March 20, 2025 • 37 mins
Nicole Janssen on Bridging AI Innovation and Government Solutions
Nicole is co-CEO of one of Canada's AI powerhouse organizations, AltaML. She's a national business leader and one of our experts in AI.
Her firm is also behind the Alberta government's innovative AI research lab, operating as an R&D hub for the big, complex problems that the government hopes AI can help address. If you're not from Canada, and maybe even if you are, you may not know that Canada is an AI pioneer. Some of the earliest technology that is powering this new digital economy was conceived of and developed in Canadian universities.
Canada has always had a very strong bond between higher education, the public sector, and industry when it comes to innovation. AI has grown here in this collaborative ecosystem. If you're feeling left behind by the pace of change in artificial intelligence at the moment, then I think you'll find Nicole's overview and perspective helpful.
She lays out what is happening in the AI industry and where things are going, and she also has great advice for government organizations just trying to get a toehold.
Resource Links:
Guest
Nicole Janssen, Co-Founder & Co-CEO, AltaML - LinkedIn
AltaML: AltaML LinkedIn
GovLab
Canadian AI Institutes
AI for Wildfire Prediction
Digital Transformation at Scale (Book) - Amazon.ca | Indigo | Kobo
This is a show about the people that make digital public service work. If you'd like to find out more, visit northern.co/311-podcast/
We're going to keep having conversations like this. If you've got ideas of guests we should speak to, send us an email to the311@northern.co.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Paul Bellows (00:30):
This is the 311 Podcast, and I'm your host, Paul
Bellows.
This is a show about the peoplethat make digital work for the
public service.
If you'd like to find out more,visit northern.
co.
Today, my guest is NicoleJanssen.
Nicole is co CEO of one ofCanada's AI powerhouse
organizations, AltaML.
(00:53):
She's a national business leaderand one of our experts in AI.
Her firm is also behind theAlberta government's innovative
AI research lab, operating as anR&D hub for the big, complex
problems that the governmenthopes AI can help address.
If you're not from Canada, andmaybe even if you are, you may
not know that Canada is an AIpioneer.
(01:14):
Some of the earliest technologythat is powering this new
digital economy was conceived ofand developed in Canadian
universities.
Canada has always had a verystrong bond between higher
education, public sector, andindustry when it comes to
innovation.
AI has grown here in thiscollaborative ecosystem.
If you're feeling left behind bythe pace of change in artificial
(01:36):
intelligence at the moment, thenI think you'll find Nicole's
overview and perspectivehelpful.
She lays out what is happeningin the AI industry and where
things are going, and she alsohas great advice for government
organizations just trying to geta toehold.
Here's my conversation withNicole Janssen.
(01:57):
Nicole, first just, first name,last name, title, just a little
bit about what you do here atAltaML.
Nicole (02:01):
Nicole Janssen co founder and co CEO of AltaML.
At AltaML, we have two sides toour business.
On one side, we're a venturestudio, and we create different
ventures, and we are thetechnical co founder.
And so we support the buildingof those ventures; we now have
seven ventures that we havecreated.
(02:22):
On the other side of thebusiness, we do services with
both public and privateorganizations to build custom AI
solutions using their data andaddressing their specific
problems.
And support them in that AIjourney from wherever they're
at.
Maybe they're at the educationstage, or maybe they're further
along and they've got some usecases identified.
(02:44):
We can start wherever anorganization's at and help get
them to value.
Paul Bellows (02:50):
So AltaML is, named after the province of
Alberta, where we're sitting andspeaking today.
And we're here in the Canadiancontext and Canada has a long
history with AI.
A lot of the technology that ispopularized today and is
capturing a lot of people'simagination today was imagined
and pioneered here in Canada.
(03:10):
And you're really at the heartof what's going on in the AI
community here in Canada.
You're one of our experts interms of where you sit and what
you can see.
Maybe a little narrative ofCanada's role of, the early days
of AI and where some of thistechnology came from right here
in our backyard.
