Architecture to scale
AI for business

AI for business that works inside your systems, not beside them

We implement AI connected to your ERP, your CRM, your WhatsApp and your data: agents that read orders, enter invoices, answer questions and build reports. You decide what a person approves.

Tell us which task to automate Take the self-assessment
  • Your data does not train third-party models
  • A person approves what matters
  • Monthly cost with a cap
One order, end to end Example
1It arrives on WhatsApp

Hi, send me 40 boxes of item 2231 for Thursday, to the usual address.

2AI reads it and checks your ERP
Customer
Recognized by phone number
Product
Item 2231 · 40 boxes
Stock
Available
Delivery
Thursday · usual address
3A person approves
Order entered in the ERPConfirmation sent to the customer on WhatsApp.
Sent back to the sales inboxThe correction is logged and the AI uses it next time.
The problem

AI is already in your company. It just isn’t useful yet.

Almost every team already uses it. But on its own, without real data and without anyone connecting it to the systems where the work happens.

ChatGPT on the loose

Everyone uses it their own way, with personal accounts and customer data pasted into the chat. Nobody knows what left the company.

Pilots that never ship

An impressive demo that never connects to the ERP or the CRM. It stays in a separate tab and people go back to copy and paste.

Made-up answers

Without access to your data, AI fills the gaps with whatever sounds right. In a quote, a price or an invoice, that gets expensive.

Costs nobody watches

Scattered subscriptions and pay-per-use with no cap. At the end of the month a bill arrives that nobody can explain.

What changes everything: AI that reads your real data, works with permissions and logs everything it does.

What it can do

AI process automation, team by team

We start with work that is done by hand today, repeats every day and follows clear rules.

Finance · Invoices, collections and reconciliation

  • Reads supplier invoices from PDFs or email and enters them in the ERP, with their accounts.
  • Reconciles the bank and flags only the transactions that don’t match, for someone to review.
  • Prepares payment reminders with each customer’s actual balance.
Connects to
OdooYour ERPEmailBank statements
A person decides

Which invoice gets paid and which difference gets adjusted. AI proposes, accounting confirms.

How it is measured

Data-entry hours per month and invoices that needed fixing.

Sales · Quotes and CRM follow-up

  • Drafts the first version of the quote with your price list and each customer’s terms.
  • Summarizes every conversation and saves it to the CRM record, with the next task scheduled.
  • Flags deals going cold before they are lost.
Connects to
CRMOdooWhatsAppPrice list
A person decides

The final price and any discount outside the list.

How it is measured

Time from inquiry to quote sent.

Customer service · Questions on WhatsApp and the website

  • Answers what is already written down in your manuals, policies and FAQs.
  • Checks an order’s status directly in your system, without anyone looking it up.
  • Hands off to a person when the question is sensitive or has no answer, with the summary ready.
Connects to
WhatsApp BusinessWeb formERPKnowledge base
A person decides

Complaints, returns and anything not in the knowledge base.

How it is measured

Questions solved without hand-off and time to first reply.

Operations · Orders, stock and logistics

  • Enters orders that arrive by message or email, and spots the ones with missing data.
  • Warns about stock shortages based on sales, open orders and restocking times.
  • Builds the day’s delivery route from the confirmed orders.
Connects to
ERPWarehouseWhatsAppSpreadsheets
A person decides

Purchasing, priority changes and any exception of the day.

How it is measured

Orders entered without errors and stockouts avoided.

Management · Ask your company

  • Answers questions in plain language about sales, margins or collections, with the numbers from your systems.
  • Builds the weekly report with what changed and why, without waiting for month end.
  • Flags deviations when a metric moves out of its normal range.
Connects to
ERPCRMBI dashboardsSpreadsheets
A person decides

The decisions. AI shows where every number comes from so it can be checked.

How it is measured

Hours spent on reports and questions nobody can answer quickly today.

AI process automation: what your company can automate today →

AI in everything we do

Not one more service: a layer over all our solutions

AI pays off when it rests on good architecture, organized data and well-maintained systems. That is why we add it to everything we do.

Start here

Technology architecture with AI

We decide where AI goes and where it doesn’t: which model to use, how it reaches your data, with what permissions, what gets logged and how much it can cost per month. Before connecting anything.

See technology architecture →
  1. Your data ERP, CRM, files
  2. Permissions only what it needs
  3. Model cloud or local
  4. Approval a person decides
  5. Log what it did and why

Custom software

AI features inside your system: read documents, classify, suggest and search in plain language.

See custom software →

Odoo and open source

Agents that work in Odoo like one more user, with that user’s permissions: they enter orders, invoices and tasks.

See Odoo and open source →

Data analytics and BI

Plain-language questions about your numbers, with the answer and where each figure comes from.

See data analytics and BI →

Mobile apps

Dictate a report, take a photo of a delivery note and have it entered in the system.

See mobile apps →

Managed hosting

Local AI on your own or a dedicated server, so sensitive data never leaves the company.

