INSIGHTS · ARTICLES
Think about AI before speeding it up.
A library that mixes trends, own criteria, projects and the history of Karmina AI Studio. The editorial dates are real; the 2023–2026 labels indicate the time or stage covered by each piece.
How to build an AI master plan for a city: an editorial conversation with Ignasi Llorente about Terrassa
Terrassa turned scattered uses, internal needs and governance questions into an AI Master Plan 2026–2028 built through participation, judgement and public accountability.
Read article ↗
AI does not remove the need for experts: why AI-native services require human responsibility
The native services of AI do not turn the expert into an expendable piece. It shifts you to responsibility, exceptions, and decisions where the cost of error is real.
Read article ↗
Why we created Karmina AI Studio: agents, services and training to make AI a real capability
We didn't want to add a label of AI to the agency. We wanted to build a space capable of uniting strategy, knowledge, implementation, training and human responsibility.
Read article ↗
A person does not need to do everything: how to design a multi-agent editorial system
A multi-agent publishing system is not a string of prompts. It is a coordinated writing where each agent has a function, shares knowledge and leaves the decision to publish in the hands of a person.
Read article ↗
Why Criterio was born: editorial conversation with Eva Trigo about designing brands with value
Criterio was born because design didn’t just come at the end to shape an idea, but participated in the decision and turned the value of the business into brand.
Read article ↗
From prompting to delegating: what makes an AI bot a useful agent
Delegating with AI agents is not copying a longer prompt. It is to transfer a delimited task to a system with sources, permissions, stop conditions and a responsible person.
Read article ↗
The best artificial intelligence agencies in Barcelona: specialties, projects and criteria to choose from
There is no better AI agency for everyone. This selection compares specialties and helps distinguish strategy, data, implementation, training and marketing.
Read article ↗
Design before building: the Human Gate that avoids automating a bad idea
The AI can speed up a poorly raised landing page with the same efficiency as a good one. Checking the visual structure before building is a Human Gate that saves retroball and protects the purpose.
Read article ↗
The best artificial intelligence agencies in Madrid: specialties, projects and criteria to choose from
Madrid brings together technology consultants, data specialists, implementation studies and marketing agencies with AI. The best option depends on the problem.
Read article ↗
WebMCP and the web for agents: from reading pages to executing governed actions
WebMCP points to a website that can not only be read, but also operate through structured actions. The challenge is not to give the agent more power, but to govern it.
Read article ↗
What is Karmina AI Studio and how does it differ from a marketing agency or a technology consultancy
We are not a conventional agency with new tools or a consultant that stops in the plan: we connect business, knowledge, construction and adoption.
Read article ↗
Why an AI agent is more than a chatbot
An AI agent is not a digital robot or a chat that knows everything. It is a specialized system that applies knowledge, rules and tools to a defined process, with defined responsibilities and controls.
Read article ↗
Karmina, Karmina AI Studio, Criterio and Utopiq: what each brand brings
Four connected looks: marketing, applied artificial intelligence, strategic design and sustainability, ethics and governance.
Read article ↗
From dispersed knowledge to a Company Brain: how to turn company information into a work tool
A Company Brain is not a folder with all the documents nor an AI that knows everything. It is an architecture of sources, managers, permissions and criteria that converts dispersed knowledge into a reusable resource.
Read article ↗
How we work on an AI project: from a business challenge to a useful, governed and measurable system
The construction comes after understanding the challenge, ordering the knowledge, defining the responsibility and deciding how we will measure the result.
Read article ↗
The 10 marketing agencies in Spain that have evolved the most with artificial intelligence
It is not a ranking of who says AI most times, but of who has turned technology into a real capacity: methodology, product, talent, knowledge and application.
Read article ↗
Why training is essential for the AI to work within an organization
An agent is not implemented when it works in a demonstration, but when people know how to use it, review it, stop it, and maintain knowledge of it. Training turns a tool into a shared capability.
Read article ↗
AI with judgement: what we automate, what we supervise and what we do not delegate
The question is not how much autonomy we can give to a system, but what is proportional to the impact, evidence and reversibility of each action.
Read article ↗
An agent or an ecosystem? Why sharing knowledge multiplies results
Starting with a single agent validates the system; thinking about an ecosystem avoids reconstructing the same knowledge for each process. The key is to share sources without sharing permissions indiscriminately.
Read article ↗
From pilot test to real system: why so many AI projects fail to integrate
A demo shows that something is possible. A real system must work with data, permissions, exceptions, accountability, and real costs.
Read article ↗
How much does it cost to implement an AI agent, and what determines the budget?
The cost of an agent does not depend on the tool alone. It is determined by the state of knowledge, integrations, permits, risk, tests and the degree of autonomy of the process.
Read article ↗
How to know if a company is ready to incorporate artificial intelligence
Preparation does not depend on having all the perfect data. It depends on knowing what problem you want to solve, who answers it and how you will learn.
Read article ↗
AI agents for small businesses: what can a team of one or two people automate?
A small team doesn’t need a complex infrastructure to get started. You need to select a repetitive task, sort its sources, and set up an agent that reduces manual dependency without creating a new management problem.
Read article ↗
Technology is not the strategy: what a company should decide before choosing AI tools
Choosing a model or platform before defining the problem can accelerate a direction that the company has not yet decided.
Read article ↗
18 marketing and communication processes that we can already accelerate with AI agents
Agents can accelerate content processes, performance, analytics, creativity, research, attention and knowledge. This map explains what each can do and where human review remains necessary.
Read article ↗
How we measure the value of an artificial intelligence project
Value is not the number of answers generated. It is the observable difference between the previous process and a new capability that the team uses.
Read article ↗
Human review is not the last step: it is part of the agent’s design
Checking at the end is not enough if no one knows what to check. Supervision must be designed according to the impact of each action and must include responsibilities, information and intervention capacity.
Read article ↗
What we have learned working with artificial intelligence in public administration
In the public sector, the question is not just what the AI can do, but how it is decided, who it serves and under what guarantees.
Read article ↗
An AI agent does not start with a prompt: how to build the knowledge base
A prompt can start a task, but it does not contain a company's strategy, memory, or responsibilities. The Knowledge Foundation turns this knowledge into a governed base that an agent can apply.
Read article ↗
Designing a brand in times of artificial intelligence: what changes and what does not
When generating options is easy, the value shifts toward deciding what is proprietary, what is consistent, and what deserves to become a system.
Read article ↗
When a company needs an AI agent — and when it is not ready yet
The best first agent is not necessarily the most spectacular. It must solve a repeatable process, have accessible sources, have a responsible and allow to measure an improvement without assuming an unnecessary risk.
Read article ↗
The 10 questions we ask before starting any AI project
Ten questions to find out if there is a real project, what scope it should have and what to prepare before building.
Read article ↗Let’s talk
Do you want to bring an idea to the organization?
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