We Believe

Responsible AI principles

AI that gives people their time back, under rules you can check.

These are the principles behind every engagement at eMediaAI. They're short on purpose. Each one is something you can hold me to in writing, and each one shows up in how I scope, build, and hand off.

20%

The line. If automating a task would return under 20% of someone's time, I recommend deleting the task.

Applied to every workflow I score

30 days

The handoff standard. A system is done when your team runs it for 30 days without me.

Every build

0

engagements taken whose goal is cutting jobs.

Written into every engagement letter

Principles only count when they show up in a contract or a build. These 3 do.

The order

People first, then process, then tech.

Every engagement runs in that order, every time. Here's what each step means in practice.

People

Start with whose day changes.

Before any tool comes up, I talk to the people who do the work and find the tasks they'd happily never do again.

Process

Fix it or delete it first.

A broken process with AI on top is a faster broken process. So the workflow gets mapped, cleaned up, or cut before anything gets built.

Tech

Pick the tool last.

The tool gets chosen against a written spec, with error handling, a data boundary, and a named owner before launch.

Most failed AI projects run that order backwards: tool first, people last.

The principles

7 principles, in plain English.

Each one is a commitment with a test you can apply to my work.

→

People first. A project goes ahead only if the people doing the work end up with time back.

→

Fairness. Any system that scores, ranks, or routes people gets checked for bias before launch and again when its data changes.

→

Safety. Every system has written error handling, and a person signs off on anything a customer will see.

→

Privacy. Your data stays in your accounts, training on your inputs is switched off wherever the vendor allows it, and the data boundary is written down before launch.

→

Transparency. I document which inputs drove each recommendation, so you can see the reasoning behind every answer.

→

Accountability. A named person on your team owns every system, and I'm accountable for everything I ship.

→

Augmentation. Every build is scored on the hours it returns to a person, and I decline work whose goal is cutting jobs.

The rule I'm known for

Eliminate before you automate.

Automation carries a build cost and a maintenance cost. A task that gives back a sliver of someone's day rarely pays either one back.

So the line sits at 20%. Under it, I recommend deleting the task. Over it, the task gets scoped and priced in hours a year.

The exception is a task where an error costs real money or quality. Then it stays, and it gets a human check.

In my own business

How I use AI at eMediaAI.

The same rules apply in-house. Examples, and they change as the tools do.

Drafting

First drafts of posts, reports, and proposals. I edit every one before it goes out.

Call notes

Calls transcribed and summarized into notes and action items. Nothing a client reads goes out unreviewed.

Research

Firecrawl and NotebookLM for gathering and reading sources. Every number gets checked against where it came from.

Builds

Claude Code and Codex for writing the systems I build, with the same error handling a client build gets.

Testing

Langfuse and Promptfoo to test prompts and agents before they reach production, and to watch them after.

How you get there

Where the principles get applied.

1. AI Opportunity Blueprint™

10 days, $5,000 flat. I map your workflows, apply the 20% rule, price each candidate in hours a year, and hand you a sequenced roadmap you can run with anyone.

2. Fractional CAIO

A Chief AI Officer on retainer for one fixed monthly fee, quoted after the Blueprint. 3-month minimum, then month to month.

3. Builds

Scoped from the Blueprint and quoted as one project fee. Your team, your vendor, or me.

If the Blueprint doesn’t identify at least 200 hours a year of removable work, each with a named mechanism and an implementation estimate, you don’t pay. You keep the document either way.

Questions owners ask before the call.

Is our company data used to train public AI models?

Not by anything I set up. Systems run in your own accounts, training on your inputs is switched off wherever the vendor allows it, and the data boundary is written down before launch.

Does people-first mean you won't automate anything?

I automate tasks all the time. What I won't take on is a project whose goal is cutting jobs. Every build is scored on the hours it returns to the person doing the work.

How do you handle AI mistakes and bias?

Anything a customer sees gets a human sign-off first. Systems that score or route people get a bias check before launch, and prompts and agents are tested before production and monitored after.

Do you follow a recognized standard?

The controls are mapped to the NIST AI Risk Management Framework. Where the EU AI Act applies to you, the policy says which parts. For most $5M–$50M companies, it doesn't yet.

Who's accountable when something goes wrong?

I am, for anything I built, and a named person on your team owns each system day to day. Every system has a written plan for what happens when it fails.

Hold me to these on the first call.

Bring the one process that hurts. Within 24 hours of the call you'll get a one-page note naming the 2 or 3 workflows where AI would pay off first, whether or not you book anything.

Read first, book later

AI whitepapers, by seat.

Summarize This Page With Your Favorite AI

© 2026 eMediaAI.com. All rights reserved. Terms and Conditions | Privacy Policy | Partner Terms