00:00:00Welcome And Guest Setup
00:00:40Why Prompting Hits A Ceiling
00:03:34From Correctness To Market Resonance
00:06:49Simulating Performance Before You Ship
00:10:13A New Week For Localization Leads
00:12:50Where To Follow The Work
ABOUT THIS EPISODE
I sit down with Augstin Da Fieno Delucci, co-founder of Trilogica Global and former Director of Global Data and AI at Microsoft, to unpack a shift that changes how global teams should think about AI: treating context as a system, not a prompt.
We’ve all gotten better at generation, better models, better prompting, better fine-tuning. Agustin argues that progress is real, but it hits a ceiling when the surrounding architecture has no memory of what “good performance” looks like across markets. When context gets rebuilt from scratch each cycle, content performance becomes unpredictable, even if everything passes QA. We explore how to move beyond linguistic correctness and brand compliance as the finish line, and start aiming for resonance: the words, framing, and emotional register that actually drive action for a specific audience.
Then we get practical. We talk about “simulation before shipping” using tools many teams already have, including semantic analysis, cultural inference, sentiment analysis, social listening, and vector databases that store market knowledge for retrieval. We also dig into what’s still emerging: the feedback loop that continuously refines audience models with real outcomes so your system gets sharper over time.
English
United States
SHOW NOTES 🔗
TRANSCRIPT 🔗
WEBVTT
00:00:00.080 --> 00:00:01.840
Hello, welcome to Field Notes.
00:00:01.840 --> 00:00:06.480
I'm Stephanie, and today we have a bit of a special episode with a special guest.
00:00:06.480 --> 00:00:20.079
So we have Agustín Da Fieno Delucchi, the co-founder of Trilogica Global, and former director of global data and AI at Microsoft, and someone who's been shaping global content systems for over 25 years.
00:00:20.079 --> 00:00:24.800
So today he's gonna walk us through something he likes to call context as a system.
00:00:24.800 --> 00:00:34.320
So really what it means to treat context as this living infrastructure rather than just a prompt, and then also what it changes for localization leaders.
00:00:34.320 --> 00:00:36.799
So, Agustin, thanks so much for joining us.
00:00:37.280 --> 00:00:39.840
Thank you so much for having me here, Stephanie.
00:00:40.320 --> 00:00:53.600
So to dive in then, we've spent these last couple years in the industry really obsessing over some things about which models to be using with AI, how to prompt better, how to fine-tune things.
00:00:53.600 --> 00:00:58.880
And then with this concept, you've been somewhat arguing that this is the wrong conversation.
00:00:58.880 --> 00:01:04.480
So, what is that wall that you're seeing teams hit when they're focusing just on those things?
00:01:04.879 --> 00:01:06.000
Yeah, absolutely.
00:01:06.000 --> 00:01:08.879
But let me actually reframe it slightly.
00:01:08.879 --> 00:01:14.879
So uh I think that it's not that the conversation about models and prompting is wrong.
00:01:14.879 --> 00:01:19.280
That works matter and it's deliver real gains.
00:01:19.280 --> 00:01:23.200
I think that the issue is that it's becoming insufficient on its own.
00:01:23.200 --> 00:01:34.959
Is that teams have gotten very good at generation, faster, more fluent, more scalable, and that progress is real and worth keeping.
00:01:34.959 --> 00:01:44.239
But there is a ceiling you hit when you're optimizing for generation in isolation from a much broader system.
00:01:44.239 --> 00:01:54.480
The output gets better, but yet organizations are still finding themselves surprised by how content performs across markets.
00:01:54.480 --> 00:02:05.040
And not because the generation has failed, but because the system around it has no model of what performing well actually means.
00:02:05.040 --> 00:02:07.120
It's a bit like guesswork.
00:02:07.680 --> 00:02:07.920
Yeah.
00:02:08.719 --> 00:02:16.240
A pattern I see repeatedly, a teams rolled out AI generating product content across, let's say, 15 markets.
00:02:16.240 --> 00:02:21.840
Linguistically, it passes every quality check with flying colors.
