Act 2: Moving Pictures

It’s been three years with GenAI in our lives.

You may think we understand what AI can or cannot do (at least those in the Global North think they do). But think again. There are entire regions of the world that are playing catch-up to the AI-first reality. And there are always newcomers to our industry who need guidance to set them on the right path.

Ultimately, progress will look different to every team and professional. It won’t be linear, and it may come at a cost. It may require us to redefine ourselves. In our quest onward, no one should be left behind, right?

In this section, read the articles that explore the human side of the (r)evolution unfolding around us.

Agustín Da Fieno Delucchi

Industry Expert and Consultant

Agustín Da Fieno Delucchi is an expert in data, AI, and localization, with decades of experience in driving global technology initiatives. A consultant, frequent speaker, and panelist at major conferences and professional podcasts, he is recognized for his thought leadership in the evolving localization and technology landscape.

In this new era of generative AI, the question is no longer whether machines can produce content across languages and modalities; it is whether they should, and under what conditions. The floodgates have opened. Multilingual outputs now flow from prompts at breathtaking scale. But as this (r)evolution accelerates, we must stop and ask: What scaffolding are we building around these outputs to ensure they mean something? To someone? Somewhere?

The excitement around GenAI in localization has often been about efficiency – how fast we can go, how many markets we can cover. But speed without quality is just multilingual noise. The deeper evolution now underway is not one of scale, but of sense-making.

Quality is not a retrospective

For years, quality assessment came at the tail end of the content lifecycle. Reviewers would flag mistranslations, misused terms, or culturally jarring expressions long after damage had already been done. But generative AI forces reframing. When there is no “source” text to anchor a review, what becomes the reference? The answer: quality must be designed into the system, before generation ever happens.

That is where semantic intent modeling and structured content governance come in. Instead of relying on post-hoc evaluation, GenAI workflows are now incorporating semantic blueprints; data-rich templates that encode the “why” of a message. Tone preferences, brand voice, factual claims, taboo phrases, cultural framing – all of these become structured signals in a semantic map that informs prompt creation.

That is precisely what frameworks like DirectGC™ are bringing to the table: a systematic, upstream approach to content generation that fuses intent, compliance, and context before a single word is generated. These frameworks do not eliminate the need for quality checks; they raise the floor from which those checks begin.

This preparation phase is not glamorous. It does not make headlines. But it is here that the most meaningful shifts in our industry are happening. Because the quality of generative outputs is not just a reflection of the model; it reflects the preconditions we set.

Measuring what matters

If preflight quality governance is the first half of the story, the second half is validation. But validating what, exactly?

Traditional localization QA focused heavily on linguistic correctness and term adherence. But those standards fall short in a generative, multimodal context. That’s why some in the field are turning to MMQEval, a multidimensional framework that evaluates content not only for accuracy and consistency, but also for resonance, engagement, and ethical grounding.

Unlike legacy QA methods, MMQEval does not treat quality as a single metric. It breaks it down across four axes:

• Normative Integrity: Does the output meet the brief and align with local norms?
• Factual & Ethical Integrity: Is the information correct, and is it safe?
• Communication Coherence: Does it sound natural and tonally appropriate?
• Experience & Engagement: Does it connect emotionally, visually, contextually?

Each axis is scored, producing a nuanced profile of quality. And importantly, it allows systems to learn over time. QA is no longer just a gatekeeper; it becomes a feedback loop. If a term keeps falling flat in a market, the terminology cluster gets updated. If an emotional tone misses the mark, the cultural vector gets refined. Quality, then, becomes a living system, not a checklist.

When quality drives the experience: A retail example

Imagine a multilingual retail campaign promoting a new eco-conscious product line across Japan, France, Brazil, and the United States. The messaging emphasizes ethical sourcing, carbon footprint reduction, and the emotional reward of sustainable living.

In a traditional localization flow, the English source copy, crafted with a U.S. audience in mind, becomes the blueprint. Translations attempt to adapt tone and claims downstream. But in Japan, the emotional appeal feels too sentimental. In Brazil, the claims raise skepticism due to regulatory differences. In France, the use of English-origin environmental terms feels alienating. And in the U.S., the message still performs but lacks cultural novelty.

Now contrast that with an intent-first generative system. Each locale receives a version built natively: the Japanese message emphasizes social harmony and craftsmanship; the Brazilian one blends ecological messaging with aspirational affordability; the French version centers around civic responsibility; and the U.S. copy leans into personal agency and status.

These variations are not only different; they are aligned. Each draws from a shared semantic intent model but filters that intent through market-specific tone maps, terminology clusters, regulatory constraints, and emotional cues. Quality is not applied after the fact. It is embedded from the start and measured at every step using frameworks like MMQEval.

The results? Fewer rewrites. Fewer cultural misfires. Faster time to market. Higher engagement.

Not because the AI was “smarter” but because the system was better prepared.

Who gets to define “good enough”?

The GenAI moment has forced us to redefine what counts as quality, and who decides. Is it the model? The LQA team? The market? The stakeholder who signed off on the prompt? Or the person who clicked “skip ad“ because it did not land?

Quality is no longer binary. It is contextual, multimodal, and increasingly subjective. But that is not an excuse for vagueness. It is a call for rigor. Evaluating global content now means incorporating perspectives that were long overlooked, cultural cues, emotional preferences, even visual taboos. The new frontier of quality is not error detection; it is alignment.

Closing the gap between creation and meaning

Here is the uncomfortable truth: Generative AI has removed bottlenecks in global content workflows. But it has also removed safeguards. In the rush to scale, we risk flattening nuance, homogenizing voice, and diluting resonance. That is why quality conversations must move upstream, and downstream.

So, here is the question we must ask ourselves as practitioners, strategists, and leaders in global content:

Are we just generating outputs, or are we building systems that care about what those outputs become in the world?

Because in the end, what we ship is not just content. It is meaning. And meaning, unlike tokens, cannot be auto-completed.

Attributions:

DirectGC™ is a system architecture developed by Trilogica Global.

The MMQEval framework is a collaborative initiative maintained by a global community of researchers, evaluators, and practitioners committed to open evaluation standards for multimodal content. Learn more at mmqeval.org.

Read the full 132-page Global Ambitions: (R)Evolution in Motion publication featuring vital perspectives from 31 industry leaders on the ongoing AI-spurred (r)evolution.

Get your digital copy today