AI process automation
Repetitive work, done on its own.
In short
AI process automation means the repetitive work —copying data, answering the same questions, chasing follow-ups— runs on its own. We build it with n8n, Make or Zapier, scheduled jobs (cron) and Claude or OpenAI where judgement is needed, connected to your CRM and tools. We start with the process that loses the most money; the rest is rules, cheaper and more predictable.
Watch it in a minute
Narrated in Spanish.
What you get
- Hours a week back for the team
- Data that moves between tools without copy and paste
- Fewer mistakes and no forgotten tasks
How it works
Illustrative example01 Comes in
An email with an order, an invoice or a question.
02 AI reads it
It extracts the data, understands what it is and classifies it.
03 Recorded
The data lands in the CRM or your tool, with no typing.
04 Done
Whoever needs to act gets the task, with a due date and a reminder.
The problem
Your team loses hours copying data, answering the same questions and chasing tasks a machine would do without mistakes.
Who it is for
Companies that want to automate the manual processes they repeat every week with AI.
What is included
- Process map and what to automate first
- AI workflows in GoHighLevel, HubSpot or n8n
- AI-generated documents, emails and proposals
- Automatic alerts and tasks for the team
A good fit if…
- The task happens many times a week and follows rules you could explain to a new hire in five minutes.
- It eats hours you can count and its failure costs money: an unanswered lead, an unsent invoice, an unconfirmed appointment.
- You want everything centralised in a CRM, whether GoHighLevel, HubSpot or another.
Not a fit if…
- The process happens twice a month: building and maintaining it costs more than it saves.
- The rules change depending on who does it: it needs defining before it can be automated.
How we do it
- 01
Centralise
Everything lands in one place before anything is automated. If contacts live in four inboxes, no automation can cope.
- 02
Automate the obvious
Create the record, acknowledge, assign and remind. No AI: this alone removes most of the leakage.
- 03
Add judgement
AI to read free text, classify requests and draft replies, on top of a process that already works and is measured.
- 04
Measure and expand
Compare volume, time and leakage with the baseline, then move to the next process. Never two at once at the start.
Timeline
A well-scoped first process goes live in 4 to 6 weeks, including the definition phase, which is usually the slowest.
Investment
- Discovery
- €1,500 – €15,000
- Implementation
- €5,000 – €25,000
- Monthly retainer
- €1,000 – €5,000/mo
The exact price is confirmed in the free consultation, not before.
Monthly maintenance
An agent or an automation does not stand still. Your tool changes its default AI model, a new version breaks an integration or a provider suffers a security breach. With the monthly plan, every change is tested before it reaches your customers, and you always know what changed.
Tests before every model change
Before switching model, version or instructions, we run a set of real conversations and cases from your business. If the result gets worse, the change does not ship.
Cost per correct answer
Every month you see what the system spent on AI and how many answers it got right. What counts is the cost of each correct answer, not the price per token.
A log of what changes in the CRM
Every change to workflows, fields, agent instructions or model is logged with date and reason. You know what changed, and it can be undone.
Keys and providers
We review who has access, rotate keys when due and follow each provider's security notices: the AI model, WhatsApp, the CRM and the automation platform.
This is the monthly retainer in the investment table: €1,000 to €5,000 a month, depending on how many workflows and agents you have in production. The exact price is confirmed in the free consultation.
How to measure and cap an agent's cost (in Spanish)Real cases
Events · Barcelona
Eventos Barcelona
Proposal automation
−85%client response time
Challenge
Each proposal took 1 to 3 days to go out. By the time it landed, the client had cooled off or already asked a competitor for a quote.
Solution
We created an intake form filled right after the call, which generates the proposal automatically in GoHighLevel, ready to review and send in minutes, with follow-up reminders scheduled in.
Read the full case studyCase video availableMarketing · AI agents
Tribeca Media
Performance and content AI agents
0campaign actions without human approval
Challenge
Their KEYWORKERS platform had an agent that read Meta Ads campaigns but did not act on them: no Google Ads, no creatives and no way to know how much it spent on AI or when it failed.
Solution
We built the Performance Agent, which reviews Meta and Google Ads, proposes pausing, cutting budget or creating campaigns, executes only with human approval, and generates new creatives when an ad fatigues. We are now building the Content Agent.
Read the full case studyCase video available
Frequently asked questions
Which process should I automate first?
The one that meets four conditions at once: it happens often, follows stable rules, takes measurable time and its failure has consequences. For most small and mid-sized businesses it is the stretch between a contact arriving and someone acting on it.
Does AI do everything?
No. Most of an automation is rules: when a form arrives, create the contact, assign it and schedule the reminder. AI comes in where the input is free text, such as a WhatsApp message or a long email. Using it where a rule would do is more expensive, slower and less predictable.
How do I know if it pays off?
Multiply how often it happens per week, how many minutes it takes and the hourly cost of whoever does it: that is the cost of keeping it manual. Then add what is lost. In sales processes without a system, leakage is usually 30% to 60% of contacts, and it almost always outweighs the hours.
How much does it cost?
It depends on scope: discovery starts at €1,500, a full build runs from €5,000 to €25,000, and there is an optional monthly maintenance retainer. The exact price is confirmed after the free consultation, not before.
Related guides (in Spanish)
Automatización de procesos con IA: qué es y por dónde empezar
Qué es la automatización de procesos con IA, qué procesos se pueden automatizar en una pyme, cuál elegir primero, cuánto tarda y cuánto cuesta.
Cómo implementar IA en tu empresa: plan en cinco pasos
Cómo implementar IA en una empresa en cinco pasos: dónde empezar, qué datos hacen falta, comprar o construir, que el equipo lo use y qué exige la ley.
12 ejemplos de automatización de procesos en una pyme, con IA
Ejemplos de automatización de procesos en una pyme, ordenados por departamento: qué hace la automatización, dónde entra la IA y cuándo no merece la pena.
Herramientas de IA para automatizar procesos: cuáles usar en una pyme
Las cinco familias de herramientas para automatizar procesos con IA, qué resuelve cada una, cuánto se puede hacer sin programar y cuándo hace falta algo a medida.
Proyectos de IA a medida para empresas: ejemplos, precio y plazos
Qué es un proyecto de IA a medida, cinco ejemplos reales en empresas españolas, cuánto cuesta y tarda cada fase y cómo evitar que se quede en un piloto.
Cuánto cuesta automatizar un proceso con IA
Qué mueve el precio de verdad, cómo calcular si compensa antes de gastar nada y por qué los presupuestos muy baratos salen caros.
RGPD y automatización con IA: qué puedes automatizar
Consentimiento, decisiones automatizadas y datos que salen de la UE. Lo que la ley permite, lo que exige probar y los fallos más comunes en España.
n8n 3.0: qué se rompe y cómo prepararte antes de diciembre
Qué cambia en n8n 3.0, qué flujos dejan de funcionar y un checklist para revisar tu instancia antes de actualizar: nodos retirados, Docker, Slack y tiempos.