00:02:21.840 --> 00:02:24.000
No doubt about it.
00:02:24.000 --> 00:02:28.879
But content performance across those markets is essentially unpredictable.
00:02:28.879 --> 00:02:32.560
Sometimes better than expected, sometimes worse.
00:02:32.560 --> 00:02:50.240
And the system can't tell you which of those in advance because it has basically no memory of what worked in those markets before, no structure model of how the audience is writing for, each generation cycle starts from scratch.
00:02:50.240 --> 00:02:53.360
Better prompts helped at the margins.
00:02:53.360 --> 00:02:58.400
But what is actually needed is a more systemic approach.
00:02:58.400 --> 00:03:05.280
One where context isn't reconstructed each time from scratch and passed along every single time.
00:03:05.280 --> 00:03:15.680
But something that lives in the architecture as a structure and let's say persistent operational data that the whole system infer from.
00:03:15.680 --> 00:03:20.319
And that's that's not instead of the better models.
00:03:20.319 --> 00:03:21.759
We still need those.
00:03:21.759 --> 00:03:24.879
It is what makes better models actually land.
00:03:24.879 --> 00:03:34.000
The two have to work together, and right now the context infrastructure side of that equation is significantly underdeveloped.
00:03:34.560 --> 00:03:40.159
So you framed this whole concept as like the next step of progression.
00:03:40.159 --> 00:03:49.520
After 40 years, we had linguistic control, we had process context, then content systems, data-driven context, now generational AI.
00:03:49.520 --> 00:03:55.680
So this would be that next shift to have that contextual ecosystem, shall we say.
00:03:55.680 --> 00:04:01.919
So, what do you see as a little bit different about this shift as compared to things in the past?
00:04:02.400 --> 00:04:08.400
So every previous shift, I think, was about getting closer to the content itself.
00:04:08.400 --> 00:04:12.960
More control, more consistency, more scale.
00:04:12.960 --> 00:04:25.920
Translation memory gave us linguistic reuse, style guides keep giving us process control, content management is giving us structure.
00:04:25.920 --> 00:04:31.040
But data-driven approaches give us feedback signals after the fact.
00:04:31.040 --> 00:04:36.800
Generative AI is giving us the ability to produce at a completely different scale.
00:04:36.800 --> 00:04:42.639
So each step is real progress, and we're seeing it.
00:04:42.639 --> 00:04:46.319
But they were all still about the content artifact.
00:04:46.319 --> 00:04:49.600
The sentence, the segment, the asset.
00:04:49.600 --> 00:04:55.199
And the bar with measuring against has always been essentially the same.
00:04:55.199 --> 00:04:56.720
Is this correct?
00:04:56.720 --> 00:04:59.519
Does it comply with the guidelines?
00:04:59.519 --> 00:05:02.639
Does it really pass the style check?
00:05:02.639 --> 00:05:06.160
I think those guard rates are still necessary.
00:05:06.160 --> 00:05:14.480
And I want to be clear about linguistic correctness, brand compliance, regulatory adherence, those don't go away.
00:05:14.480 --> 00:05:16.160
They still need to be there.
00:05:16.160 --> 00:05:19.360
But they are a floor, not a ceiling.
00:05:19.360 --> 00:05:24.240
So for a long time we've been treating the floor as the goal.
00:05:24.240 --> 00:05:30.720
Think about you know, a major retailer manages seasonal campaigns across, let's say, 20 markets.
00:05:30.720 --> 00:05:40.720
Today, even with sophisticated tooling, the approach is essentially produce content, localize it, ship it, and wait to see what happens.
00:05:40.720 --> 00:05:43.120
The feedback loop is reactive.
00:05:43.120 --> 00:05:47.279
You learn something went wrong when the market tells you it did.
00:05:47.279 --> 00:05:53.600
What becomes possible now is to move well beyond correctness.
00:05:53.600 --> 00:05:59.279
The model that campaigns against structured market data before it shifts.
00:05:59.279 --> 00:06:01.040
Well before it shifts.
00:06:01.040 --> 00:06:08.160
And it shows and surfaces not just what's technically compliant, but what's actually going to resonate.
