INBOX INSIGHTS: Local AI Migration Part 3, Practical Responsible AI Part 2 (2026-09-16)

INBOX INSIGHTS: Local AI Migration Part 3, Practical Responsible AI Part 2 (2026-09-16) :: View in browser

👉 Take our new course, AI for Writers and learn how to make AI write better!

Watch this video on YouTube

🎧 Listen to this newsletter as an MP3

Local AI Migration: Part 3

Why hasn’t this moved in a week?

I told you at the end of the last issue that we’d dig into what I learned once I finally created the PRD.

I didn’t create the PRD.

I was on vacation the week of September 7th, which I planned, and I was offline from work. Then I came back on Monday the 14th to a calendar that had filled in behind me and an out-of-office responder I forgot to turn off, which cheerfully told four different clients to go contact Kelsey while I sat at my desk answering their emails. By the time I went looking for a clear hour to sit down and think about local AI, there wasn’t one left. The project did not move an inch.

Here’s my discomfort with that. I’ve now published two issues about the discipline of writing requirements down before you spend money, and then I went a full week without touching my own requirements document. If you were reading along and planning to do this alongside me, I left you standing there.

The question isn’t “how do I find more time?” I’m not going to find more time. What I want to know is why this specific project was the one that was lost when a dozen other things I was equally behind on somehow got done.

The Accountability Factor

Look at what actually moved on Monday: two decks, built back to back for two different clients; a set of workbooks for a tech stack audit, because there’s a client on the other end expecting them; a written response to an executive about the six topics he wants covered in next week’s read-out, because he emailed me directly; a kickoff call for an October workshop, because the organizers were already on the Zoom; and a first conversation with a prospect, because they’d put it on my calendar.

Every one of those had a person attached who would notice if it didn’t happen.

My local AI project has exactly one person attached to it, and that person was on vacation.

I will drive twenty minutes to do something I won’t do in my own living room. If I sign up for a class, somewhere, at a time, with other people in the room, I’ll go every time. If I tell myself I’ll do that same workout at home on Tuesday, I will find something else to do on Tuesday. It’s easy to find something “better” to do even though it’s the same thing I paid to do in someone else’s space. The only variable is whether anyone is expecting me. That’s my trap, and I know it’s my trap, which is the part that bothers me. I’m usually pretty good about this. Until I’m not.

Accountability isn’t a character trait you either have or you don’t. It’s a structure you built. The class works because I paid in advance, somebody takes attendance, and there’s a visible empty spot if I skip. The living room version has none of that, so it rests entirely on me being a disciplined person on a Tuesday, which is something I’d like to believe about myself and which the evidence does not support.

My client work is the class. My local AI project is the living room.

My carbon footprint had an excellent week

There is one upside that is worth reporting: my AI usage last week was exactly zero, not reduced—zero.

After two issues of careful thinking about the energy cost of the way I work, the single most effective intervention I’ve found so far is going on vacation. I’d love to bottle it. The trouble is that the method doesn’t scale, my clients have opinions about it, and it turns out the thing I was actually optimizing for was never the electricity.

My own data says this isn’t a one-week thing. I went and looked, which I don’t recommend doing on a Monday. I have 100 open tasks assigned to me in Asana. 30 of them are past the date I gave them. The oldest is from February.

It gets more specific than that. Two of those aren’t loose tasks at all, they’re entire project plans that stopped: a service launch, last touched in February, and a lead generation program, last touched in January. Both are things I decided were important. Both are internal. Neither has a client waiting on the other end.

Kelsey, whose job it is to keep me on track, put two focus hours on my calendar this week specifically so I’d have protected time. I declined both of them Monday morning to make room for client work.

This is a People problem, not a Process one

My instinct was to blame my calendar, which means blaming Process. Better time blocking. A better system. That’s the comfortable diagnosis because it’s the one you can buy software for.

It’s also wrong. In the 5P Framework by Trust Insights™ (Purpose, People, Process, Platform, Performance), “People” isn’t only about who does the work. It’s about who’s accountable and who notices. My local AI project has one name in the People box, and that name also runs a company. There’s no second person, no check-in, and nobody who’s going to ask me Thursday how it went.

The process was fine. I wrote a solid PRD outline. No process survives an audience of one.

The version that costs real money

Run this same week against an enterprise AI pilot, and nothing changes except for the number of zeros.

MIT’s NANDA initiative published a report in 2025 finding that roughly 95 percent of enterprise generative AI pilots delivered no measurable return, and that number got repeated just about everywhere. It’s also worth reading the pushback from SmarterX, which argues the finding has been badly oversimplified in the retelling. Both things are true: the number is softer than the headlines made it seem, and plenty of pilots really do stall.

