INBOX INSIGHTS: Local AI Migration Part 5, Practical Responsible AI Part 4 (2026-09-30)

INBOX INSIGHTS: Local AI Migration Part 5, Practical Responsible AI Part 4 (2026-09-30) :: View in browser

👉 Take our new course, Introduction to Deep Research with AI!

Watch this video on YouTube

Local AI Migration: Part 5

This weekend I was on the couch with football on, reviewing the technical plan for this project on my phone. Somewhere in the second quarter of the game, it occurred to me that I was doing the exact thing I wouldn’t be able to do with this project.

Working from my phone, from anywhere, with an AI that remembers what I decided last week, is a level of ease I’ve gotten very used to. Moving some of my work onto my laptop, or off AI entirely, means using less. It also means giving some of that ease back.

In my experience, every update to these tools makes it a little easier to stay connected and a little easier to reach for them. (Except when they keep changing where some of the buttons are; that’s a different grievance.) There’s always an app on the phone, always a chat window open, always something within reach. That’s a slippery slope, and it’s easy to lose sight of why I started this project while I’m sliding down it. (Especially when the game is no longer interesting.)

The tasks I’m starting with aren’t thinking tasks. Converting a document from one format to another is clerical work. But who’s to say they’ll stay that way? The same convenience that lets me convert a file from the couch also lets me hand off a judgment call from the couch, and I’m not sure I’d notice the day that starts to happen.

Back to the point, while reviewing the plan, I realized it stated that I would borrow a smart plug. The purpose of using a smart plug is to check how much power each job uses. Well, I don’t own a smart plug, and even if I borrowed one, opening an app to check and report on usage is one extra step I’m not prepared to do.

That was the moment I realized I had been reviewing a plan written for a person far more disciplined than me. It was thorough and correct, and I wasn’t going to follow it. I am fairly disciplined, but I can procrastinate with the best of them.

I write requirements for a living. I tell clients all the time to design for how people actually behave, not how we hope they’ll behave. I was one approval away from signing off on a plan that assumed a version of me who checks dashboards, remembers extra steps, and has more free mornings than I do. I mean, I do those things for other people, but not for myself.

The question isn’t whether it’s a good plan. The question is whether I’ll actually follow it on a regular and consistent basis to keep it moving forward and actually affect change.

The plan I’d have written for someone else

Once I started looking for it, I had built in extra complexity all over the plan for the sake of just getting it done. I needed to slow down and really examine it thoroughly, which I did (yes, still from the couch, still while the game was on). Once I dug in deeper (yes, to the plan I wrote), I realized there was extra complexity everywhere and I wasn’t setting myself up for success.

The smart plug went. My Mac has a power reader built in. It only measures the chip, not the whole laptop, so the numbers will come in a little lower than if I were using the plug. That said, a number I’ll actually collect beats a better number I won’t.

Three folders became two. My notes for this project were going to live in three layers with different rules for each. That’s overkill. It’s now one folder for notes and one for instructions. While good in theory, I did not need a filing system with a philosophy.

A file I’d have to keep updated disappeared. The date on my newest note tells me how current everything is. I was overcomplicating the change log with scheduled tasks and text documents, and really, it’s just me on this system. So I ditched the log because it’s one less thing to forget, manage, or even look at.

Testing got lighter. The heavy comparison now runs once per option instead of every single time. As much as I love to thoroughly test things to be sure, that’s a trap for me. I would spiral and get lost in testing. Once is enough. If the results are going sideways, I can make decisions at that point.

Nine working sessions became six. Anything that doesn’t need my judgment happens between sessions. This means that the time I do spend working in the systems goes to deciding, not waiting. If getting this project off the ground felt daunting, I would have more reason to keep putting it off.

Boring, on purpose

The first thing that actually changes is converting documents. When I ask for a Word or PDF version of something, the AI will no longer perform this task. Instead, I’ll have a small script do the conversion. This saves usage and tokens and aligns with my overall project goal.

It’s the most boring option on my list, but it’s also the lowest risk. The result is either right or wrong the moment I open it. The bonus is that I don’t have to learn anything new. I can ask for documents exactly the way I do now.

The first change to a process you’ve spent years getting comfortable with should be one you can’t mess up. The more ambitious changes can wait until you can prove success. Second on the list is a monthly Asana report. Again, it’s low risk and boring, but it keeps me moving in the right direction.

This is what change management actually looks like: meeting people where they are, including when the person is just me. I’m sure there are apps, commands, and software that could do these tasks better. If I were going to use them, I already would be. Creating a plan built around tools I won’t use is a plan that fails. This is the best place for me, personally, to start.

If your first instinct reading this is to reply with “you should,” you’ve lost my attention. (I can hear some of you typing it now.) That’s exactly where change management goes wrong. You can’t always adapt every plan around your people and how willing they are. But understanding where they are and acknowledging it sets you up far better in the long run. If you lead with “you should,” you’re doing it wrong.