00:06:08.160 --> 00:06:15.839
Which words carry the most weight for this audience, which framing drives action, and which one gets ignored.
00:06:15.839 --> 00:06:19.120
And that is even before a single asset goes out.
00:06:19.120 --> 00:06:21.839
And that's a fundamentally higher bar.
00:06:21.839 --> 00:06:25.199
We are not just getting better at producing content.
00:06:25.199 --> 00:06:32.079
We are starting to get serious about choosing the content that matters for a specific market at a specific moment.
00:06:32.079 --> 00:06:36.480
And that is a new territory for the industry at that level of detail.
00:06:36.480 --> 00:06:43.279
Of course, with all the assets we have today, there are more generic indicators and there is responsibility about it.
00:06:43.279 --> 00:06:48.800
But the opportunity to unleash all this extra stuff is big.
00:06:49.120 --> 00:07:02.319
So this concept of kind of simulation before shipping, even, so predicting how content will land in the market before it goes out, what does that actually look like in practice?
00:07:02.319 --> 00:07:06.720
Is that something that's real today, or is it something that we're heading towards?
00:07:07.040 --> 00:07:10.879
Yes, and it's partially real and partially where we are heading.
00:07:10.879 --> 00:07:14.480
And I think it's important that we are honest about that distinction.
00:07:14.480 --> 00:07:19.759
We can today build structured representations of audiences and markets.
00:07:19.759 --> 00:07:21.439
Not just demographic data.
00:07:21.439 --> 00:07:27.040
We are talking about behavioral patterns, linguistic preference, emotional registers.
00:07:27.040 --> 00:07:30.079
What resonated and what's fallen flat?
00:07:30.079 --> 00:07:31.360
We can track that.
00:07:31.360 --> 00:07:37.279
The technology is making this possible right now are things most teams already have access to.
00:07:37.279 --> 00:07:47.920
Large language model, semantic analysis, cultural inference, vector databases for storing and retrieving market unloaded knowledge.
00:07:47.920 --> 00:07:56.959
Sentiment analysis, you can do that on social listening tools that we all use that feed real behavior signal into the system.
00:07:56.959 --> 00:08:05.040
And also structured data schemas that formalize that we know about given markets into something that we can actually interpret.
00:08:05.040 --> 00:08:13.519
What's less common is the architectural discipline to connect those pieces into a coherent system rather than using them as isolated tools.
00:08:13.519 --> 00:08:15.920
And I think that's where the main difference is.
00:08:15.920 --> 00:08:21.519
We can run content against those models before deploying and then surface those risks.
00:08:21.519 --> 00:08:25.839
Cultural misalignment, semantic drift, register mismatch, all that.
00:08:25.839 --> 00:08:34.639
But if you're launching, for example, a financial services campaign in Brazil and Japan simultaneously, let's say.
00:08:34.639 --> 00:08:50.879
A structured content system can flag before anything is published, that the directness of the call to action is likely to read aggressive in one market, and the visual metaphor anchoring the copy carries an unintended connotation in the other.
00:08:50.879 --> 00:08:53.600
And that's no guesswork.
00:08:53.600 --> 00:09:01.279
Is the system really interpreting model a model to of those audiences that has been built and refined over time?
00:09:01.279 --> 00:09:05.919
And today that kind of signal often comes from a human reviewer after the fact.
00:09:05.919 --> 00:09:12.799
And it comes, if it comes at all, I think what is still developing is the feedback loop.
00:09:12.799 --> 00:09:20.000
This ability to continuously refine the model, to retrofit based on the actual performance.
00:09:20.000 --> 00:09:24.240
So the system gets better at predicting outcomes over time.
00:09:24.240 --> 00:09:26.080
And that's the full version of it.
00:09:26.080 --> 00:09:29.919
It's closing the loop and finalizing the circle.
00:09:29.919 --> 00:09:33.120
And we are not uniformly there yet.
00:09:33.120 --> 00:09:43.840
But here's what I would say to someone skeptical: the organizations that start building that content infrastructure now are accumulating something that really compounds.
