Taker
- What you do yourself
- Practice: how to request, verify and use the results.
- Where you put the effort
- Adoption, criteria of use and learning of the team.
- What you keep
- Agreed Uses, Licenses and Review of Responses.
KarminaAI StudioAgents · Services · Training
Knowledge · The AI adoption model
Taker·Shaper·Maker
Three ways to bring AI into your organisation.
One decision: what do you really need?
Adopt a tool, adapt it to your context or build your own capacity. The model helps decide how much technology, knowledge, and dedication each process needs.
Karmina application system created by Ignasi Llorente.
It's not a race to Maker. It is a map to find the degree of adaptation that adds value to each case.
How to read the model
It is about more than technology.
It is about the role you take on.
With Taker, the weight is in use. With Shaper, in the adaptation. With Maker, in the construction and operation of its own capacity. Change what the organization prepares, controls and must maintain.
The question is not “how can we get higher?” but “what do we need to make this process work better?” Volume, uniqueness of knowledge, risk and maintenance capacity guide the response.
Adopt
Taker
What can we solve with what already exists?
Value is how you use the tool.
Taker is to adopt an available AI capability and incorporate it into a task. There is no need to develop your own solution: the person provides the context, asks for a result, reviews it and decides how to use it.
The important change is not to buy a license. It’s about moving from improvised use to a shared practice: knowing what to ask for, what information to input, and how to recognize an incorrect answer.
What is involved in the practice
- Select a tool suitable for a specific need.
- Agree on instructions, examples and basic quality criteria.
- Review the answers before incorporating them into the work.
When it makes sense
When the task is limited, a standard solution already gives a useful answer and manual intervention continues to be assumable.
The Limits You Need to Keep in Mind
If each result requires reconstructing the same context, copying data between tools or correcting the same deviations, the process may already need to be adapted.
What should be observed
Preparation time, quality after review and real utility for the team.
The same case, three approaches · Illustrative example
Answer a customer inquiry.
A person prepares a draft response to a consultation with an AI tool. It enters the authorized context, checks the information and sends the message itself.
- Person + context
- Eina existing
- Review and use
An assisted task. The person drives every step.
Adapt
Shaper
What do you need to know and respect the AI to be useful here?
Value is in the context that you incorporate into it.
Shaper is to adapt an existing technology to a specific way of working. Instead of explaining everything in each conversation, the solution has stable instructions, authoritative sources, formats, and review criteria.
This adaptation can include a documentary basis, templates, connections with tools or adjustments of the model when they are justified. Not everything requires training: often the difference is to consult the right knowledge and apply it well.
What is involved in the practice
- Sort the sources and define who has authority.
- Configure instructions, formats and treatment of exceptions.
- Prove behavior with usual cases and difficult cases.
When it makes sense
When the task is repeated and the result depends on the vocabulary, data, conditions or the organization’s own criteria.
The Limits You Need to Keep in Mind
A documentary base is not maintained alone. If the sources are contradictory or nobody assumes the update, the adaptation can reproduce errors with more consistency.
What should be observed
Accuracy with respect to sources, consistency, necessary corrections and updating of knowledge.
The same case, three approaches · Illustrative example
Answer a customer inquiry.
An assistant prepares the answer by consulting the catalog and the current conditions. Respect the agreed tone, indicate the sources and refer to a person the questions he can not solve.
- Query + sources
- Adapted Assistant
- Review with criteria
A contextualized result. Knowledge is reusable.
Build Capacity
Maker
What capacity do we want to govern and evolve?
Value is in the system you can sustain.
In the applied reading of Karmina, Maker means to build an operational capability of its own: a system that connects tasks, knowledge, tools and responsibilities. It can take advantage of existing models; self-development is on the whole.
It’s not enough for a demonstration to work. It is necessary to define what happens when data is missing, a tool fails or the result requires a human decision. Also who can stop the system, review the activity and approve a change.
What is involved in the practice
- Design the complete flow and connect only the necessary tools.
- Establish permits, records, tests and points of human approval.
- Assign responsibility for operation, incidents and improvement.
When it makes sense
When the process has strategic value, there is evidence that the solution works and there are resources to operate it beyond the pilot.
