ABOUT THIS EPISODE
Someone on your team ships a translated page overnight, looks like a hero, and nobody filed a localization request. Then you stumble on the copy later and think, “Did we do this?” That moment has a name: shadow localization. We dig into why it shows up even in mature programs, why AI and machine translation make it explode, and why treating it like a turf war is the fastest way to lose trust and relevance.
We talk through the real-world patterns: the small team that built a translation workflow years ago and never connected with localization, the “turnkey” vendor that bundles translation into a project and then asks us to sanity-check the output, and the random discovery of low-quality “translations in the wild” that ignore terminology, brand voice, and basic QA. From there, we share a practical response: reach out with curiosity, run a quick diagnostic, fix what truly needs fixing, and use the moment to onboard teams to better processes, shared SLAs, or volume pricing without forcing one rigid workflow on every use case.
The bigger takeaway is strategic: if we position ourselves as the team that translates, people will assume ChatGPT can replace us. If we position ourselves as the team with international intelligence, market context, and a plan for coherent multilingual experiences, we become essential. Listen, then share this with a localization peer and leave a review if it helps. Where are you seeing shadow localization pop up in your organization?
SHOW NOTES 🔗
TRANSCRIPT 🔗
00:00:00.320 --> 00:00:03.439
Hello, I'm Stephanie and back for another episode of Field Notes.
00:00:03.439 --> 00:00:10.480
And this time we're actually going to be starting a new thread in Field Notes with Giulia Greco from Giulia Greco Consulting.
00:00:10.480 --> 00:00:17.920
And she's going to be approaching things from more of that interpersonal organizational side of things within LOC programs.
00:00:17.920 --> 00:00:20.719
So super excited to start this new series with her.
00:00:21.039 --> 00:00:21.920
Thank you, Stephanie.
00:00:21.920 --> 00:00:23.519
This is very exciting for me too.
00:00:23.519 --> 00:00:30.480
I'm thrilled to be participating in this new iteration of Field Notes and grateful for the opportunity.
00:00:30.480 --> 00:00:37.039
And I think we're going to have some really interesting chats that hopefully can be useful to other localization professionals.
00:00:37.039 --> 00:00:40.240
So we can take it away with our first topic.
00:00:40.240 --> 00:00:41.600
All right, let's do it.
00:00:41.759 --> 00:00:52.560
So this first topic is one that many of you have probably been hearing about or maybe hopefully not experiencing, but it's something we like to call shadow localization.
00:00:52.560 --> 00:01:05.120
So this is what happens when you're doing your regular localization job, you have a good system in place, and all of a sudden you find out, oh, this other department has spun up their own workflows.
00:01:05.120 --> 00:01:14.159
Maybe they didn't realize you existed, maybe they just wanted to do their own thing, but now all of a sudden you have two separate workflows happening in the same company.
00:01:14.159 --> 00:01:16.480
Maybe they're doing great, maybe not.
00:01:16.480 --> 00:01:18.400
This is a situation.
00:01:18.400 --> 00:01:23.920
So AI has made translation something that anyone can do now, it seems.
00:01:23.920 --> 00:01:27.840
So you don't really need a translation team anymore, right?
00:01:28.319 --> 00:01:33.280
Well, that's what a lot of people seem to think.
00:01:33.280 --> 00:01:40.239
People that are not in the localization field, obviously, many people in leadership or other over-eager folks.
00:01:40.239 --> 00:01:43.359
And sometimes I think, honestly, can you blame them?
00:01:43.359 --> 00:01:49.359
The barrier to entry into translating has disappeared.
00:01:49.359 --> 00:02:03.120
But I also want to say something first, because this whole thing of shadow localization existed even before AI, at least in big companies that may have a tendency to be a little bit silosed.
00:02:03.120 --> 00:02:15.599
There's always like the little team that nobody knows about in an obscure part of the company doing something that kind of flies under the radar, that doesn't really interact with other teams.
00:02:15.599 --> 00:02:23.280
And they sometimes live under a rock and they just don't know that now there is a localization team.
