Automation Meets AI: Atlassian's Localization Evolution

Reconnecting with Melanie Heighway, Head of Internationalization at Atlassian, offers a fascinating glimpse into how localization teams can adapt and thrive during periods of rapid technological change. 

Four years after her first appearance in Global Ambitions, Melanie reveals how her team now leverages automation and AI to manage the localization of 30+ products across 20+ languages while maintaining quality standards with a lean team.

Episode Transcript

Stephanie Harris-Yee (00:21):
My name is Stephanie and I'm your host
today.
This time we have a specialepisode with Melanie Heighway
from Atlassian.
Now, if you've been with us fora while, you might remember
that Melanie was one of our veryfirst guests on this show.
We thought it was high time tocheck back in with her and see
what's going on in her world.
Welcome back, Melanie.

Melanie Heighway (00:41):
Thank you so much.
Thank you so much for having meagain.

Stephanie Harris-Yee (00:43):
It was a delight the first time, very
excited for the second time.
Very excited to hear what youhave to say and to hear your
updates.
So first off, let's see what'syour role at Atlassian these
days.

Melanie Heighway (00:55):
So my role hasn't changed too much since we
spoke about four years ago.
Actually, I leadinternationalization and product
internationalizationlocalization at Atlassian, and
my team ensures that ourproducts are available in over
20 languages, supporting ourcustomers across the globe.
We localize over 30 productsand platform areas into 20 plus

(01:15):
languages and we partner veryclosely with engineers,
designers and product managersto embed internationalization
into the product developmentlife cycle.
So my role these days is quitesimilar in that I am still
driving strategy, scalingoperations and ensuring that our
processes are very efficient,automated and aligned with
Atlassian's growth goals.

(01:36):
A significant part of my focusat the moment is around
leveraging automation and AI tostreamline our workflows and
reduce our manual effort anddeliver, you know, obviously
high quality localizedexperiences at scale.
One of the major changes thathas happened since I last spoke
to you is that we have movedorgs.
So I think when I spoke to you,m y team was sitting in the

(01:59):
go-to-market org and that'sright, and I was sitting next to
the marketing localization team.
Shortly after that, we movedinto the trust engineering
organization, so that was backover in R&D and what drove that
change was that, although wereally benefited from being in
go-to-market and being alongsidethe marketing localization team

(02:19):
, particularly in standardizingour tooling and our translators
and our processes.
We were too far removed from ourstakeholders, particularly
engineering, who we work themost closely with, and so we had
a roadshow essentially todifferent R&D engineering
leaders, product leaders, tofigure out the right fit, and at
that time we just started thetrust organization.
So that fit in really nicelywith the vision for the trust

(02:42):
org, which was very much aroundparts of the product that
customers expect to be in theproduct by default, so things
like security, reliability, alsoother teams like risk and
compliance.
So we were there for a coupleof years and most recently we
moved over to the TechnicalProgram Management Org, the TPM
Org, as we centralized our craftso we've moved alongside other

(03:05):
teams that are program managersof other technical areas.
So obviously very different tolocalisation but still running
technical program management,and we've been there for around
a year now.
But that's also within R&D andengineering as well.
So the wonderful thing is thatwe can partner very closely with
engineering and get thatinfluence.

Stephanie Harris-Yee (03:26):
Wow.
Well, that's a lot of changeshere and there through the
organization.
Well, I'm sure alongside thatthere have been big changes in
the industry all of thesedifferent things in the past
four years.
So what's kind of the mostchallenging thing that's come
across your plate in, say, thelast 12 months?

Melanie Heighway (03:42):
In the last 12 months?
Well, it has been a veryinteresting time since we did.
Last week, obviously, openaireleased ChatGPT I think it was
November 2022.
And that made huge waves intech, but obviously in our
industry as well.
So, linked to that, I would sayone of the most significant
challenges that I've faced inthe past year has been driving

(04:03):
engineering and processefficiency, the rapid growth of
Atlassian and also the explosionof tech and innovation.
That's happened globally and asthe company scales as well, the
demand for localisationservices has dramatically
increased and obviously withthat comes that expectation of
efficient and seamless processes.
Atlassian also has a very strongfocus on clock speed as well as

(04:25):
efficiency.
You know, obviously it's aboutcustomer experience, but it's
about making sure our processesare as efficient as possible.
So for us, that has been, youknow, transitioning from more
manual kind of traditional stylelocalization processes to more
automated and even AI drivenworkflows or leveraging AI to
help us with things that havebeen more manual in the past.