00:09:43.840 --> 00:09:53.200
Every market signal, every outcome, every data point and refinement goes into a system that gets sharper.
00:09:53.200 --> 00:10:01.120
And not just by updating translation memory, not just by updating terminology, by updating real market signals.
00:10:01.120 --> 00:10:12.320
The ones waiting for the technology to mature before they start building will be starting from zero when their competitors are already operating from a significant base.
00:10:12.960 --> 00:10:16.399
So let's bring it back to say like the localization manager, right?
00:10:16.399 --> 00:10:27.440
So you talk about these sort of experts who are now moving from that reviewing the outputs to then designing the system from the end of the line to the instrument panel, as it were.
00:10:27.440 --> 00:10:35.600
For a localization lead listening right now, what does this actually mean for how they would spend their week six months from now?
00:10:35.919 --> 00:10:36.320
Right.
00:10:36.320 --> 00:10:37.600
It's a great question.
00:10:37.600 --> 00:10:40.559
So the two things that I have very practical.
00:10:40.559 --> 00:10:49.039
The first is going back to the drawing board and how context actually travels through your pipeline, to your workflow.
00:10:49.039 --> 00:10:59.120
Now, not how it's supposed to travel, because we know that we all have uh ideal situations, but how it actually happens can be different.
00:10:59.120 --> 00:11:00.480
So map it.
00:11:00.480 --> 00:11:01.679
That's important.
00:11:01.679 --> 00:11:05.840
Where is it that the context is created?
00:11:05.840 --> 00:11:07.759
Where is it handed off?
00:11:07.759 --> 00:11:12.000
Where exactly does it get dropped?
00:11:12.000 --> 00:11:25.200
Most teams, when they do this honestly, find the context is being reconstructed from scratch, if at all, at multiple points during the public, because there is no persistent place where it lives.
00:11:25.200 --> 00:11:28.639
That map is your starting point.
00:11:28.639 --> 00:11:31.840
It tells you where the architecture needs attention.
00:11:31.840 --> 00:11:42.320
The second, I think, is like look at your current system and ask honestly whether they can tell you or they can anticipate performance.
00:11:42.320 --> 00:11:50.960
Not measuring it after the fact or separately, but modeling it before the content gets goes out and is created.
00:11:50.960 --> 00:11:59.039
Can your system tell you really before deployment that a given piece of content is likely to underperform in a specific market?
00:11:59.039 --> 00:12:04.000
If the answer is no, that's the gap you're designing towards.
00:12:04.000 --> 00:12:12.879
So six months from now, a localization lead doing this work isn't spending their weekend reviewing outputs.
00:12:12.879 --> 00:12:16.480
They are asking, what does the system know about this market?
00:12:16.480 --> 00:12:19.039
What is it telling us before we ship?
00:12:19.039 --> 00:12:34.960
That's a fundamentally different professional orientation, and I think a significant, more valuable one, where you are looking at the signals that the system is doing, understanding why, and then modeling the system to actually do that.
00:12:34.960 --> 00:12:50.240
And that requires a lot of also linguistic effort in fine-tuning, but most importantly, it's about understanding your market, understanding where you're going, mapping that into your brand, into your contest goals, and all that.
00:12:50.559 --> 00:12:53.440
We're about out of time and we've just scratched the surface here.
00:12:53.440 --> 00:13:00.240
So is there somewhere online where someone could go to either connect with you or to learn more about these concepts?
00:13:00.559 --> 00:13:01.519
Yes, absolutely.
00:13:01.519 --> 00:13:04.159
I think the best place to start is LinkedIn.
00:13:04.159 --> 00:13:10.480
I publish regularly on this work, including pieces of the framework we've been talking about today.
00:13:10.480 --> 00:13:19.759
And if you want to go deeper into what Trilogica Global is building in this space, the website is Trilogicaglobal.com.
00:13:19.759 --> 00:13:25.759
And there is context there on our approach to global content systems and where this work is heading.
00:13:25.759 --> 00:13:32.240
We are also developing more formal published materials, articles, additional talk sessions.