Nicole (03:25):
The interesting thing is Canada was investing in AI long
before it was sexy.
Back in the seventies, we wereheavily investing in this
technology and have been alwayssince then.
We have three of what are calledthe godfathers of AI that are in
Canada.
Located currently in Edmonton,Montreal and Toronto.
(03:48):
And they associate with thethree different AI institutes
that we have, Amii, Vector andMila, that are really at the
forefront of research.
I have to tell you the level ofresearch and the talent
generation we have here is it's,you can't compare it, because
we've been investing in it forso long in a very Canadian way,
(04:10):
humble, quietly, trucking along,just getting shit done.
And the interesting part is mostpeople, even myself, who I was
in tech, and 2018, living in acity that has one of the AI
institutes.
I had no idea that the AIinstitute existed.
(04:30):
I had no idea that we had one ofthe godfathers of AI at our
University.
I had no idea that people fromaround the world fight tooth and
nail to have a spot at theUniversity to learn about
machine learning and AI.
No idea.
I think most Canadians are inthat same camp.
We are at a place where we havethis absolutely amazing
(04:54):
foundation that today, honestly,if we tried to start today, we
could never get here.
Because of the dollars it wouldtake to get us here, because of
the hype now.
The talent that comes out ofeach of these universities that
focuses on AI is incredible.
Bar none, that is why we have anAI company here, because it is,
(05:18):
while it's one of the hottestjob titles to hire, we don't
have difficulty hiring, that jobtitle, because of the flow of
super incredible talent and therelationship that we have with
the different universitiesacross Canada.
We do a lot of internships, etcetera, and we have
relationships with some of thebest profs who will say, I know
(05:40):
AltaML offers an exceptionalexperience for an internship.
And so they'll send their beststudents to us.
And we end up having just thisincredible pipeline that we
actually also leverage for allof our clients because those
interns are working on ourclient projects.
And so often they become ourclient's talent just because
(06:01):
we've had such an effectiveapproach to the internship
program.
Paul Bellows (06:06):
I want to get into the next level of some of the
work that you do here at AltaML,but I think it'd be helpful.
Large language models and theparty trick at scale that has
appeared in our culture recentlyof, robots that will talk to you
and answer questions and writeessays, et cetera, it's
engaging, it's compelling, butit's not the sum of what AI is.
AI is a whole net of things.
(06:28):
Is it fair to ask, could yougive us a short primer on how do
you think about the categoriesof AI and what is possible out
there?
I think it's helpful to skippast the thing that has
everyone's imagination right nowand remind folks there's a lot
of different things that are AIwhen we talk about that.
It's category.
Nicole (06:45):
First I have to give a little background on me.
When we launched AltML in 2018,we saw an opportunity in this
space.
My co founder is quitetechnical, I am not.
So in, in 2018 I knew that AIstood for Artificial
Intelligence, and that was theextent of my knowledge.
And often, I think that peoplelike to hear me talk about it,
(07:08):
because I have to talk about it,not in a technical way, but how
I've come to understand itthrough more of a business lens.
I laugh, actually, because Ithink of it in three categories,
but one is called and referredto often as classic AI.
That's hilarious to me, because,the capabilities we have within
(07:31):
this classic version of AI areincredible and nowhere near
adopted to the level that theyshould be.
So don't feel like you're out ofdate if you're working on a
classic project.
And things that fall in there,computer vision, natural
language processing, those typesof things.
(07:52):
Then you've got, the newerversion that came out in 2021
with ChatGPT, the generative AI.
And so you're creating, someform of content, whether it's
software code, whether it'swritten text, videos, that's
new.
And so you're building somethingthat didn't exist before,
whereas the other versions areanalyzing the data that exists,
(08:16):
learning the rules of how tomake best decisions in those
areas.
But then, now we've got theagentic starting, and this is
the new thing around taking thatgenerative piece and adding that
next component to it whereyou're saying this is what we
(08:37):
should do and now I'm going tohave all of these agents go do
it on my behalf.
So those are how I bucket thethree.
