KarminaAI StudioAgents · Services · Training

Manifest · Karmina AI Studio

Artificial intelligence is more than a tool.

It is a new way of working, deciding and creating. This manifesto gathers the principles with which we design agents, services and training so that the technology expands capabilities without diluting the criterion.

01

Technology contrasts.
Knowledge is oriented.
People decide.

Where he was born

An AI studio born within an agency with years of experience working with people, brands and change.

Karmina AI Studio is born within Karmina to connect artificial intelligence with the real knowledge of organizations. Let’s not separate the technology from the strategy, nor the automation from the culture that will make it work.

That’s why we start with processes, sources, criteria and people. Then we decide if the right step is to train, order, automate, build an agent or simply not add more technology.

Karmina team working together in a hybrid session
People before interfaces.Barcelona · Madrid · Online-first

Five Principles

The manifesto is a way of making decisions within the project.

The principles are translated into concrete questions: who answers, what sources have authority, what permissions the system needs, what to review and how we will know if it adds value.

01

Technological Humanism

Technology has broad capabilities. It does not automatically become authority.

Each system has a mission, a responsible person and limits proportional to its impact. Sensitive decisions remain in the hands of those who can understand the context, take the consequences and stop the process.

In practice

  • Identifiable human responsibility
  • Review before sensitive actions
  • Actual ability to correct, ignore, or stop
Meet the people behind
02

Collective growth

The AI should grow the team’s capacity, not concentrate it in a black box.

We document knowledge, explain decisions and transfer capacity. The objective is not that an organization depends forever on a supplier, but that it can understand, govern and evolve what it incorporates.

In practice

  • Training linked to the project
  • Sources with responsibility and validity
  • Reusable knowledge without erasing authorship
See how we build teams
03

Curiosity with limits

Experimenting is not improvising with data, people, or critical processes.

We provide a hypothesis in a small enough environment to understand what is going on. Let’s agree before what I would mean that works, observe the errors and decide with evidence if it is convenient to scale, redesign or close.

In practice

  • Delimited and reversible pilots
  • Success criteria before testing
  • Errors, changes and documented learning
Level of adoption
04

Transparency and trust

A convincing answer is not the same as a reliable result.

We make visible when the AI intervenes, what sources it uses, what is given, what is inference and what still needs validation. Trust does not come from hiding complexity, but from being able to inspect it and act when something does not fit.

In practice

  • Fonts, permissions and visible versions
  • Expressed Limits and Uncertainty
  • Human Gates before publishing or acting
Explore strategy and governance
05

Comprehensive sustainability

More speed doesn’t always mean more value.

We value maintenance, technological dependence, review cost, safety, resource consumption and impact on team. We do not automate a process that should be simplified before or build more infrastructure than necessary.

In practice

  • Complexity proportional to the problem
  • Measured value and incidences
  • Right to maintain, change or disconnect the system
See how it takes shape in projects

From Principle to System

An agent does not start with a prompt.

It begins with a process that is worth improving, knowledge that can be authorized and people able to review the result.

01

Understanding Before Proposing

We listen to the team, observe the friction and decide if an agent, an automation, a training or a simpler solution is necessary.

02

Order before connecting

We build the Knowledge Foundation with sources, vocabulary, managers, permissions, examples and quality criteria.

Enter Knowledge Foundation ↗
03

Designing the supervision

Human Gatess are part of the flow: before publishing, sending, modifying, sharing data, affecting money or making sensitive decisions.

04

Measuring before climbing

We compare time, quality, errors, adoption and maintenance cost. If the system does not create net value, we do not convert it into infrastructure.

Autonomy with responsibility

Just because something can be automated does not mean it should be.

An agent can
  • Consult authorized sources.
  • Apply agreed rules and formats.
  • Prepare drafts, classifications and reports.
  • Perform reversible and low-risk steps.
  • Climb doubts, conflicts and exceptions.
An agent must not
  • Invent the criteria that the organization has not defined.
  • Disguise uncertainty or lack of evidence.
  • Extend permissions on your own.
  • Infer sensitive data outside the mission.
  • Publish, contact or decide without the intended authorization.

Supervision is effective when a person can see the sources, understand the errors, correct them and give naturalness to the result.

An idea that becomes work

Agents, services and training. Three starting points, one shared judgement.

Karmina team collaborating in person and remotely

The company we are building

A new Karmina capability, beyond a catalogue of tools.

Karmina AI Studio combines the agency’s marketing, communication, creativity and business knowledge with AI strategy, knowledge systems, automation, training and governance.

This allows us to start with a specific need — a repetitive task, a poorly informed decision, a team that needs guidance — and build only what is needed to make the change usable and sustainable.

We want the AI to enter organizations with context, accountability and a utility that can be demonstrated.

Continue exploring

From the manifesto to a specific decision.

Let’s talk

What decision of AI do you need to make with more criteria?

Tell us about your situation, even if the challenge is not fully defined. A member of our team will respond to understand the process and discuss a first step.

hello@karmina.ai