00:02:23.280 --> 00:02:32.960
And maybe there wasn't two years ago or five years ago when they were assembled and started doing some work and realized that they needed translations.
00:02:32.960 --> 00:02:36.479
Maybe there wasn't any infrastructure and no people doing this.
00:02:36.479 --> 00:02:40.960
So they took matters in their own hands, but they still kept on flying under the radar.
00:02:40.960 --> 00:02:46.479
And this actually happened to me in my previous company with the patent team.
00:02:46.479 --> 00:02:47.840
Nobody knew about it.
00:02:47.840 --> 00:02:49.599
They were like three people.
00:02:49.599 --> 00:03:00.960
They had easy access to procurement and legal because of the nature of what they do, clearly, a clear mandate to get it done no matter what.
00:03:00.960 --> 00:03:12.400
And it one day, I guess one of them talked to somebody else in the company and found out that there was a localization team that was providing translations to the entire company except for them.
00:03:12.400 --> 00:03:18.000
And they were like, We have a contract for translations with the LSP.
00:03:18.000 --> 00:03:19.199
So they called me.
00:03:19.199 --> 00:03:25.360
I sat down, I had a chat, and uh, and I found out that the LSP was a reputable one.
00:03:25.360 --> 00:03:27.039
I found out about their volumes.
00:03:27.039 --> 00:03:32.960
I asked them, Do you think you want to come on board and have my team handle your work?
00:03:32.960 --> 00:03:37.280
Or is this something that you are okay continuing with the way you are?
00:03:37.280 --> 00:03:39.840
And they were, we don't really know what we don't know.
00:03:39.840 --> 00:03:49.360
And what I realized was that they could probably benefit from being onboarded to my LSP because I said you can benefit from volume discounts.
00:03:49.360 --> 00:03:58.080
At this point, you get under the umbrella of our general contract and SLAs, but you can still manage the whole pipeline on your own.
00:03:58.080 --> 00:04:06.400
Because I think they wanted to maintain independence and the ability of just do things the way they had always done them.
00:04:06.400 --> 00:04:07.840
Everything worked out.
00:04:07.840 --> 00:04:19.120
It became a question of can you get a better price if you hop on our bus with this LSP, with whom we have a big contract, versus keep going as a small thing with your own LSP.
00:04:19.120 --> 00:04:20.319
And that's what they did.
00:04:20.319 --> 00:04:25.120
I put them in touch directly with my LSP and I said, they're on their own.
00:04:25.120 --> 00:04:26.720
I don't need to take this on.
00:04:26.720 --> 00:04:28.800
God knows we have enough work as it is.
00:04:28.800 --> 00:04:32.000
I'm not looking for more work for me and my team.
00:04:32.000 --> 00:04:39.920
And they were happy, and then it just became an addendum to our regular contract with the negotiated rates and everything.
00:04:39.920 --> 00:04:41.439
And Bob's your uncle.
00:04:41.439 --> 00:04:42.959
Everybody was happy.
00:04:42.959 --> 00:04:45.120
And this was pre-AI days.
00:04:45.120 --> 00:04:53.439
So this is just to say that it can happen, it was happening even before, but now it has a bit of a different flavor.
00:04:53.439 --> 00:05:23.600
Now it's like teams who are like, oh man, I used to have to go through the localization team and having to fight for budget and make a business case, uh, and now I can just go to ChatGPT and get it translated immediately and move by myself and look like the hero who's shipping stuff, and I don't have to deal with those annoying localization people who are telling me the qualities or tell me that I have to get in line because there's more important things in the queue first.
00:05:23.600 --> 00:05:26.959
So, this is what I mean when I say can you blame them.
00:05:26.959 --> 00:05:30.720
The reality is that for a lot of teams, uh, there is a queue.
00:05:30.720 --> 00:05:50.800
Yeah, translation is not immediate when you're going through a localization team, and sometimes localization people get to hang up on the consistency and the quality conversation without realizing that is of absolutely no interest and no consequence to their interlocutor.
00:05:50.800 --> 00:06:11.360
And I think that's a conversation that we need to have because if there is shadow localization in your org right now, this is a signal that maybe something needs to change so that it doesn't become more widespread and eventually you end up appearing as irrelevant and not needed.
