The 3 moves: analyse, optimise, scale with AI

Every week we receive the same message: “We want to integrate AI into our company.” Every week we give the same answer: “Where is your data?”

Most do not know. The data lives in spreadsheets, WhatsApp, emails and the heads of three people who have worked at the company for ten years. That is a problem, because AI without clean data is not intelligence. It is an expensive parrot.

The order matters more than the budget

We have seen companies invest €50,000 in an AI chatbot before cleaning their customer database. The result: the bot answered questions with incorrect information, hallucinated products that did not exist and created more work than it eliminated. It was not an AI problem. It was a sequencing problem.

At SOM-OS, we work in three sequential moves. We do not skip or mix them. Each builds the foundation for the next.

Move 1: Analyse

Before writing a line of code, we map the complete system. This is not a 15-minute interview. It is a systematic exercise that answers three questions:

• Where does your data live today? Excel, Google Sheets, WhatsApp, email, physical paper, the boss’s head. Everything. No exceptions.
• Which tools do not talk to each other? The CRM does not know what the calendar is doing. The calendar does not know what the invoicing system billed. The invoicing system does not know what the team recorded in the field.
• Which processes consume the most manual time? Copying and pasting between systems. Forwarding emails. Calling to confirm something already written somewhere. Tasks a human does that do not require human intelligence.

This move ends with a map. A 5–8-page PDF detailing where the chaos is, which bottlenecks exist and which process creates the most friction. We call it the Chaos Map. The client keeps it, whether or not they pay for the next step.

Move 2: Optimise

This is where we build. Custom software, on the existing foundation, replaces the processes that consume the most time. It is not an isolated app. It is a system connecting what was separate.

The goal of this move is not to make everything digital. It is to make data flow without human intervention. For a quote to become an automatic work order. For completed work to generate an invoice. For an invoice to synchronise with the payment schedule. Without copying anything by hand.

We build on Foundation, our monorepo template: Nuxt 4 + NestJS 11 + PostgreSQL + Redis + BullMQ + LangChain. We do not start from zero. We assemble and adapt. That is why we deliver in 3–5 weeks what other agencies take 3–5 months to deliver.

When this move ends, the company has working digital architecture. Data lives in a structured database, not scattered sheets. Processes flow from one point to the next without manual intervention. And — this is the key — the architecture is ready to receive intelligence.

Move 3: Scale with AI

Only now do we deploy AI. Not before. It is not a generic chatbot. It is an intelligence layer operating on the clean data and digital architecture from the previous move.

What makes this possible? AI now has context. It knows which products exist, the status of each order and which client needs what, and can make decisions with correct information. It does not hallucinate. It does not invent. It executes.

Types of agents we deploy during this move:

• Qualification agents — talk with leads before they reach a human. Suggest the right product or rule them out. As at Sando Capital: −70% of the time spent qualifying investors.
• Operational agents — automate repetitive decisions. Assign work, generate reminders, escalate alerts. They work on system data, not generic prompts.
• Analysis agents — cross-reference data that previously required hours in Excel. Automatic reports, demand forecasts, customer segmentation. Without exporting or importing anything.

Skipping steps guarantees failure

The most expensive mistake we see: companies wanting move 3 without moves 1 and 2. They buy AI, plug it into analogue chaos and expect magic. What they get is faster, more expensive digital chaos.

AI on dirty data = incorrect answers at industrial speed.

AI on processes that have not been digitised = automation of manual jobs that should not exist.

AI without architecture = another isolated product added to the list of tools nobody uses.

The order is not optional. It is not a technical detail. It is what separates a system that generates a return from an expense that does not.

Real case: Nemus Arboricultura

Nemus managed its operation with five isolated tools: Excel for quotes, WhatsApp for field work, a calendar for training and two apps nobody used well. Copying and pasting was the main tool.

We applied the three moves:

• Analyse: we mapped the five systems, identified bottlenecks and the processes consuming the most time.
• Optimise: we built a single app (web + mobile) unifying quotes, field work, training and invoicing. Data flows from one process to the next without manual intervention.
• Scale with AI: on the digital architecture, we deployed agents automating quote qualification and work follow-up.

Result: 5 → 1 tools. −5h/week in administration. 0 copying and pasting.

“They are building us an app to unify everything: quotes, field work, training students, invoicing and follow-up. At last we will have control in one place. That peace of mind is priceless.”

— Carolina Ferrer, Coordinator, Nemus Arboricultura

The complete framework

In one sentence: analyse before building, optimise before scaling, AI after clean data.

It is not complex. It is discipline. Most companies fail not for lack of budget but for lack of order. They want the outcome without the process. The process is what makes the outcome last.

If your business has processes that scale badly, start with a free 30-minute discovery call → som-os.dev. We tell you which move you are stuck on and what you need to reach the next.