00:13:32.240 --> 00:13:36.960
So following either of those channels is perhaps the best way to connect us.
00:13:37.200 --> 00:13:40.080
So yeah, we'll include those links in our episode pages.
00:13:40.080 --> 00:13:42.799
So yeah, feel free, go ahead and connect with him there.
00:13:42.799 --> 00:13:43.360
Okay.
00:13:43.360 --> 00:13:45.360
Agustin, thank you so much for joining us.
00:13:45.679 --> 00:13:47.120
Thank you so much, Stephanie.
00:00:00.080 --> 00:00:01.840
Hello, welcome to Field Notes.
00:00:01.840 --> 00:00:06.480
I'm Stephanie, and today we have a bit of a special episode with a special guest.
00:00:06.480 --> 00:00:20.079
So we have Agustín Da Fieno Delucchi, the co-founder of Trilogica Global, and former director of global data and AI at Microsoft, and someone who's been shaping global content systems for over 25 years.
00:00:20.079 --> 00:00:24.800
So today he's gonna walk us through something he likes to call context as a system.
00:00:24.800 --> 00:00:34.320
So really what it means to treat context as this living infrastructure rather than just a prompt, and then also what it changes for localization leaders.
00:00:34.320 --> 00:00:36.799
So, Agustin, thanks so much for joining us.
00:00:37.280 --> 00:00:39.840
Thank you so much for having me here, Stephanie.
00:00:40.320 --> 00:00:53.600
So to dive in then, we've spent these last couple years in the industry really obsessing over some things about which models to be using with AI, how to prompt better, how to fine-tune things.
00:00:53.600 --> 00:00:58.880
And then with this concept, you've been somewhat arguing that this is the wrong conversation.
00:00:58.880 --> 00:01:04.480
So, what is that wall that you're seeing teams hit when they're focusing just on those things?
00:01:04.879 --> 00:01:06.000
Yeah, absolutely.
00:01:06.000 --> 00:01:08.879
But let me actually reframe it slightly.
00:01:08.879 --> 00:01:14.879
So uh I think that it's not that the conversation about models and prompting is wrong.
00:01:14.879 --> 00:01:19.280
That works matter and it's deliver real gains.
00:01:19.280 --> 00:01:23.200
I think that the issue is that it's becoming insufficient on its own.
00:01:23.200 --> 00:01:34.959
Is that teams have gotten very good at generation, faster, more fluent, more scalable, and that progress is real and worth keeping.
00:01:34.959 --> 00:01:44.239
But there is a ceiling you hit when you're optimizing for generation in isolation from a much broader system.
00:01:44.239 --> 00:01:54.480
The output gets better, but yet organizations are still finding themselves surprised by how content performs across markets.
00:01:54.480 --> 00:02:05.040
And not because the generation has failed, but because the system around it has no model of what performing well actually means.
00:02:05.040 --> 00:02:07.120
It's a bit like guesswork.
00:02:07.680 --> 00:02:07.920
Yeah.
00:02:08.719 --> 00:02:16.240
A pattern I see repeatedly, a teams rolled out AI generating product content across, let's say, 15 markets.
00:02:16.240 --> 00:02:21.840
Linguistically, it passes every quality check with flying colors.
00:02:21.840 --> 00:02:24.000
No doubt about it.
00:02:24.000 --> 00:02:28.879
But content performance across those markets is essentially unpredictable.
00:02:28.879 --> 00:02:32.560
Sometimes better than expected, sometimes worse.
00:02:32.560 --> 00:02:50.240
And the system can't tell you which of those in advance because it has basically no memory of what worked in those markets before, no structure model of how the audience is writing for, each generation cycle starts from scratch.
00:02:50.240 --> 00:02:53.360
Better prompts helped at the margins.
00:02:53.360 --> 00:02:58.400
But what is actually needed is a more systemic approach.
00:02:58.400 --> 00:03:05.280
One where context isn't reconstructed each time from scratch and passed along every single time.
00:03:05.280 --> 00:03:15.680
But something that lives in the architecture as a structure and let's say persistent operational data that the whole system infer from.