As of 2021, everybody wanted totalk about just generative AI.
As of now, everybody only wantsto talk agentic AI.
And what I see is, let's findthe tool that solves the problem
(09:00):
that you're trying to solve.
Let's not ask the question, whatcan I do with AI?
But rather ask the question ofwhat business challenges am I
facing?
And is one of these types andversions of AI the right tool to
get me to where I need to be?
Because if you're just lookingto build cool AI, you're wasting
(09:21):
your time.
Paul Bellows (09:22):
Which is why you need a shop like AltaML to help
you traverse the broad landscapeand start to actually connect
technology to business problems,right?
Nicole (09:30):
Yes.
Paul Bellows (09:30):
Yeah, so so this gets interesting I want to talk
now a little bit about thebusiness model of how you work
here at AltaML because it's oneof my it's one of my favorite
things about what you do is howyou do it.
You're not a traditionalsoftware firm.
You talked a little bit aboutyou have your venture studio,
and then you have your servicesside where you do things for
people.
Can you go a little bit, justsome examples of on the Venture
(09:52):
Studio side, what do thosepartnerships look like?
How do you work alongside yourcustomer or your partner in that
context versus the relationshipon the services side?
Nicole (10:02):
So in the Venture Studio, we are truly co founders
with the founders.
Sometimes these ideas will comefrom within our team, and we
will put a CEO in place.
But in most cases, this is afounder who's come to us with an
idea, they don't have thetechnical background, so they
(10:22):
have an exceptionalunderstanding of the industry
that they're trying to solve theproblem in, they understand that
this is a problem we're solvingand they have the expertise to
run this, they just don't havethe technical piece.
So we bring in the technicalpiece and become that co founder
at the technical level.
We also support, backend thingslike HR and finance, and those
(10:47):
things that often distract a CEOat the onset and suck up a ton
of time.
We've got a strong team that cansupport in all of that.
So that's not where that founderhas to focus, and that, we can
really just hone in on whatwe're trying to build and get it
built and get it into market.
And then on the services side wework with our clients in a way.
(11:10):
We work in a couple ways.
So one is we'll work with aprivate sector organization
where we really help them honein on what are the most
impactful use cases in whatorder, and also keeping in mind
that I would venture to say that70 percent of an AI project is
(11:30):
the change management piece.
And the technical component isnot as large.
And we also help organizationssee that maybe IT is not where
an AI project should sit.
It needs to partner with IT, butwe actually have to truly
understand the business problem,who the end users are, who will
(11:52):
be impacted by this, so that itactually gets into operation.
So we work that way.
We also work, especially in thepublic sector, in a very
collaborative way, where we say,okay, we have a model with the
Government of Alberta, forexample, it's called GovLab and
we work within that.
They allow public organizationsto collaborate with them.
(12:15):
So we've had the City ofEdmonton, we've had the City of
Calgary all involved in this andthe IP that's built can be
shared.
Why would the City of Calgarypay for something, and then the
City of Edmonton pay forsomething?
This is all taxpayer dollars.
Why don't we, one of them buildit with us, and then the other
one can leverage it.
(12:37):
And so that's how GovLab hasworked, is it's allowed that IP
to be shared.
And so that we're not all havingto pay a bunch of extra money
with our tax dollars.
Instead, let's do this together.
Paul Bellows (12:50):
So the GovLab piece is what I was really
excited to get to in ourconversation here because I
think it's a really innovativemodel and it's something that I
think public sector folks acrossNorth America should be paying
attention to as an example ofhow to bring innovation into
government.
Now innovation always, it's aword that makes anyone who's
been in government long enoughhears the word and they cringe.
(13:11):
But, I like to, I'm a bit of anetymologist, so I like to go
back to the root of things.
And it really just comes fromthe root of renewal.
So we talk about innovation andsay, how else could we do this?
And I love that you alreadycalled out, AI is so much more
often business technology thantraditional IT technology.
It's really about do weunderstand the business problem.
