Ethics in AI projects does not have to be an insurmountable block: it can grow with the project, step by step, and become a competitive advantage.
In the race to adopt Artificial Intelligence, it is common to view the ethical dimension as merely a set of restrictions or compliance barriers. Here at Taqtile, we see it differently: we believe that ethical responsibility is not a brake, but a competitive advantage capable of protecting the brand's reputation and generating new value opportunities.
We know the topic is complex and often difficult to make tangible in the day-to-day of projects. However, ignoring this layer is not an option. Ethical care is, ultimately, care for the longevity of the business. Even though regulatory and social demands are still maturing, the brands that get ahead build much more solid foundations for scaling their innovations.
A common tension in innovation and Artificial Intelligence projects is that ethical challenges, by being vast and complex, risk becoming an "elephant in the room". Everyone knows it is there, taking up space, but the difficulty of addressing all its ramifications often leads to postponement or even the problem being forgotten altogether.
And in part this makes sense, because every innovation project faces many barriers, and a barrier of this magnitude can be decisive for whether a project continues. This is where one of Taqtile's key learnings about the ethical challenge comes from: the complexity of the topic must not overshadow the scale of the initiative being discussed. It needs to grow as the project advances.
As will be explored below through conversations with Tuanny Martins and Danilo Toledo, who lead these discussions at Taqtile, between halting a project and avoiding the subject altogether, the path is to work through it incrementally.
The goal of this piece is therefore to share what we have learned at Taqtile about how to prevent the topic from being set aside. The key is not having all the answers and commitments on day one, but turning ethics into a manageable process, one that is addressed and matures at the same pace as the project advances.
The importance of making it a subject
Following the development of this theme reminded me of when I first encountered the research of Sheila Jasanoff, an Indo-American social scientist who is a leading reference in the field of Science, Technology and Society (STS) studies. Studying the impact of new technologies on society, the author became known for demonstrating a gap in how we think about society: the failure to account for how much technologies affect the production of social orders.
In her studies, Jasanoff introduces the concept of the sociotechnical imaginary, referring to collectively held and publicly performed visions of desirable futures as something achievable through scientific and technological advances. In this scenario, the author shows, we tend to look at innovation only through the lens of the solution, overlooking how it affects and produces new social orders.
In this sense, the author argues that the greatest risk arises not from the technology itself, but primarily when aspects of responsibility and ethical challenges are not part of these sociotechnical imaginaries of technology.
Not coincidentally, this is one of Taqtile's central concerns: beyond seeking notions of compliance, studying the topic with the aim of making it manageable and possible to address in the corporate environment. To share what we have already learned through internal discussions, we have consolidated three strategies for bringing the conversation into the corporate context.
1. The topic must keep pace with the project
How, then, to bring the subject into practice without stalling innovation? The first strategy suggested by Danilo is the incremental approach.
"Ethics cannot be an afterthought. It needs to be present at the starting point, and evolve in complexity alongside the project." For this reason, the role of consulting is to raise awareness of broad risks from the outset, but address tensions proportionally to the stage of development. This prevents ethics from becoming a sensitive, insurmountable block, allowing it to be addressed step by step, keeping pace with the growing complexity and impact of the solution.
That is what we did in the AI agent that runs car subscription sales: the agent's autonomy grew in stages, as it proved it was getting things right, and sensitive decisions, such as credit analysis, stayed with a person.
2. Approach it through the lens of efficiency: the cost of not doing it
Tuanny also points to a pragmatic dimension worth addressing. Treating ethics from the beginning is a matter of efficiency.
For example, the cost of remediating a poorly built AI model is exponentially greater than developing it with responsible criteria from the start. "AI models learn and amplify patterns. If data bias is not addressed at the source, it becomes embedded in the architecture of the system. Correcting this afterwards can require anything from costly retraining to discarding the model entirely. Furthermore, technical aspects such as explainability, the system's capacity to detail how it reached a conclusion, often cannot be added in after the fact."
Beyond that, over the long term, thinking about compliance and responsibility from "day zero" is about protecting the investment, especially in a market that is set to become increasingly regulated. Addressing the ethical aspect of the technology helps anticipate where regulation will arrive first, allowing companies to get ahead.
3. Ethics is also about generating value
Something Tuanny highlights is that "it is fundamental to shift the perspective: move from a defensive posture ('block AI to avoid harm') to a proactive one. Ethics does not just mitigate risks; it generates value opportunities."
A practical example she brings illustrates this point: "while developing an internal prototype, we noticed that the AI was using predominantly masculine language and examples, reflecting the bias in its training data. By identifying this, it was possible to adjust the model to adopt neutral, inclusive language with a precision and scale that would have been difficult to achieve manually. Here, the ethical lens was not a brake, but a lever to deliver a technically superior and socially more appropriate product."
Looking at the proactive side of what Artificial Intelligence can deliver, therefore, is not only about its desirable (and distant) futures. It is also about understanding how the technology's capabilities can solve real problems today.
How Taqtile is going deeper on the challenge
At Taqtile, as in the broader market, the complexity of the topic is still being mapped and elaborated. Building proprietary tools, the focus is not only on raising awareness, but on bringing the subject into the practical reality of companies and projects.
Out of this front came Responsible AI, Taqtile's AI responsibility assessment. It evaluates an AI project or product, in production or still in development, across 10 dimensions: identification, data and privacy, bias and equity, transparency, labor displacement, quality, governance, sustainability, maintenance and security.
Each dimension is covered by questions classified by criticality. Some are non-negotiable and, left unanswered, cap the project's score. Others need to be resolved before scaling, and others come in progressively. This classification is what makes the subject manageable. The result is not a pass or fail, but a maturity map by dimension, the project's regulatory exposure and a roadmap of practices, in the order in which they are worth adopting.
If you want to understand how Responsible AI applies to your project, write to us. We run the assessment together with your team, at whatever stage the project is in.