00:03:15.680 --> 00:03:20.319
And that's that's not instead of the better models.
00:03:20.319 --> 00:03:21.759
We still need those.
00:03:21.759 --> 00:03:24.879
It is what makes better models actually land.
00:03:24.879 --> 00:03:34.000
The two have to work together, and right now the context infrastructure side of that equation is significantly underdeveloped.
00:03:34.560 --> 00:03:40.159
So you framed this whole concept as like the next step of progression.
00:03:40.159 --> 00:03:49.520
After 40 years, we had linguistic control, we had process context, then content systems, data-driven context, now generational AI.
00:03:49.520 --> 00:03:55.680
So this would be that next shift to have that contextual ecosystem, shall we say.
00:03:55.680 --> 00:04:01.919
So, what do you see as a little bit different about this shift as compared to things in the past?
00:04:02.400 --> 00:04:08.400
So every previous shift, I think, was about getting closer to the content itself.
00:04:08.400 --> 00:04:12.960
More control, more consistency, more scale.
00:04:12.960 --> 00:04:25.920
Translation memory gave us linguistic reuse, style guides keep giving us process control, content management is giving us structure.
00:04:25.920 --> 00:04:31.040
But data-driven approaches give us feedback signals after the fact.
00:04:31.040 --> 00:04:36.800
Generative AI is giving us the ability to produce at a completely different scale.
00:04:36.800 --> 00:04:42.639
So each step is real progress, and we're seeing it.
00:04:42.639 --> 00:04:46.319
But they were all still about the content artifact.
00:04:46.319 --> 00:04:49.600
The sentence, the segment, the asset.
00:04:49.600 --> 00:04:55.199
And the bar with measuring against has always been essentially the same.
00:04:55.199 --> 00:04:56.720
Is this correct?
00:04:56.720 --> 00:04:59.519
Does it comply with the guidelines?
00:04:59.519 --> 00:05:02.639
Does it really pass the style check?
00:05:02.639 --> 00:05:06.160
I think those guard rates are still necessary.
00:05:06.160 --> 00:05:14.480
And I want to be clear about linguistic correctness, brand compliance, regulatory adherence, those don't go away.
00:05:14.480 --> 00:05:16.160
They still need to be there.
00:05:16.160 --> 00:05:19.360
But they are a floor, not a ceiling.
00:05:19.360 --> 00:05:24.240
So for a long time we've been treating the floor as the goal.
00:05:24.240 --> 00:05:30.720
Think about you know, a major retailer manages seasonal campaigns across, let's say, 20 markets.
00:05:30.720 --> 00:05:40.720
Today, even with sophisticated tooling, the approach is essentially produce content, localize it, ship it, and wait to see what happens.
00:05:40.720 --> 00:05:43.120
The feedback loop is reactive.
00:05:43.120 --> 00:05:47.279
You learn something went wrong when the market tells you it did.
00:05:47.279 --> 00:05:53.600
What becomes possible now is to move well beyond correctness.
00:05:53.600 --> 00:05:59.279
The model that campaigns against structured market data before it shifts.
00:05:59.279 --> 00:06:01.040
Well before it shifts.
00:06:01.040 --> 00:06:08.160
And it shows and surfaces not just what's technically compliant, but what's actually going to resonate.
00:06:08.160 --> 00:06:15.839
Which words carry the most weight for this audience, which framing drives action, and which one gets ignored.
00:06:15.839 --> 00:06:19.120
And that is even before a single asset goes out.
00:06:19.120 --> 00:06:21.839
And that's a fundamentally higher bar.
00:06:21.839 --> 00:06:25.199
We are not just getting better at producing content.
00:06:25.199 --> 00:06:32.079
We are starting to get serious about choosing the content that matters for a specific market at a specific moment.
00:06:32.079 --> 00:06:36.480
And that is a new territory for the industry at that level of detail.
00:06:36.480 --> 00:06:43.279
Of course, with all the assets we have today, there are more generic indicators and there is responsibility about it.