How I am actually starting

Getting started on anything is the hardest part. It’s why I spent so much time procrastinating in the documentation phase. So, like everything else, I am going to keep it simple. I am going to start with one step, once a week, 30 minutes or less. I can build that time into my schedule and protect it. Because I am keeping it simple, it comes before client work steals all my attention. Last month proved that client work wins every time the two are in the same room.

If a step finishes early, I stop. The next one waits a week. No catching up, no doubling up.

If I skip a week, nothing is lost. The next step is simply still next.

For accountability, I’ve looped in Kelsey, who asks whether I did the thing. We also created a simple Asana project where she can see if I’m keeping up (or not). Kelsey asks one question a week: did Wednesday’s step happen? She doesn’t review anything. She just asks.

Most rollouts are designed for the ideal employee

If you’re leading an AI rollout, look at what your plan assumes people will do. Check a dashboard every day. Remember a new step in a process they’ve run for years. Find an hour nobody has.

Plans like that look impressive and aggressive in the kickoff deck and stall by week three. Stakeholders get excited about the potential for progress, so they sign off. However, they are written for the team you wish you had, not the one that is struggling to keep up with what is already on their plate.

A simpler plan I’ll actually follow beats a highly technical plan I won’t.

Where have you written a plan for the person you wish you were? 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.

How are you managing the expectations of using AI?

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 to transform overwhelming platform guides into clear, actionable steps. You will discover how to turn complicated platform rules into simple daily actions. You will learn which three profile changes push your work into front feeds. You will watch a quick framework that shows how to track your growing audience without diving into complex spreadsheets. You will see how to test your posting schedule and keep only the tactics that bring new viewers.

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

Last time on So What? The Marketing Analytics and Insights Livestream, we looked at deep research. Catch the episode replay here!

This week on So What? we’re talking LinkedIn algorithm changes and measurement. 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.

This week, we dig into part 4 of 5 about what responsible AI is in practice. Responsible AI is one of those amorphous terms that can mean whatever the speaker wants it to mean, and it has perilously few practical tactical things we can do.

In this five-part series, we aim to correct that by providing practical tactical things you can do.

We have five core practices you can use to reduce your AI impact:

Use AI for what it’s good at.
Use cloud AI as little as possible.
Use AI for the tasks you hate.
Use local AI as much as you can.
Prohibit AI for what it’s bad at.

Part 4: Use local AI as much as you can

Back in part 2 of this series, we showed you what, on average, different cloud-based AI models cost in terms of resources per million tokens, and how something as simple as opening a one page Word document in a midrange model could consume the equivalent of a bottle of drinking water.

The antidote we mentioned in that issue was to avoid cloud AI and use local AI. This week, let’s expand on that.

First, a bit of table setting: local AI is the use of AI models that run on premises in your company. They can run on local servers, or even run on your devices, depending on how much compute power your devices have.

Any laptop computer with 32 GB of RAM or more can run at least some form of AI model, and the more memory (especially video memory, or VRAM) you have, the bigger and faster your local AI can be. For companies with bigger hardware footprints or perhaps existing graphics workstations that can be repurposed for AI, you might be able to do considerably more locally. To paraphrase, you can never have too much VRAM or too much disk space.

Getting started with local AI is an entire course or set of courses, but the basics begin with understanding what tasks are best suited for local AI. Generally speaking, local AI models are smaller, faster models that are not as intelligent as their frontier counterparts and thus require more task decomposition.

More task decomposition means taking a task and breaking it down into smaller, easier to navigate parts. This is especially important if you are using a model that has a relatively small amount of working memory, such as 32,000 or 64,000 tokens, which corresponds to 20,000 to 40,000 written words.

The current general best models for local use come typically from Chinese model makers. These models called open weights models are models that can be downloaded and run on your infrastructure. Because they are run on your own infrastructure, they are as safe as the rest of your IT infrastructure. You do not have any data privacy or security issues when you run the models yourself.

If you use the TRIPS framework by Trust Insights to analyze the tasks that you perform, which are time-intensive, repetitive, low importance, painful, and have sufficient good examples of what success looks like, identify those tasks that score the highest and are relatively small in scope or can be decomposed to small individual tasks. Those are the tasks to start using local AI with.

For example, each week, Trust Insights publishes an automated AI newsletter of the top AI news of the week and what it means for you. Typically, this is a newsletter of 10 to 15 articles and short summaries all generated by the Alibaba Qwen model that runs on a MacBook Pro. This uses no data centers, no freshwater, and the energy from it is entirely renewable because it’s powered by the solar panels on my house.

Each week, this consumes 3-5 million tokens per newsletter, which amounts to 3,000 watt-hours and 900 fluid ounces of water at a hyperscaler data center, or about 600 smartphone charges and 54 bottles of water. We save that entirely each week because we use local AI for it.

One of the secrets to local AI is that the harness, aka the environment that you use the model in, matters just as much as the model itself. A small lightweight model can get a tremendous amount done if it has good guardrails provided by its harness. Our process at Trust Insights is to use a frontier model to tune the harness, and then the local AI model runs itself inside the harness week after week, consuming far fewer resources.

Stay tuned next week as we cover the fifth and final part in our series, which is not using AI for things it’s bad at.

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?

MEMA, October 2026
Women in Engineering, 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