(04:47):
Also, the challenge is,particularly in our industry, is
keeping up with the sheervolume of localization words to
be translated with a very leanteam.
You know we haven't been ableto keep up using manual methods,
so the challenge is designingthose processes and making sure
that we keep that high bar forquality, but also making sure
that these processes scale toadapt to our company's needs.

Stephanie Harris-Yee (05:08):
So, as part of that challenge trying to
meet the scaling as well as theautomation and all these, do
you have any particular stepsthat you did in order to kind of
reach that and to maintainthose high levels of quality?

Melanie Heighway (05:21):
Yeah.
So the three areas we investedin to tackle these challenges
automation number one.
We've made significant stridesin automating repetitive and
time-consuming tasks.
So, for example, when it comesto context gathering, even
quality checks as well, soquality checks of the source
English we do a lot of that.
Even translators themselves,obviously they're taking a very

(05:43):
close view of it.
They pick up a lot of mistakesthat get through as well,
particularly even just verysmall nuanced language mistakes.
So leveraging linked scriptsand so on to automatically parse
our strings and do those checks, rather than have humans do
that and then also post checksas well when the strings come
out as well, identifyingincorrect pluralization and

(06:05):
whatnot.
But for context gathering inparticular, we've also made
strides to semi-automate that asmuch as possible.
Obviously, there's only a levelthat you can get context
automatically, so there's stillan element of manual process
there.
But we are making some greatefforts and some great
technology coming out in thatarea.
In terms of AI integration,we've had to spend a lot of time

(06:26):
traditionally in the past goingand gathering context or going
back to individual engineers,even content designers, to
gather context or have themrewrite strings that weren't
localizable.
So using AI to gather contextcan be very useful.
So even just within Atlassiansystems, we use a robo a lot.
If a string comes up, forexample, we're not sure what the

(06:47):
context is, we can say what isthe context of this product
feature in Confluence or Jira,and it will be able to search
our internal knowledge base andbe able to come back with.
This string is used here andthis relates to this feature,
which is so much faster thangoing and manually speaking to
someone, so that's been reallyhelpful as well.
It's also really helpful if astring has internationalization

(07:10):
issues, particularly when it'sin a different language and it's
say quite gnarly, quite long,has nested plurals or something.
Copy that into AI and say what'sthe internationalization issue
here and it can actuallydiagnose it very quickly.
So that makes it a lot faster.
You can also recommend whatneeds to be fixed so that we can
easily have our engineeringteam give them instructions and
have them fix that.
And then, lastly, in terms ofkeeping a high bar, establishing

(07:33):
standards with our teams,particularly with our content
designer teams, creates thecontent and our engineering team
as well about frameworks theyneed to use and best practices,
but also forging thoserelationships, so being very
visible in their orgs, going totheir town hall meetings,
getting to know them even.
You know, anytime there's achance to meet in person, such

(07:55):
as we've got an internalengineering conference, for
example, next week, a lot of theteams still go into the office
meeting up with them and justforging that great relationship,
getting to build empathy forwhat we're doing and ensure that
they give us great strings towork with so that we can deliver
a higher quality.

Stephanie Harris-Yee (08:10):
Well, that sounds like you have a lot of
different areas that you havegood steps on over the past 12
months.
So if someone is maybe in asimilar position, where they've
had a lot of changes happening,they're struggling with maybe
just one of these areas or allof these areas do you have any
recommendations on?
Hey, this is a good tip for you.
Try this.

Melanie Heighway (08:32):
Three things that I think I can think of here
as well.
Firstly, identify the mostlabor-intensive manual tasks.
So, particularly, you know,when you've got a very lean team
and you've got a lot of wordsto localize, look at your entire
workflow and look at wherethere's efficiencies to be
gained.
So say, for example, when itcomes to context gathering, as
we were talking about, always abig challenge in our industry.