00:06:43.279 --> 00:06:48.800
But the opportunity to unleash all this extra stuff is big.
00:06:49.120 --> 00:07:02.319
So this concept of kind of simulation before shipping, even, so predicting how content will land in the market before it goes out, what does that actually look like in practice?
00:07:02.319 --> 00:07:06.720
Is that something that's real today, or is it something that we're heading towards?
00:07:07.040 --> 00:07:10.879
Yes, and it's partially real and partially where we are heading.
00:07:10.879 --> 00:07:14.480
And I think it's important that we are honest about that distinction.
00:07:14.480 --> 00:07:19.759
We can today build structured representations of audiences and markets.
00:07:19.759 --> 00:07:21.439
Not just demographic data.
00:07:21.439 --> 00:07:27.040
We are talking about behavioral patterns, linguistic preference, emotional registers.
00:07:27.040 --> 00:07:30.079
What resonated and what's fallen flat?
00:07:30.079 --> 00:07:31.360
We can track that.
00:07:31.360 --> 00:07:37.279
The technology is making this possible right now are things most teams already have access to.
00:07:37.279 --> 00:07:47.920
Large language model, semantic analysis, cultural inference, vector databases for storing and retrieving market unloaded knowledge.
00:07:47.920 --> 00:07:56.959
Sentiment analysis, you can do that on social listening tools that we all use that feed real behavior signal into the system.
00:07:56.959 --> 00:08:05.040
And also structured data schemas that formalize that we know about given markets into something that we can actually interpret.
00:08:05.040 --> 00:08:13.519
What's less common is the architectural discipline to connect those pieces into a coherent system rather than using them as isolated tools.
00:08:13.519 --> 00:08:15.920
And I think that's where the main difference is.
00:08:15.920 --> 00:08:21.519
We can run content against those models before deploying and then surface those risks.
00:08:21.519 --> 00:08:25.839
Cultural misalignment, semantic drift, register mismatch, all that.
00:08:25.839 --> 00:08:34.639
But if you're launching, for example, a financial services campaign in Brazil and Japan simultaneously, let's say.
00:08:34.639 --> 00:08:50.879
A structured content system can flag before anything is published, that the directness of the call to action is likely to read aggressive in one market, and the visual metaphor anchoring the copy carries an unintended connotation in the other.
00:08:50.879 --> 00:08:53.600
And that's no guesswork.
00:08:53.600 --> 00:09:01.279
Is the system really interpreting model a model to of those audiences that has been built and refined over time?
00:09:01.279 --> 00:09:05.919
And today that kind of signal often comes from a human reviewer after the fact.
00:09:05.919 --> 00:09:12.799
And it comes, if it comes at all, I think what is still developing is the feedback loop.
00:09:12.799 --> 00:09:20.000
This ability to continuously refine the model, to retrofit based on the actual performance.
00:09:20.000 --> 00:09:24.240
So the system gets better at predicting outcomes over time.
00:09:24.240 --> 00:09:26.080
And that's the full version of it.
00:09:26.080 --> 00:09:29.919
It's closing the loop and finalizing the circle.
00:09:29.919 --> 00:09:33.120
And we are not uniformly there yet.
00:09:33.120 --> 00:09:43.840
But here's what I would say to someone skeptical: the organizations that start building that content infrastructure now are accumulating something that really compounds.
00:09:43.840 --> 00:09:53.200
Every market signal, every outcome, every data point and refinement goes into a system that gets sharper.
00:09:53.200 --> 00:10:01.120
And not just by updating translation memory, not just by updating terminology, by updating real market signals.
00:10:01.120 --> 00:10:12.320
The ones waiting for the technology to mature before they start building will be starting from zero when their competitors are already operating from a significant base.
00:10:12.960 --> 00:10:16.399
So let's bring it back to say like the localization manager, right?
00:10:16.399 --> 00:10:27.440
So you talk about these sort of experts who are now moving from that reviewing the outputs to then designing the system from the end of the line to the instrument panel, as it were.
00:10:27.440 --> 00:10:35.600
For a localization lead listening right now, what does this actually mean for how they would spend their week six months from now?
