1. Home
  2. Approach

Four levels. Built with your people.

How I look at AI in your company, level by level, how we work together, and when I'll tell you it's the wrong time.

The four levels

The four levels, in detail

Each level builds on the one below, and each one holds up the others. Company-wide only where it makes sense.

  1. Employee

    Most people get access and then good luck. Nobody tells them what to watch out for when they write with AI, or why a chat gets sloppy the longer it runs.

    Or why there are different models, what the effort setting does, why the chat, the desktop app and the terminal behave differently, and what plugins, skills and MCPs are for.

    I go through how AI behaves and what to watch out for, so everyone gets good output. Then we find the use cases for each role: a UX designer works differently from a marketer or a product manager. From there, people figure out the rest with AI as a partner.

  2. Department

    Everyone brings their own workflows, scheduled jobs, artifacts and skills. The output stays personal, and only the person who built a process can run it. The department needs a shared base: which tools, why, where and how.

    An example from code: one shared repository as the source of truth for everyone, and each person's own Claude on top of it.

    A department also works towards its own goals and numbers. Use cases get picked by what moves them, so the output meets the department's bar instead of turning into an AI Frankenstein. Commercial teams don't get software-team workflows.

  3. Leadership

    Leadership sits at every level: team leads, heads of department, management. It's about decisions, not job titles. Leadership needs a few things clear:

    • The direction. Who goes AI-enhanced, who goes AI-first, and which teams don't need AI at all, and why.
    • The investment. How much to spend, and based on what. More output isn't more revenue, so the gains get measured.
    • The lines. What gets rewarded, and what gets stopped.
    • The people. Not everyone is excited about AI. Some start to question their role. They get a plain answer.
  4. Company

    The company has to keep up in its market, even when it isn't a tech company. That takes a foundation everyone stands on:

    • The baseline. Clear rules for how to use AI and what to watch out for, so nobody just tries things at random. The goal: 2 prompts to good output, not 20.
    • The path. How the people who want more get support.
    • The knowledge. Your company knowledge sits where AI can use it, in your own systems, with no lock-in to one provider.
    • The pace. AI moves fast. The setup keeps up with new models, and people stay interested without getting overwhelmed.

How we work together

Four steps. You can stop after any of them.

  1. Strategy

    First I talk to management, team leads and staff, and map all four levels: how each person, team and department uses AI today, and where it gets stuck.

    Then we decide with leadership what's company-wide and what stays per team, who needs which level and why, who owns what, and what data never goes into a model.

    You'll know what your AI licences cost and what you should get from them.

  2. Setup and knowledge

    I set up the tools, accounts and admin settings in your own accounts, so you're not tied to me or to one provider.

    Then the knowledge. Your projects, documents and ways of working become sources the model can read. This is where AI stops being a dumb kid.

  3. People and first workflows

    Most people don't have time to figure out AI on their own. So each person gets trained to the level their role needs, on their own tasks.

    Together with your people I make the first workflows or agents. They own them and run them after handover. I don't keep anything running for you.

  4. Care

    Models, settings and staff change every month. A monthly check keeps the setup and the knowledge sources current.

    New people start from the same setup. You can stop any time with a month's notice.

Fit

When AI strategy is the wrong buy

I only take on AI strategy when your core business already works and there's budget for it. If that's not you, I'll tell you on the first call.

  • Wrong problem. AI speeds up what is broken. If sales, channels or prices are the problem, faster tools make the same mistakes faster.
  • No money for it. Strategy, setup and getting people ready cost money and time. If there's no budget, I'd rather say so than do half a job.
  • No payoff this quarter. AI work pays off once people work with it. It won't save this quarter.

In those cases I say so, and we look at the business first.

Turnaround

First the business, then AI

Sometimes AI isn't what you need right now. Sales are flat, the website loses people, or the prices were never tested. Then I start with the business.

  • What I look at. How people find you on Google, your website, your channels and your pricing.
  • What you get. A report with steps you can act on, in order.
  • What happens next. I help carry the steps out with your team, as far as you want me to.
  • Where AI comes in. It speeds up my analysis, so you get the report sooner.

Fixed price per project, set after a free 30-minute call.

Limits

What I don't do, and what I can't

Better you hear it now than halfway through a project.

  • I don't build apps or run software for you. We build the first workflows and agents together, and your people own and run them.
  • I don't promise savings, revenue or usage numbers. Those depend on your team. I promise my own work, on the dates we agree.
  • I don't do legal, data protection or security work. When the AI Act or data rules come up, your lawyer or data protection officer answers them. Compliance stays with you.
  • I don't set up Microsoft Copilot. I work with Claude and OpenAI (ChatGPT and Codex).
  • I don't bill by the hour. Every package has a fixed price.
  • I can't fix a broken business with AI. If that's the problem, we start with the business.
  • I can't be in ten places at once. I'm one person, so I take on a few clients at a time. You always work with me, never a junior.

No lock-in.

The setup and the knowledge stay with your company, in your own accounts. You can stop after any step.

Challenges you may recognise

  • i.Everyone got access, then good luck. Nobody said what to watch for. I start with a baseline per role: a marketer and a UX designer need different things.
  • ii.A few want to go further, the rest has no time. I give the few a base to build on and share, so the rest copies what works instead of starting alone.
  • iii.The business itself is stuck. Then AI only speeds up what is broken. I say so in the first call and start with Google, the website, channels and pricing.

Contact

Tell me what is stuck

Book 30 minutes on video, or write me a few lines about which AI tools your team uses and where it gets stuck.

Book a 30-minute call

consulting@mario-behrends.com
LinkedInEnglish and German.