00:06:11.360 --> 00:06:13.199
Yeah, that's the number one thing.
00:06:13.519 --> 00:06:15.600
It sounds like things can come in two ways.
00:06:15.600 --> 00:06:22.240
So, one, like you mentioned in your previous company, where they actually realize that, oh, there's a localization department now and they'll come to you.
00:06:22.240 --> 00:06:28.639
And then other cases where you'll go out and you'll say, wait, this suddenly popped up out of nowhere.
00:06:28.639 --> 00:06:29.920
Here is this thing that's happening.
00:06:29.920 --> 00:06:30.240
Yeah.
00:06:30.240 --> 00:06:34.160
And either case, sometimes it'll be working well, sometimes it won't.
00:06:34.160 --> 00:06:42.800
What does happen if you come across one of those situations where you found someone, they've made this whole system, and it's a mess?
00:06:43.120 --> 00:06:45.600
Yeah, so this is the other way, exactly.
00:06:45.600 --> 00:06:50.639
To your point, you usually find out in two ways that there's shadow localization.
00:06:50.639 --> 00:06:55.199
Either because somebody contacts you out of the blue and says, Hey, I didn't know you guys existed.
00:06:55.199 --> 00:06:59.360
We've been doing translations for a while, like that patent team that I just mentioned.
00:06:59.360 --> 00:07:08.319
Otherwise, you just accidentally, randomly run into some content in the wild, as I like to call it, like translations in the wild.
00:07:08.319 --> 00:07:10.720
You're like, wait a minute, what is this?
00:07:10.720 --> 00:07:11.920
Did we do this?
00:07:11.920 --> 00:07:14.240
This doesn't look like something that we did.
00:07:14.240 --> 00:07:20.560
And usually it's stuff that attracts your attention because, to my previous point, the quality is not there.
00:07:20.560 --> 00:07:27.920
And so it's harder to surface, but it does happen quite a bit, and it did happen to me and my team quite a bit.
00:07:27.920 --> 00:07:50.000
Yeah, it's harder to get to the bottom of it, and it's honestly way more concerning because it is that other aspect of teams who maybe they never came to you, but maybe they did, they tried a few times, and the process felt either too slow for their timeline or somehow it didn't satisfy their needs.
00:07:50.000 --> 00:07:56.639
And so they ended up building a workaround, and they have no reason to stop unless you give them a reason to stop.
00:07:56.639 --> 00:08:09.199
So when that happens, if you're able to get to the bottom of it and find the team or the person responsible for this, uh, what I did in the past was just reach out and say, hey, we noticed that you did this translation.
00:08:09.199 --> 00:08:10.240
Can you tell me more about it?
00:08:10.240 --> 00:08:16.959
I come from the localization team, so I'm interested in everything multilingual, and I'm curious as to how you guys got this done.
00:08:16.959 --> 00:08:25.279
Also, just FYI, we noticed that language XYZ has some significant mistakes.
00:08:25.279 --> 00:08:39.679
There's been other instances where other teams have done localization through a third-party vendor, and so they sell a turnkey type of project where they include translation and make it sounded like it's gonna be so easy, you don't have to worry about a thing.
00:08:39.679 --> 00:08:44.960
And people that are already overworked and under resourced are like, yeah, okay, sure, sounds good, but they'll take care of it.
00:08:44.960 --> 00:08:53.840
But then they get a little doubt, and so they may say yes, fully knowing that there is an internal localization team, and then they'll go to the localization team and say, Hey, can you check this?
00:08:53.840 --> 00:08:58.480
My vendor did it, and we wanna make sure that they did it correctly.
00:08:58.480 --> 00:09:04.159
And I hated that because that's basically you're getting called in as a cleanup crew, basically.
00:09:04.159 --> 00:09:22.399
Having said that, it's annoying, but it is an incredible opportunity, just like it's a good opportunity when you find the shadow localization in the wild, because either way, if you are able to establish a connection or if they come to you asking to check, this is your opportunity to run a diagnostic and demonstrate your value.