00:10:35.919 --> 00:10:36.320
Right.
00:10:36.320 --> 00:10:37.600
It's a great question.
00:10:37.600 --> 00:10:40.559
So the two things that I have very practical.
00:10:40.559 --> 00:10:49.039
The first is going back to the drawing board and how context actually travels through your pipeline, to your workflow.
00:10:49.039 --> 00:10:59.120
Now, not how it's supposed to travel, because we know that we all have uh ideal situations, but how it actually happens can be different.
00:10:59.120 --> 00:11:00.480
So map it.
00:11:00.480 --> 00:11:01.679
That's important.
00:11:01.679 --> 00:11:05.840
Where is it that the context is created?
00:11:05.840 --> 00:11:07.759
Where is it handed off?
00:11:07.759 --> 00:11:12.000
Where exactly does it get dropped?
00:11:12.000 --> 00:11:25.200
Most teams, when they do this honestly, find the context is being reconstructed from scratch, if at all, at multiple points during the public, because there is no persistent place where it lives.
00:11:25.200 --> 00:11:28.639
That map is your starting point.
00:11:28.639 --> 00:11:31.840
It tells you where the architecture needs attention.
00:11:31.840 --> 00:11:42.320
The second, I think, is like look at your current system and ask honestly whether they can tell you or they can anticipate performance.
00:11:42.320 --> 00:11:50.960
Not measuring it after the fact or separately, but modeling it before the content gets goes out and is created.
00:11:50.960 --> 00:11:59.039
Can your system tell you really before deployment that a given piece of content is likely to underperform in a specific market?
00:11:59.039 --> 00:12:04.000
If the answer is no, that's the gap you're designing towards.
00:12:04.000 --> 00:12:12.879
So six months from now, a localization lead doing this work isn't spending their weekend reviewing outputs.
00:12:12.879 --> 00:12:16.480
They are asking, what does the system know about this market?
00:12:16.480 --> 00:12:19.039
What is it telling us before we ship?
00:12:19.039 --> 00:12:34.960
That's a fundamentally different professional orientation, and I think a significant, more valuable one, where you are looking at the signals that the system is doing, understanding why, and then modeling the system to actually do that.
00:12:34.960 --> 00:12:50.240
And that requires a lot of also linguistic effort in fine-tuning, but most importantly, it's about understanding your market, understanding where you're going, mapping that into your brand, into your contest goals, and all that.
00:12:50.559 --> 00:12:53.440
We're about out of time and we've just scratched the surface here.
00:12:53.440 --> 00:13:00.240
So is there somewhere online where someone could go to either connect with you or to learn more about these concepts?
00:13:00.559 --> 00:13:01.519
Yes, absolutely.
00:13:01.519 --> 00:13:04.159
I think the best place to start is LinkedIn.
00:13:04.159 --> 00:13:10.480
I publish regularly on this work, including pieces of the framework we've been talking about today.
00:13:10.480 --> 00:13:19.759
And if you want to go deeper into what Trilogica Global is building in this space, the website is Trilogicaglobal.com.
00:13:19.759 --> 00:13:25.759
And there is context there on our approach to global content systems and where this work is heading.
00:13:25.759 --> 00:13:32.240
We are also developing more formal published materials, articles, additional talk sessions.
00:13:32.240 --> 00:13:36.960
So following either of those channels is perhaps the best way to connect us.
00:13:37.200 --> 00:13:40.080
So yeah, we'll include those links in our episode pages.
00:13:40.080 --> 00:13:42.799
So yeah, feel free, go ahead and connect with him there.
00:13:42.799 --> 00:13:43.360
Okay.
00:13:43.360 --> 00:13:45.360
Agustin, thank you so much for joining us.
00:13:45.679 --> 00:13:47.120
Thank you so much, Stephanie.
Disclaimer: The podcast and artwork embedded on this page are from Stephanie Harris-Yee, Argos Multilingual, which is the property of its owner and not affiliated with or endorsed by Listen Notes, Inc.
EDIT
Thank you for helping to keep the podcast database up to date.