Can you give me an example ofwithin, I want to talk a little
(13:34):
bit about how GovLab is working.
It's the next level ofunderstanding there.
But maybe the best way to getinto that is what kinds of
problems are you able to solveat GovLab the government would
not be able to solve if theywere trying to do this on their
own, work independently from anorganization with kind of
expertise for how to solve theseproblems.
Nicole (13:52):
So a really important aspect of the GovLab model is
that this is not traditionalprocurement.
We don't go project by project.
Instead, we look at AI as awhole across the government.
And we have, we support all thedepartments in coming up with
ideas and use cases, help themsuss out what's the feasibility,
(14:16):
what's the date, do we have thedata, what's the ROI on this,
all of the different pieces, andthen they pitch it to the
governance group, who decideswhich use cases are going to get
funded for the next quarter.
And so you're not going to getfunded forever and always until
it gets into operation, you'vebeen funded to see over the next
(14:38):
quarter.
Are you worthwhile continuingafter that quarter?
So that allows us to cut offprojects that aren't showing ROI
and get rid of them.
But it also allows us to hone inon the best projects and really
put focus on those.
Some of the projects that we'veworked on, I have two favourites
that I'll share with you.
One is sexy, one is absolutelynot sexy at all.
(15:01):
So I'll start with the sexy one.
It's around wildfires.
So we can predict now, with an80 percent accuracy, 24 hours in
advance, where a wildfire willbe in Alberta.
That allows the resources to beeffectively stationed where they
need to be ready to fight thatpotential fire.
(15:21):
That also allows a significantsavings in overtime of, gosh, I
don't know, it seems like, thewhole province might go up
tomorrow, I'm going to getpeople everywhere.
This really allows the dutyofficers to use their experience
combined with the model to say,okay, We're going to focus in
these three areas, or whereverit might be, and we're going to
(15:45):
keep our resources in thoseplaces.
Because every time you moveresources for firefighting,
you're at risk.
You're moving helicopters,you're moving all sorts of
things.
Those are not things that youwant to be moving a lot.
You want to make sure they're inthe right places and ready to
fight the fires that come.
Now the interesting thing aboutthat model is we've just started
the next version of it where wecan, we're looking at the fuel
(16:08):
grid.
How many dead trees, how manydiseased trees are in these
areas and can we be doingpreventative maintenance?
So that if we get a wildfirestarted in that area, it won't
be quite as dramatic anddamaging as it would have been.
We're combining the two to bethis, information system for
duty officers to be making thebest decisions.
(16:30):
That's my sexy one.
My not so sexy one is aroundsafety codes.
So we have a lot of safety codesin the province.
Not the best reading unlessyou're wanting to fall asleep.
But there's a lot of questionsthat come in around safety codes
from builders, developers, etc.
And they're often waiting onthose questions to move forward.
(16:54):
And so the wait time was oftenweeks, sometimes up to a month,
waiting to find out the responseto these questions because it
was a lot of information to gothrough.
And they had to have highlyskilled engineers.
providing these responses.
We've created a model, agenerative AI model, that
(17:16):
produces that response, that canonly pull data from the existing
safety code, so you're notgetting the, the issues of
ChatGPT, where you have no ideaof what it's saying is true.
Instead, we're saying, you canonly provide examples from this
content, and it gives the sourceof this, which safety code it's
(17:36):
in, what clause it is, etcetera,drafts the email.
We've got a human in the loopbecause that's an important part
of responsible AI to take alook.
Is this the right response?
And so now response time is twominutes instead of many weeks.
And those individuals who arehighly trained engineers, are
now actually getting to dosomething in their work that's a
(17:59):
little bit higher value and whatthey were actually trained to
do.
And so that has been a big winfor that department, and we see
that opportunity there to takethat and expand it across really
anywhere that has legislation orregulation that needs a
response.
Because if you think about howmany things and categories we
(18:21):
have of that in government, howmany places could this be an
impactful model?
So not so sexy, but super, superimpactful.
Paul Bellows (18:30):
Useful though.