What I’d add from my own week is a fairly boring mechanism for the “why.” Most internal AI pilots have no client on the other end. They have a sponsor who is enthusiastic in the kickoff meeting and busy by week three. They’re the living room workout. Not unimportant, just unwitnessed, and the first thing to yield the moment a paying client needs something.

Nobody explicitly kills those projects, that’s what makes them so hard to spot. They just never get chosen first. [insert sad trombone]

What I’m actually doing about it

I’m giving the project a witness. The PRD review goes on the agenda for my standing Friday session with Chris. Not because Chris needs to approve anything, but because he’ll ask, and then I’ll have to have something to show him.

I’m publishing the deadline to you. The PRD gets finished this week, and part 4 is what’s in it. You’re the audience I was missing. That’s a mildly manipulative use of a newsletter, and I’m doing it anyway.

I’m not adding a system. No new tool, no new tracker, no calendar rebuild. That’s “Platform” thinking, and I’d only be doing it to feel productive.

If you have an internal AI project that hasn’t moved in a month, don’t ask whether it’s still a priority. Everyone says yes to that. Ask who would notice on Friday if it didn’t move again this week. If you can’t name that person out loud, you don’t have a stalled project. You have a project that was never going to start.

How are you keeping your internal projects honest?

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 discusses modern billing challenges. You’ll discover how to charge fairly while keeping your unique skills central. You’ll learn to spot which chores machines can handle and which need your personal touch. You’ll find ways to organize your daily routine so buyers see clear worth no matter how fast the work finishes. You’ll walk away with a simple method to track whether your new shortcuts improve results.

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

Last time on So What? The Marketing Analytics and Insights Livestream, we talked through how to analyze podcast data in new and different ways. Catch the episode replay here!

This week on So What? we’re examining billing models of different kinds of agencies in the AI era. 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 2 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 2: Use Cloud AI as Little As Possible

When we talk about cloud AI, we’re specifically referring to AI from hyperscale providers like Meta, OpenAI, Anthropic, Google, Microsoft, and others. These are the companies building massive data center projects around the world, data centers that are often consuming vast quantities of electricity and fresh water.

How much they consume is a matter of debate and varies wildly from provider to provider. Some providers focus more on sustainable usage than others; for example, Google in its annual sustainability reports shows that they’re about 64% water neutral, meaning for every 3 units of water they consume, they put 2 back.

However, the reason to avoid hyperscalers is specifically because of this architecture. Yes, as some advocates argue, a hyperscaler can process a lot more data a lot more efficiently in theory, but their size creates problems with things like heat that fresh water cooling has to address. Edge-distributed computing – running workloads closest to the user – may be less efficient in aggregate, but more overall resource efficient because a data center doesn’t have to be powered and cooled when no work is being done.

How much power and water do data centers consume? We can make an educated guess based on SEC filings, sustainability reports, and prices of compute. In accounting, there’s a concept called cost of goods sold (COGS), which is what it costs a company to manufacture a product. Big tech providers publish this in their SEC filings. And we can infer from that plus what providers charge for API costs what the underlying likely costs are, in energy and water.

For example, for 1 million tokens, Anthrophic charges $10 per million tokens input, and $50 per million tokens output for their Claude Fable 5.1 model, their flagship model. They charge $2 per million tokens input and $10 per million tokens output for Claude Sonnet 5, 20% of Fable’s costs. Claude Haiku is half of that again.

When we look at the margins disclosed in earnings reports and SEC filings for big hyperscalers like Google and Meta, and extend that to other models using the same COGS formulae, we end up with this basic set of estimates. (download the full report for free, no form to fill out, no registration, from our Instant Insights here)

To give you a sense of added perspective, in the average AI system, opening and summarizing a one-word Word document in Claude Code, absent any other plugins and addons besides the docx skill, consumes 52,000 tokens. By the basic math in the chart above, 52,000 tokens on a typical model (Sonnet 5) consumes a bottle of water or the same amount of carbon as driving a quarter mile.

A single Word document.

When we use non-cloud AI (local AI) as Katie’s been sharing in the newsletter these past few weeks, our individual laptop may not have the scale of a big hyperscalers, but depending on the laptop you have, you may be using considerably less energy. For example, the MacBook Pro M4 Max, the one I have, draws 160 watts of power at peak usage. An entire hour of usage at full power is 16% of what a single million token run costs in terms of power consumption for a midrange model.

If you have the ability to run non-cloud AI models, do so, especially if you have the ability to run mainstream-capable models (which often need generous amounts of RAM/compute locally). Barring that, the most responsible thing you can do when it comes to using AI, at least from a sustainability perspective, is to use it only when you absolutely need to.

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?

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