// the method

The 5-step framework, in detail

One document does not transform a company. A sequence does. Treethree runs on five steps that build on each other, from a map of what is possible to a system running in your business. Here is what each step produces, what it costs, and why the first one is free and useful on its own.

Most AI advice arrives as a single artefact: a report, a slide deck, a workshop. You read it, you nod, and nothing changes, because a one-off document cannot carry a company from "we use ChatGPT" to "our operations run on agents". That distance is covered in steps, each one gated on the last, each one sharper than the one before.

The framework has five. The first is an online assessment you can run today. The other four are an in-person engagement that only makes sense once the first has shown you where you stand. You never pay for a step until the previous one has earned it.

1 2 3 4 5 free Landscape Direction Deep Dives Plan Execution online in person in person in person in person
Landscape of Choices, Direction Setting, Path Deep Dives, Tailored Implementation Plan, Execution Cascade. Each gated on the one before. Step 1 is online, steps 2 to 5 are in person.

Step 1: Landscape of Choices

The full map of what running on AI could look like for you. This is the online assessment, and it is where everyone starts. You answer questions about how your team works today, and you get back a real sample report: where you sit across the four modes, a function-by-function read, and an honest line on what should stay human. A free calibration sharpens it before you pay for anything.

The sample is genuinely free. The full report, the version that prices the tools, sequences the build, and lays out your first 90 days, is 450 EUR, and you only pay once you have read the free version and decided it is worth it. Most companies could stop here and have more direction than a 20,000 EUR consulting deck would have given them.

Step 2: Direction Setting

A working session on your functions, team, and tools. We set the direction together. Not a form. The report tells you where you stand. Step 2 is where a person sits with you and turns that into a decision: which functions move first, what your team can realistically absorb this quarter, which tools you already own that should anchor the build.

Real margins, the brittleness of your integrations, how fluent your team actually is: these come out in conversation, not in a dropdown. It is also where we say no out loud, when a function is better left alone for now.

Step 3: Path Deep Dives

Function by function, in payoff order. With the direction set, we go deep on the functions worth automating first, in the sequence that pays back fastest. Each deep dive covers the specific tools and models, the data the agents need to read, the handoffs to people, and the failure modes to watch. You finish with a concrete picture of how one function actually runs on AI, not a generic promise that it could.

Step 4: Tailored Implementation Plan

Phased, costed, every step gated on the last. The deep dives become a build plan: what to stand up first, what each phase costs to run and to set up, and the gate that has to be cleared before the next phase starts. Nothing is "do everything at once". Each phase has to earn the next, the same logic that governs the five steps themselves.

Step 5: Execution Cascade

The live build: an iterative plan that keeps evolving and adapts as your team builds and executes. This is the part most frameworks skip. A plan on paper meets reality, reality pushes back, and the plan has to move. Step 5 is the working surface where the build actually happens, with hand-holding only where you want it. Your team owns the system. We keep it honest as it grows.

A report is not a transformation. It is the first step that makes the next four worth taking, and the only one you have to take to get value.

Why a sequence, and not a single report

The reason to build this as five gated steps is the same reason most AI projects fail. The MIT NANDA study found in 2025 that 95 percent of enterprise AI pilots showed no measurable P&L impact. The pattern underneath that number is almost always the same: a company bought one big thing, all at once, with no sequence and no gate, and discovered too late that the groundwork was missing. A gated sequence makes that failure cheap to catch. If step 2 reveals the data is not ready, you fix the data before you spend on step 4.

Free sample Full report, 450 EUR In-person engagement, steps 2 to 5 Step 1, the report (online) The engagement
You only pay for the next step once the last one has earned it.
1 of 5
the free report is the first step, and useful on its own
450
EUR for the full report, only after you have seen the free version
~10 min
to your free sample report, no card required

Where to start

You do not commit to five steps. You take the first one, which is free, and decide from there. If the report is enough, use it and go. If it makes you want the costed plan, the full report is one click and 450 EUR. If you want a person to set the direction with you, that is step 2, and not before. The sequence is built so the value shows up early and you never pay ahead of it.

See your own version of all five steps.

The free assessment is step 1: your four-mode position and a function-by-function read in about ten minutes. No card required, and the full report is only 450 EUR if you decide to go further.

Start assessment
Keep reading
The four AI modes Data as the groundwork Eight mistakes to avoid