Nicole (18:32):
Yes, very useful.
Paul Bellows (18:33):
So useful, I mean it's just, if you think of that
as a paradigm of government hasdata, the data is complex, it
needs interpretation, or search,or discovery, and you've got
very high paid people, whichmeans you can't have that many
of them, because you just can'thave everyone sitting benched,
and when things get busy, itslows down business, and having
done some commercial real estatedevelopment in the past, just
(18:54):
through offices opening things,every day you're waiting for an
answer, costs go up.
Nicole (18:59):
Yes.
Paul Bellows (18:59):
Openings are delayed, I love this, that's a
brilliant example.
Anyone who's ever had to getsomething done alongside
government would appreciategovernment having access to
those tools.
Nicole (19:09):
Yes.
Paul Bellows (19:10):
So I want to just a couple of questions one thing
you said that I love it I justwant to put some punctuation
behind is I love you talkedabout procurement and how this
is a different model ofprocurement.
You talked about the abilitywhen something's not working we
can stop, which is remarkable ingovernment procurement.
Usually, you do a public RFP,you buy the thing that you need,
(19:32):
and then you've bought it, andyou have to see it through to
the end, right?
It's hard to stop once you'vebought something and struck a
contract.
This is a structure in which badideas can fail fast.
And without public visibilityand people getting embarrassed,
it's just, we were all sure thiswas a good idea.
We were wrong.
We can be wrong without spendinga lot of time and money or
(19:53):
having to build something thatwe know people will never use.
That's just one of my favoriteanecdotes you've pulled out here
so far.
Tell me a little bit more, theGovernment of Alberta has said
we're going to put somepermanent, time based funding in
place, to create a certaincapacity for AI work, on a
quarterly basis.
You say, how are we going tospend that capacity?
What problems are we going tosolve?
And you have to continuallyjustify the value that's getting
(20:15):
produced, am I understandingthat correctly?
Nicole (20:16):
Yes, definitely.
Paul Bellows (20:18):
And have you seen this happening anywhere else in
Canada, the U.
S.?
Is this a novel thing happeninghere in Alberta, or is this
something more uptake withingovernments?
Nicole (20:26):
I haven't seen it elsewhere, but I can tell you
that every government that Italk to is excited and
interested about it.
We're talking with a bunch ofprovinces and we have talked to
some states in the US as well,where there's, they see how
difficult it is to do AI ifyou're not either buying it off
(20:51):
the shelf, which often is notgoing to get you what you need.
Sometimes it is, and then youshould just do that.
But a lot of times it's not andthen they look at their existing
procurement.
And they think this isimpossible.
How, because you can't, youdon't know what the end state is
truly going to look like with AIuntil you get started.
(21:12):
And you actually see is the datagoing to give us the accuracy we
need?
And then if it isn't, is thereanother way to solve this
problem with a differentapproach?
You have to have that agility.
And so because that hasn'texisted really before in
traditional procurement, It'sjust so incredibly challenging.
And so I would say that, I wouldsay a year from now, we will
(21:36):
have, you'll see a whole bunchmore governments starting to
deploy this model.
Hopefully with us.
Paul Bellows (21:42):
They're all going to hear this podcast.
Yeah, absolutely.
So next question then.
So it's stable funding, thegovernment can control the
spend, see this is the amountwe're willing to commit.
We get as much work done withinthat as we can.
We produce models, we produceintellectual property.
Who owns that, at the end of theday?
Nicole (21:59):
This is another interesting part of the model,
is that the government owns theIP.
But the Government of Alberta,for them, it's very important
that they're a leader in thecommercialization of AI because
Canada really has struggled inthat area.
And so they have allowed us tohave the ability to license that
(22:20):
IP and build products on top ofit and commercialize those as
long as they get a share in theprofits.
Their goal is over time, andwhen you're building products,
that doesn't happen overnight.
So we don't have any fully inmarket yet.
But once we build thoseproducts, and we have a few that
we think are just about there,then their AI program will start
(22:43):
generating revenue.
