Behrends Consulting

You switched on AI. Now make it work.

Your people use AI, each in their own way. Nobody sees the full picture, and the results don't add up for the company. I look at it from the outside, find what's missing, and build it with your people.

Mario Behrends
Mario Behrends

01  The problem

Sounds familiar?

  • i.Your team has AI tools. Apart from typing into a chat, not much has changed.
  • ii.Everyone got access. But is there a plan for how to use it, and why?
  • iii.Your people are busy with their real work. Nobody has time to figure out AI on the side.
  • iv.The few who want to go further start on their own. Their results stay personal. What does the company get?
  • v.You pay for AI licences every month and can't say what you get for it.
  • vi.Nobody has asked staff, team leads and management how they actually use AI. So nobody sees the full picture.

And then?

  • a.Now the whole company has AI. Teams start building for themselves. Worse, so do single employees. Every process gets personal and only works for the person who built it.
  • b.Colleagues have no plan to follow. No clear rules for how AI should be used, how it works, or what to watch out for.
  • c.Nobody knows how to get high-quality output, or how to stop the token burn they don't even know is happening.
  • d.Everyone's AI work runs side by side. The results don't add up for the company.

So where do we go from here?

02  The approach

Four levels. Each one builds on the one below.

It starts with people who know how to work with AI. Departments, leadership and the company build on that, and each level holds up the others. Pick a level to see what it solves.

The problems from above, by level

  • i.Just a chat
  • iii.No time
  • c.Weak output, hidden token burn
  • iv.People go it alone
  • a.Processes only the builder can run
  • d.Results don't add up
  • ii.No plan
  • v.Unclear what licences bring
  • vi.Nobody sees the full picture
  • b.No rules for using AI
  1. "I got access. Now what?"

    Each person learns how AI behaves and what to watch out for: where it gets its context, why long chats get sloppy, how to get good output without endless prompting. Then we find the use cases for their own role. From there they figure out the rest with AI as a partner, wherever it's useful.

    Solves

    • Just a chat
    • No time
    • Weak output, hidden token burn
  2. "We all use AI differently."

    Everyone brings their own workflows and setup, and the output stays personal. We agree on a shared base: which tools, why, where and how, and one source of truth, with each person's own AI on top. Use cases get picked by what moves the department's goals.

    Solves

    • People go it alone
    • Processes only the builder can run
    • Results don't add up
  3. "Where do we go with AI, and why?"

    Who goes AI-enhanced, who goes AI-first, and who doesn't need it, backed by business reasons. How much to invest, and based on what: more output isn't more revenue, so we measure the gains. It's clear what gets rewarded and what gets stopped.

    Solves

    • No plan
    • Unclear what licences bring
    • Nobody sees the full picture
  4. "How do we keep up?"

    Every employee gets a baseline: clear rules for how to use AI and what to watch out for (the goal: 2 prompts, not 20), and those who want more get a path. Your knowledge sits where AI can use it, with no lock-in to one provider, and the setup keeps up with new models at a pace that doesn't overwhelm anyone. Company-wide only where it makes sense.

    Solves

    • No rules for using AI

If AI is the wrong first step, I'll say so and tell you where I'd start.

Read the full approach

03  Fit

When I say no to AI strategy

I take on AI strategy only when your core business works and the budget is there. If it doesn't fit, I tell you.

  • Wrong problem. AI makes a broken process run faster. It doesn't fix it.
  • No budget. Setup and training cost money. If it's not there, now isn't the time.
  • No payoff this quarter. AI strategy won't fix this quarter's numbers. If those come first, we start with the business.
What I don't do

Turnaround

When the business itself is stuck

Then I start there, before any AI. I find what holds the business back: Google, your website, your channels, your pricing.

You get a report with steps you can act on, and my help carrying them out if you want it.

How turnaround works

04  Sample Scan  /  Fictional company

See what you get

A 150-person firm gave every team AI tools. The Scan showed who uses them, for what, and which step comes first.

Read the sample Scan (in preparation)

05  Proof of practice

I run my own businesses this way

Each one has its own page that shows how it works, step by step.

The office behind all of it

  • I run my work through 7 AI agents: CEO, UX, sales, marketing, legal and money, QA and review.
  • They share one ticket board. A task is only done when there's proof it works.
  • I set the goal and make the calls.
How the office runs

Consumer service, live

771agencies in 14 Dutch cities

RenterIntro

One payment, and a renter's intro lands in the inbox of every rental agency in their city.

How it runs

Free jobs site, live

2.658jobs at 195 companies

First Job Daily

A free jobs site for internationals: open jobs at employers on the IND sponsor register. Updated every day.

How it runs

StudyfinderAI outreach, paused

4.199emails to 1.163 institutions in 19 countries

Outreach on fixed rules

Contact research, sending and reply sorting for a two-founder startup. AI proposes, rules decide, a person takes every warm reply.

How it ran

StudyfinderAI, website relaunch

30days from first commit to live

A website built by AI agents, checked by rules

Every claim checked against a register, every change through automated tests. Accessibility 100 in the lab.

How it was built

06  About

Mario Behrends

I'm a one-person business. I work close to owners, managers and leads, and we build the work together.

Before this I fixed aircraft at Airbus, built my own Shopify brand and supported e-commerce teams.

More about me

07  How I work

Five guiding principles

They decide how I work with your team, with AI and with your numbers.

AI is a dumb kid when a chat opens, and a professor once you point it to the right sources.

  1. Clear sourcing. Every number I give you comes with its source. If I don't know, I say so.

  2. Use the best model on real work. I use the best models out there on the tasks your people do every week.

  3. No sugarcoating. I tell you what I see, plainly, even when it's uncomfortable. When AI or more spending isn't the answer, I say so.

  4. We work together. AI isn't an intern. Treat it like a coworker. Your people work with it side by side, and they own what we build.

  5. Context first. AI is a dumb kid when a chat opens, and a professor once you point it to the right sources. So the first job is a clean foundation: your knowledge, in order, where the model can read it.

    Quoted above

Tools

I work with Claude and OpenAI (ChatGPT and Codex).

No lock-in. The setup and the knowledge stay with your company, in your own accounts.

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.