passing tests is mandatory – never any less, because AI has no emotions and can’t get frustrated – and if we’re going to replace human effort with machine effort, then we should hold machines to our ideal standards and not our human ones.

This wraps up our vibe coding introductory series. As you can tell from the past 5 issues, there’s a lot more we could explore, and we likely will – stay tuned to the Trust Insights Academy in the months and years to come!

Almost every AI course is the same, conceptually. They show you how to prompt, how to set things up – the cooking equivalents of how to use a blender or how to cook a dish. These are foundation skills, and while they’re good and important, you know what’s missing from all of them? How to run a restaurant successfully. That’s the big miss. We’re so focused on the how that we completely lose sight of the why and the what.

This is why our capstone course, the AI-Ready Strategist, is different. It’s not a collection of prompting techniques or a set of recipes; it’s about why we do things with AI. AI strategy has nothing to do with prompting or the shiny object of the day — it has everything to do with extracting value from AI and avoiding preventable disasters. This course is for everyone in a decision-making capacity because it answers the questions almost every AI hype artist ignores: Why are you even considering AI in the first place? What will you do with it? If your AI strategy is the equivalent of obsessing over blenders while your steakhouse goes out of business, this is the course to get you back on course.

👉 Take this course now to become an AI leader

Here’s a roundup of who’s hiring, based on positions shared in the Analytics for Marketers Slack group and other communities.

Are you a member of our free Slack group, Analytics for Marketers? Join 4000+ like-minded marketers who care about data and measuring their success. Membership is free – join today. Members also receive sneak peeks of upcoming data, credible third-party studies we find and like, and much more. Join today!

Where can you find Trust Insights face-to-face?

Spotlight, September 2026
LPA, September 2026
MEMA, October 2026
MAICON, October 2026
SMPS AEC.AI, November 2026
MarketingProfs B2B Forum, November 2026

Going to a conference we should know about? Reach out!

Want some private training at your company? Ask us!

First and most obvious – if you want to talk to us about something specific, especially something we can help with, hit up our contact form.

Where do you spend your time online? Chances are, we’re there too, and would enjoy sharing with you. Here’s where we are – see you there?

Read our disclosures statement for more details, but we’re also compensated by our partners if you buy something through us.

Imagine a world where your marketing strategies are supercharged by the most cutting-edge technology available – Generative AI. Generative AI has the potential to save you incredible amounts of time and money, and you have the opportunity to be at the forefront. Get up to speed on using generative AI in your business in a thoughtful way with our workshop offering, Generative AI for Marketers.

Workshops: Offer the Generative AI for Marketers half and full day workshops at your company. These hands-on sessions are packed with exercises, resources and practical tips that you can implement immediately.

👉 Click/tap here to book a workshop

Some events and partners have purchased sponsorships in this newsletter and as a result, Trust Insights receives financial compensation for promoting them. Read our full disclosures statement on our website .

Thanks for subscribing and supporting us. Let us know if you want to see something different or have any feedback for us!

Need help with your marketing AI and analytics?

Leave a Reply Cancel reply

Pin It on Pinterest

Watch This Newsletter