00:09:22.399 --> 00:09:31.840
And the review usually uncovers problems because of course it does, because there was no style guide, there was no terminology adherence, there were no branded terms.
00:09:31.840 --> 00:09:35.519
And so usually you can find quite a bit that needs to be changed.
00:09:35.519 --> 00:09:42.320
So you do it, you fix it, you go back, and you're like, we've done the cleanup, it was quite significant, and here's why.
00:09:42.320 --> 00:09:49.759
And next time, feel free to just call us, we can do this for you, and it will be done correctly from the get-go.
00:09:49.759 --> 00:09:52.960
And that starts the conversation as to this is how you contact us.
00:09:52.960 --> 00:09:56.320
You basically use this as an opportunity to onboard a new team.
00:09:56.480 --> 00:09:56.720
Yeah.
00:09:56.720 --> 00:10:05.039
So is that usually the answer then to try to draw them back in and bring it back under the localization team as much as possible?
00:10:05.039 --> 00:10:07.840
Or is that just a case-by-case thing?
00:10:08.159 --> 00:10:09.519
I think it's a case-by-case thing.
00:10:09.519 --> 00:10:09.679
Yeah.
00:10:09.679 --> 00:10:14.799
And that's why I wanted to start this whole conversation with the patent team.
00:10:14.799 --> 00:10:18.000
Because in that case, I didn't bring them under my umbrella.
00:10:18.000 --> 00:10:19.919
You don't always want to bring it under it.
00:10:19.919 --> 00:10:36.159
And this is a little bit uncomfortable because I think it's in localization circles, there's this idea that no, everything has to go through you and centralizing it, not in terms of org, but in terms of all the work needs to flow through you, it's considered the right thing to do.
00:10:36.159 --> 00:10:39.279
And I'm like, maybe not, maybe not in every situation.
00:10:39.279 --> 00:10:42.480
Not everything needs the full localization treatment.
00:10:42.480 --> 00:11:08.799
And so, unless you are incredibly nimble as a team, you can accommodate very different ways of working very quickly, suddenly, because a team that's doing stuff on their own on the side, because it's faster, because they want it immediately, and because maybe they're just translating a landing page that literally will stay online for 48 hours.
00:11:08.799 --> 00:11:25.519
We had a lot of those in my previous companies, like in country teams that were organizing in real life events in various markets and needed just to throw a landing page up for registration with some information, and then it would come down after a week.
00:11:25.519 --> 00:11:31.440
Do I really need to get hung up on it needs to be done 100% correctly, or can they just write it on their own?
00:11:31.440 --> 00:11:33.519
In some cases, it wasn't even localization.
00:11:33.519 --> 00:11:40.720
It would just get completely created in the local languages from scratch, published, and it's fine.
00:11:40.720 --> 00:11:49.600
So it's more a question of thinking about in tiers, thinking about what is this content, what is its longevity, what is it supposed to be doing.
00:11:49.600 --> 00:11:57.759
There are things that obviously you want to be in the driver's seat, but it goes back to like quality tiers that needs to reflect content tiering, right?
00:11:57.759 --> 00:11:59.279
But it's the same situation.
00:11:59.279 --> 00:12:08.799
The patent team has something very specific, they know exactly how it needs to be done better than any localization professional on the localization team.
00:12:08.799 --> 00:12:12.559
They already have a situation where they're getting their needs fulfilled.
00:12:12.559 --> 00:12:14.000
Do we need to get involved?
00:12:14.000 --> 00:12:14.879
Not really.
00:12:14.879 --> 00:12:20.720
There were no complaints, there were no quality issues, it was done to compliance, it was done the way it was supposed to be doing.
00:12:20.720 --> 00:12:22.080
Fine, get it done.
00:12:22.080 --> 00:12:24.879
Then you gotta ask yourself, why is it happening?
00:12:24.879 --> 00:12:45.759
I've always taken any type of initiative like this and any complaint as a flag that something in our setup is not working, something in our setup is leaving someone behind, and we need to figure out how to not leave them behind because otherwise they will find a way that is a different way.