(08:55):
But automation can reallyalleviate that, whether it's
writing a script to catchlocalizability issues or it's
just automating routine andrepetitive processes, just so
that you have that free time toactually work on the more meaty
and more impactful tasks.
And then, on top of that, interms of the time that you would

(09:15):
gain back, I would reallyinvest that into collaborating
with your stakeholder teams,particularly with engineering.
One thing that we've found youknow a lot of these manual tasks
that we do, we obviously don'tjust do them for fun.
We're doing them becausethey're needed, particularly if
the strings are coming to us andthey're not have localizability
issues.
They're not localizable.
We often, you know, go back andhave those rewritten and

(09:37):
whatnot.
But by actually going back andworking very closely with
engineering so that you know weestablish guidelines, we share,
that we also educate them.
We don't get those errors inthe first place.
Otherwise you kind of find thatyou're covering gaps and that
isn't particularly efficient foryour team and also the business
doesn't know they have aproblem if you're covering the
gaps.
So making sure you have thosepartnerships and you're

(09:59):
educating those stakeholders tomake sure that they can help you
with what you're doing, andthen everybody will be
successful.
And then, lastly, I'd say,making sure that you have robust
data tracking so that you canactually see what you're facing
Like where am I localizing?
Where's the highest amount ofvolume that I'm localizing?
Is it Confluence?
Is it Jira?
Oh, it's over here.

(10:19):
I wouldn't have expected thatfor that product.
Is that the highest prioritythat I should be working on?
But even you know, looking atthe usage in different regions,
which is obviously very standardin our industry, to see, you
know number of people who areviewing localized UI, but even
just looking at the number ofwords that you're localizing and
revenue that's coming in, anykind of data analytics that you

(10:40):
can access is always reallyhelpful, particularly when
you're trying to measure impactas well and make improvements.
You want to know where you'restarting and what the goalposts
could be, and that's also a verycompelling story to the
business that you can tell interms of saying, well, if you
invest here, I can deliver thisoutcome, and by tracking the
baseline and having the datathere, that will help you do

(11:01):
that.

Stephanie Harris-Yee (11:02):
Well, we're coming close to the end of
our time, so I do have onefinal question for you.
With all of these changes thathave been happening in the
industry, do you have anythoughts on how localization
leaders like yourself can reallycontinue to thrive?

Melanie Heighway (11:19):
I think in today's rapidly evolving
industry, we all must embracechange and really stay on top of
the great tech and innovationthat's coming out of our
industry, really just to keep upwith what our company needs
from us.
And, as we just talked about atthe start of the recording,
things are evolving very rapidlyand so it's really important, I
think, to keep up with all thenews.
Talk to different companiesproviders, vendors, technology

(11:41):
vendors talk about what they'reworking on and what's coming out
.
Going to webinars andconferences is really useful as
well for those kinds of thingsas well, because essentially, we
would need to tap into all ofthe resources that we have, all
the tech that we have to keep upwith speed that this industry
is going at, tech in general, sodefinitely keeping up with the
evolutions in our industry andscalability as well.

(12:03):
So companies are obviouslyexpanding their global reach and
it's always really important tokeep that in the back of your
mind about how you scale thiswithout compromising quality and
keeping up with such a vastamount of volume with such a
lean team.
So making sure that you buildflexible and robust processes
that can adapt to the growingneeds of your company and,

(12:25):
obviously investing in scalabletechnologies and language
service provider partners tohelp you reach those goals as
well.
Lastly, never stopping to pushthe strategic importance of
localization internally as well.
We always have to have thatseat at the table an executive
sponsor, and making sure thatthat's at the top of people's
minds so that we can ensure thatwe're fully built into the

(12:48):
process and part of the process,because if you want to build
global products, you have to bea part of the global process
that the company is using.
So definitely those threethings would be the pieces of
advice that I would give.

Stephanie Harris-Yee (13:02):
Wonderful.
Well, thank you so much,Melanie.
This has been truly a pleasureto have you back with us again.