AI that works for your people, not instead of them.

I'm Felix Gallo. I run Humanetic, a one-person consultancy in Wausau, Wisconsin. I help organizations work out where AI can be useful to them and how to use it safely. Both are harder than they look. I've spent more than thirty-five years managing and shipping technical projects for Amazon's Alexa, Activision, Disney and others, and much of what I learned there applies to organizations of any kind.

Email me The first conversation is free.

The core problem

Tools like Claude write so fluently that it's easy to treat them like a colleague. A colleague who has worked with you for years can take a vague request and fill in the gaps, because they know your organization, your members, and what's happening this week. Claude knows none of that. Whatever you leave out, it fills in with the average of everything comparable it has read, and that average is usually boilerplate.

The same tool can give one person filler and the person at the next desk good work. Usually the difference is that the second person supplied the specifics. Doing that well is a skill anyone can learn, and it doesn't need a technical background. It's the first thing I teach, and the more interesting uses of AI build on it.

AI 101 – Getting Started

The first hour

A one-hour session for your whole staff. Skeptics are welcome; they're usually the best at catching the tool's mistakes, which makes them good reviewers. Everyone reads a short document, then watches three live demonstrations using your own work. In the third, I make Claude fail on purpose so your team can see where it breaks before anyone is asked to rely on it.

A six-week trial

The trial uses real tasks, and each one gets a timed baseline before AI is involved. Your team builds a shared prompt library and keeps a log of every fabrication caught in review. Each stage ends with a go/no-go decision. Internal work comes first, and outward-facing drafts come only after the team has shown it reviews carefully. For a small team the software costs under $500 for the whole trial, on month-to-month plans. At the end you keep the uses that measurably saved time and drop the others. If none of them work, you still have a documented record of what AI can't do for your organization, which is useful the next time the question comes up.

Data and review rules

A data rule short enough for an index card: public material: use freely; routine internal material: permitted, on a plan that doesn't train on your data; anything shared with you in confidence: never entered. And a review standard: nothing leaves the building until a named person has reviewed it and is accountable for it.

AI 201 – Next Steps

Once the habits are in place and the trial has shown which uses are worth keeping, there's a lot more you can do. A follow-on engagement might cover:

Or whatever your situation calls for. I enjoy unusual problems.

Subtle Risks

  1. It makes things up. When an AI reaches the limit of what it knows, it can invent facts, and it can even claim to have checked them. It says these things in the same confident tone it uses when it's right; a made-up phone number reads exactly like a real one. So facts should go into the tool from you, and nothing should come out of it and reach the public without someone verifying it.
  2. Unedited AI writing is easy to spot. It has recognizable habits, and more of your customers notice them every year. A letter that was obviously machine-written tells a member they weren't worth a person's time, which defeats the purpose of writing to them. Every draft should get a human edit, a detail only you would know, and a real signature.
  3. It shouldn't make decisions. It has no stake in the outcome, no accountability, and no knowledge of the people involved. Use it to lay out a decision (the options, the criteria, the case against) and have a named person make the call. Anything it produces needs review by an accountable person before an email goes out or a post goes up. On legal, tax, HR, or safety matters, it can offer advice and nothing more.

The pocket card

The handout your team keeps after the first-hour session. It summarizes the method on one card.

Working with Claude Humanetic LLC

The main idea: whatever you leave unspecified, it fills in with the average of everything it has read. The more you specify, the less it has to guess.

1Say what “good” meansEnd each request with pass/fail criteria, and have it check its draft against them before showing you.
2Brief it every timeGive it a minute of context at the start of every session. It knows nothing about you until you tell it.
3Show it your voiceGive it two or three of your best past pieces and ask it to match them. Describing your voice in adjectives doesn't work.
4Let it ask questionsIf the task is fuzzy, ask it for “the five most useful questions, one at a time” before it proposes anything.
5Push back on draftsGive specific feedback, more than once. If a draft looks perfect, read it again more slowly.
6Check every factFacts come from you, not from it. Verify every name, number, and date.
7Make the decision yourselfAsk it to argue against a plan before improving it. It can lay out a decision; a person makes it.

Who this is for

Organizations that do a lot of writing and structured paperwork, such as startups, small and mid-sized businesses, established companies, nonprofits, professional offices, and community groups, and that have always had more work than hours. In a town the size of Wausau, that's most of Main Street. I work in person across central Wisconsin and remotely elsewhere. The habits, prompt library, and policies I help set up belong to your staff, so you can carry on without me once we're done.

Get in touch

If you're wondering what AI is good for in your organization, and what it's bad at, I'd be glad to talk it through. The first hour is free, and the coffee or New Glarus is on me. It helps to know what your organization does, which work takes up the most hours, and how your team feels about AI right now. Most teams have both enthusiasts and skeptics. I usually reply within one business day.

felixgallo@humanetic.ai

Felix Gallo

Founder, Humanetic LLC · Wausau, Wisconsin