And their hope is, over time, itbecomes self sustaining.
And in fact, they're not puttingany money into AI, other than
the money they're getting backfrom it.
And so that is a reallyinteresting approach to AI that
I think is, you can't expectthose returns immediately, but
with the long term plan, thatcan be really advantageous.
(23:06):
Then the other piece that wasreally important that I haven't
talked about is the talentdevelopment, because with with
the public sector, let's just behonest, it's not the first stop
for your newest grads coming outof tech with the hottest, job
title.
But a startup or a scale up likeus is and so that's why we have
(23:28):
no problems attracting teammembers for our internship.
Then they work on these projectsin government and realize, wow,
there is some incredibleproblems that government can
solve in AI because the datathat is here is incredibly
valuable and can really solvevery complex very interesting
(23:51):
and impactful problems.
So, every person that theGovernment of Alberta has hired
in this area has now comethrough GovLab as an intern.
And so this has become theirtalent pipeline.
And then whoever doesn't becomea Government of Alberta or an
AltaML staff person goes intothe ecosystem.
(24:12):
And is someone who doesn't justhave the academic background,
but they have real worldexperience developing these
solutions and getting them intooperation.
Which is valuable to our, to ourmarket.
Paul Bellows (24:25):
One of the things that I think, as a practitioner
in the digital software spaceworking alongside government,
one of the things that I'vealways found is the most
fascinating is just the sheerscale of what government does
and does every single day.
The scale of the data, the scaleof the operations, the scale of
the staff, the organization, thebudgets.
Government is a large, behemoth,which I think often makes it a
(24:48):
target for why does it cost somuch, that'll be a separate
conversation that I don't liketo weigh in on, I'm not
qualified.
But, to say, hey, working withgovernment, you get access to
data, but also to do things thatimpact real people every day,
like it's useful what you get tocreate when you do good things
with government.
I think that's really excitingas a talent pipeline.
And again, some of the talentthe government wants to hire,
(25:11):
you need to build a relationshipbecause what the market's
offering for salaries andcompensation, in start ups and,
venture backed organizations,versus what government can offer
in compensation, they're justdifferent categories.
But sometimes you fall in lovewith the work, and maybe a lower
comp, but more stability, morepatience of an organization,
longer timescales, and theability to have impact, I think,
(25:32):
can be really powerful for youngpeople.
who look around and say, I wouldlike to contribute to the
betterment of the world ratherthan just pad my own bank
account.
What, how can I work?
I think that's really exciting.
Nicole (25:42):
We're doing a project with Alberta Cancer Care right
now that's a part of the GovLab.
And we're working on how do weget individuals who've been
diagnosed with cancer quickeraccess to oncologists?
How do we remove the wait timeand start reducing it?
Now, obviously, eventually,we'll roll that work out,
hopefully, across the healthcaresystem, but we're starting with
(26:05):
cancer.
Now, imagine coming into aninternship and getting to work
alongside that, thinking, mywork is impacting the wait time
of somebody with canceraccessing the system.
That's incredible.
Paul Bellows (26:21):
What
Nicole (26:21):
a thing worth getting out of bed for in the morning, I
love it.
That's brilliant.
Something that maybe peopledon't understand clearly is what
is the skill set you're hiringfor?
You're talking about internsfor, and again, for folks who
don't really get into the weedsof what AI is and the types of
technology, what kinds of skillsets are you able to bring to
government?
Paul Bellows (26:41):
Are these design people, coders, math people, all
of the above?
Can you talk a little bit aboutthe kind of talent that would
sit within GovLab?
Nicole (26:50):
We would have software engineers we would have machine
learning engineers but we alsohave both internships and a lot
of roles in the business side.
Some of the most valuable peopleare those who can listen to the
business side and translate tothe technical side.
I refer to them as ourtranslators, but they're,
(27:12):
project delivery, they'reproduct owners, they're those
individuals who might not be theperson doing the code.
But they can help dive down intoexactly the right problem to
start solving that's going tobring value.