00:12:46.080 --> 00:12:48.960
So it sounds like there's a couple of different situations.
00:12:48.960 --> 00:12:51.279
You respond to them on a case by case.
00:12:51.279 --> 00:12:53.919
How do you take all of these different things?
00:12:53.919 --> 00:12:58.480
So some you're cleaning up messes, sometimes you're just giving a little bit of advice.
00:12:58.480 --> 00:13:10.240
How do you organize your response so that at the end of the day you're ending up with a little bit more influence and more visibility than what you started with?
00:13:10.559 --> 00:13:11.600
Yeah, great question.
00:13:11.600 --> 00:13:15.840
So it is by reframing what your function is.
00:13:15.840 --> 00:13:26.080
If your localization team is positioned as that team that translates stuff, then of course everyone thinks that they can replace you with ChatGPT or any other AI tool.
00:13:26.080 --> 00:13:41.679
But if you're positioned as the team that has strategic advice and a strategic point of view on international markets that ensures that international experiences are coherent, they are on brand, and they're effective.
00:13:41.679 --> 00:13:49.600
Like a team that has like international intelligence that can be deployed and help you be better at what you're trying to do.
00:13:49.600 --> 00:13:55.519
That's a completely different value proposition that AI can't replicate on its own.
00:13:55.519 --> 00:14:00.080
Every time you clean up someone else's messes, you're collecting intelligence.
00:14:00.080 --> 00:14:08.159
You're learning which teams are going international that you didn't know of, what content they're producing, what markets they're targeting.
00:14:08.159 --> 00:14:24.720
And that's important organizational knowledge that you should have as a localization professional because you can take that to your own leadership and say there are X number of teams across the company doing translation work outside our function and our systems.
00:14:24.720 --> 00:14:27.759
Here's the quality risk, and here's what I propose.
00:14:27.759 --> 00:14:31.919
You're not going to your leadership to throw them under the bus and complain.
00:14:31.919 --> 00:14:34.720
You're complaining that they are being bypassed.
00:14:34.720 --> 00:14:44.000
You're going to them as a strategic move to present a strategic picture that you can see and nobody else is seeing.
00:14:44.000 --> 00:14:45.039
So it's an advantage.
00:14:45.039 --> 00:14:57.919
So the goal is not to control everything, it is to be the person in the company that understands the full picture of how your organization communicates across markets and across languages and across cultures.
00:14:57.919 --> 00:15:00.080
That is the definition of influence.
00:15:00.080 --> 00:15:11.519
And ironically, you will get there faster if you help people, no matter what happened and how they came to you, than by gatekeeping them or policing them.
00:15:11.519 --> 00:15:12.960
Nobody likes to be policed.
00:15:12.960 --> 00:15:13.519
Yeah.
00:15:13.759 --> 00:15:14.000
Okay.
00:15:14.000 --> 00:15:19.279
If you had to give someone one thing that you'd want them to take away from this chat, what would it be?
00:15:19.600 --> 00:15:32.159
It would be that if someone in your company is localizing without you, don't panic and don't get territorial, get curious instead, and use this as an opportunity to figure out what's happening.
00:15:32.159 --> 00:15:42.399
Clean up what needs to be cleaned up, be helpful, and use what you learn in this process to make your function definitely more visible and more essential.
00:15:42.399 --> 00:15:48.320
Yeah, I think that shadow localization is a symptom, not the problem necessarily.
00:15:48.320 --> 00:15:51.039
And it tells you how your function is perceived.
00:15:51.039 --> 00:15:52.320
So it's a data point.
00:15:52.320 --> 00:15:57.440
Use it as a data point and then use it to expand your influence.
00:15:57.440 --> 00:15:58.480
Okay.
00:15:58.879 --> 00:15:59.759
Thank you, Giulia.
00:15:59.759 --> 00:16:02.159
And I'm excited to continue this series.
00:16:02.159 --> 00:16:03.840
So yeah, we'll talk more next time.
00:16:04.240 --> 00:16:04.799
So am I.
00:16:04.799 --> 00:16:05.759
Thank you so much.