See managed hosting →

Technology assessment

The starting point: which tasks to automate first, which not to, and what needs sorting out before.

See the assessment →
How we work

How to implement AI in a business: one task, measured

No giant projects up front. We test one real task, with a measure agreed before we start, and continue only if it works.

  1. Assessment

    We review your processes, systems and data, and pick the task where AI saves the most with the least risk.

    You getThe ranked task list and what needs sorting out first.

  2. Measurable pilot

    AI does that task with your real data, alongside your team. We measure against what was agreed at the start.

    You getReal numbers: time saved, errors and monthly cost.

  3. Integration

    If the pilot works, we connect it properly to your ERP, CRM or WhatsApp, with permissions, approvals and logging.

    You getThe task running inside your systems, not beside them.

  4. Follow-up

    We review quality and costs, fix what fails and add the next task when it makes sense.

    You getA dashboard showing what the AI does and what it costs.
Our rules

Four things we don’t negotiate

Your data is yours

We use services that don’t train their models on your information, or models that run on your server.

A person approves

Anything that moves money, commits the company or reaches a customer goes through someone on your team.

Capped cost

Each task has a monthly budget and a dashboard showing how much was spent and on what.

It helps, it doesn’t replace

It takes the repetitive work off your team so they can spend their time on what needs judgment.

Private AI or cloud: where does it run?

It depends on what data it touches and how much you will use it. We decide it in the assessment, task by task.

In the cloud

Models from the major providers, on business plans.

  • The most capable models
  • Fast to start, no hardware to buy
  • Pay per use, with a monthly cap
Best whentasks don’t touch sensitive data or volume is low.

On your server

Data stays in

Open models installed on your own or a dedicated server that we manage.

  • Data never leaves the company
  • Fixed cost, no usage surprises
  • Smaller models, chosen per task
Best whenyou handle health, legal, financial or customer data that cannot leave.

Hybrid

Sensitive work runs on your server and the rest goes to the cloud.

  • Each task where it fits best
  • Less data exposed
  • Balanced cost
Best whenyou have both kinds of tasks, which is the most common case.
How we use it ourselves

We tried it in-house first

We built a regulatory compliance platform with AI agents working in parallel. What we learned is what we now apply with every client:

  1. Specify before you automate. Writing down what the system must do saves weeks later.
  2. Green tests are not enough. An agent walks through the app like a person and finds what the tests miss.
  3. Every fixed bug stays watched. It becomes an automated test so it doesn’t come back.
  • 4AI agents working in parallel
  • 2days writing what it had to do, before a line of code
  • 2weeks to the first demo
  • 1,800+automated tests passing

Pick one task. We test it with your data.

Tell us what is done by hand in your company. We reply with:

  • Whether AI is the right tool or a simpler automation is enough.
  • What it would take in your systems and data to make it work.
  • What the pilot would look like, with the success measure and a capped monthly cost.
Tell us the task on WhatsApp Take the self-assessment

No commitment · Remote, Spain and Latin America · Your data stays put unless you say so

FAQ

What people ask us about AI

If your question isn’t here, message us and we’ll tell you how we would do it in your company.

Where should a company start implementing AI?

With a task that is done by hand today, repeats every day and follows clear rules: entering invoices, reading orders, answering common questions. In the assessment we pick the one that saves the most with the least risk and test it against a measure agreed up front.

Will AI replace my team?

That is not the goal. It takes the repetitive work off them so they can spend time on what needs judgment. The decisions that matter stay with a person.

Is my data safe?

We use business services that don’t train their models on your information, or models installed on your own server. AI only reaches what each task needs, with permissions, and every action is logged.

How is this different from using ChatGPT?

ChatGPT answers in a separate tab with whatever you paste into it. What we build works inside your systems: it reads your real data, enters information in your ERP or CRM and logs what it did.

Does it work with Odoo, my ERP or a custom system?

Yes. With Odoo and most ERPs and CRMs it connects through their API. If your system doesn’t have one, we build it or work on the database with the right permissions.

What is an AI agent and what is it good for in a business?

It is an AI that, besides answering, can do things: query a system, enter data, send a message. That is why we set it up with narrow permissions and human approval at the important steps.

Can we have private AI, without sending data out?

Yes. We install open models on your own or a dedicated server that we manage. They are smaller than cloud models, so we choose the model per task. Many companies combine both options.

How much does it cost to implement AI in a business?

It depends on the task, the volume and whether it runs in the cloud or on your server. After the assessment we give you the pilot cost and an estimate of the monthly cost, with a cap that isn’t exceeded.

What happens if the AI makes a mistake?

That is why there is human approval where it matters and a log of every action. Corrections are used to adjust the rules, and we track errors as part of the follow-up.

Do you work with companies in Spain and Latin America?

Yes. We work remotely with teams in Spain and across Latin America, in their time zone and following their data protection rules.