The Limits You Need to Keep in Mind
Build involves maintenance, dependencies and recurring costs. More integration does not have to mean more autonomy: every action needs its limits.
What should be observed
Quality of the complete process, incidences, cost of operation and ability to monitor, stop and recover.
The same case, three approaches · Illustrative example
Answer a customer inquiry.
A system receives the consultation, recovers the context, prepares the response and leaves it pending approval. After validation, you can execute the authorized shipment and record the result.
- Entry + tools
- Flow with approval
- Action and registration
A connected process. Autonomy is explicit and delimited.
DUBTE BACK
It is normal not to know where to start.
There is no need to come up with a definite solution. In a free video call we can understand what concerns you, separate noise emergencies and order a possible first step.
Video call · free · no obligationBook a free video call ↗A short conversation to understand your starting point and bring clarity.
The difference, at a glance.
What changes from one approach to another?
It doesn’t just change the outcome. It also changes where you put the effort and what you need to take care of so that it continues to work.
Shaper
- What you do yourself
- Context: sources, instructions, formats and working rules.
- Where you put the effort
- Quality of knowledge, configuration and testing.
- What you keep
- Current sources, adaptations and evaluation criteria.
Maker
- What you do yourself
- The system: flows, integrations and operation capacity.
- Where you put the effort
- Development, supervision and operational responsibilities.
- What you keep
- Operation, permits, costs, incidents and evolution.
Choosing with criteria
Start with the process.
Not the label.
Before adopting, adapting or building, put four questions on the table. They do not give an automatic score: they help to make the decision explicit.
Does an existing tool already solve it well?
If the answer is yes, Taker may be enough. Check the result on the actual job, including the time you spend reviewing it.
Does self-knowledge change the answer?
If internal information, stable criteria, or specific formats need to be applied, explore Shaper. These sources must first be ordered and in force.
Do you need to connect a whole process?
If the challenge combines various tasks and tools, it values a capability of its own. Define permissions, exceptions, and decisions that continue in people’s hands.
Who will be doing it in six months?
The project does not end with the first delivery. Without a manager, maintenance time and criteria to measure value, it is necessary to reduce the scope.
What does not change
More capability.
The same responsibility.
In any approach it is necessary to know what data can be used, what is considered a good result and who responds when something fails.
Autonomy is not included in the name. A Maker system may require human approval in every relevant action, just as a Taker tool needs a risk-appropriate review.
To finish understanding him.
Questions about
the model.
You need to go through Taker, then Shaper and finally Maker?
It is not a mandatory itinerary or a classification of better and worse. A standard tool can be a definitive solution. It makes sense to change focus when a specific need appears that the current one does not resolve well enough.
Can a company combine the three approaches?
Yeah. The map is most useful when applied to use cases. A team can use a tool to summarize documents, an assistant adapted to consult procedures and a system of its own to coordinate a recurring flow. There is no need to put a single label on the entire organization.
Does Shaper mean to train a model with own data?
Not necessarily. The response can be adapted with instructions, examples and source query, without modifying the model. The adjustment of a model is a technical option that must be justified by evidence, not a requirement to begin with.
What does Maker mean in the original frame and on this page?
In the technical formulation of reference, Maker describes the construction of foundational models of its own. Here we explain a reading applied to the organization: build and govern a system of its own, also on existing models. Developing an agent or an integration is not, in itself, equivalent to creating a foundational model.
Does having an AI agent make us Maker?
No. An agent can be a standard product we adopt or a solution we adapt to our knowledge. In this applied reading, the step to Maker has to do with assuming the design, operation and evolution of a system of its own, not with putting the name of agent.
How do we know it’s worth changing focus?
It is necessary to compare the result with the current way of working: quality, review time, errors, total cost and maintenance load. A useful pilot includes exceptions and stop conditions. If adaptation or development does not improve the balance, keeping the option simpler is a good decision.
Frame of reference: McKinsey · A CIO and CTO guide. Examples and operational capability-oriented reading are an explanatory adaptation of Karmina, not customer cases.
Let’s talk
What approach does your process need?
Explain to us what you want to solve, what tools you use and where you find the limit. We will help you assess whether you need to adopt, adapt or build.
hello@karmina.ai