And then communicate that to thetechnical folks so they can dive
in right away.
Because if you just handtechnical folks a bunch of data
(27:34):
and say, What cool stuff can youbuild with this?
I'm going to be honest, a lot oftimes there's not a lot of value
there.
There's a lot of cool things!But it might not be the most
valuable thing to theorganization.
And so those individuals arereally important too.
The skill set that thoseindividuals need to have is an
awareness of the capabilities ofthe technology and then being
(27:58):
able to really dive in and digdeep into understanding a
business problem.
Paul Bellows (28:04):
I think one of the most important questions that
many technologies fail to answeris, what should we do?
There's a problem.
Do we understand it?
Have we analyzed it?
Have we gotten to the root ofthat problem?
And then, yeah, we have a lot oftechnology, but those are just
tools.
They only ever tell us how, notwhat and not why.
Yeah, the people who canconnect, a problem space to a
(28:25):
technology space and create abridge, that's powerful.
I'd like to think that's whatI'd use.
I'm glad to hear that's usefulin the world.
I want to talk just for one moreminute, coming back to where we
started.
So historically, and there's somany parallels with everything
Canada has ever done, which iswe're really good at resources.
We have all these wonderfulresources and we've always
(28:46):
relied on other people tocommercialize those resources to
some degree, that it's the, ourgreat critique is, and then we
ship it off and someone elsebuilds the house out of it.
Someone else upgrades our energyand does things, and it sounds
like there's a bit of thathappening with AI.
We've pioneered a lot of this.
We've invented a lot of it.
We've got these deep wells ofexpertise here.
We've gotten used to havingthese people here.
We have these ecosystems thatproduce more of them.
(29:08):
So we have all of this, but wehaven't always harnessed it to
produce value, wealth,solutions, whatever you wish to
have at the end of the day.
So as you sit where you sit,working alongside government, In
a moment where I think Canada'sbeing challenged to level up in
2025 here and get to the nextlevel.
What do you see as some of theproblems that we need to solve
(29:29):
in the business ecosystem, inthe technology ecosystem, in our
public service ecosystem?
What do you think we need toconfront as governments,
provincially, municipally,federally, right across Canada
right now?
How do you, see that?
Nicole (29:42):
So I think that the two biggest challenges are both
commercialization and adoption.
On the commercialization side itbecomes tough when your
investors and your customersaren't here.
It's very easy to raise capitalin, in AI in the US.
It's not so easy in Canada.
(30:02):
It's also, to give perspective,our sales cycle.
In Canada is 18 months, in theU.
S.
it's four.
I am a proud Canadian and wewill always be, a Canadian
company.
But, we also, do work in theU.S.
because that's a big differenceto a business, right?
(30:24):
You can't ignore those facts.
And how do we as as a countrydecide who we want to be and
what we want to do, really,frankly, for everything, but
specifically in AI, what are thelayers that we want to be
exceptional at?
Because there's a massive supplychain to AI.
There's the infrastructure,there's the foundational models,
(30:48):
there's the application layer,there's all of these different
pieces.
Are we trying to be excellent atall of them or some of them?
What's our brand going to be?
And I, I think the federalgovernment is really tired of
hearing from me continuallysaying our brand needs to be
that we're the global leader ofresponsible AI.
And if you buy AI in Canada, youknow it's ethical and built with
(31:10):
responsible AI principles.
That's a brand that aligns verywell with who we are.
But, that's a side.
I have to tell everyone I evertalk to about that because
someone will start listening.
But, so that's the kind ofcommercialization side.
But then the adoption side alsorelates to it because it's about
(31:31):
the customer side.
A piece of commercialization.
With that 18 month sales cycleversus four months, with AI, the
impacts it starts having on anorganization are very quick.
Once it's in operation, itbegins to make an impact
immediately.
But the life cycle of getting tooperation, isn't short with AI,
(31:56):
it is an investment of timebecause you can build this
really cool thing that solvesthis really impactful problem,
but there's change that has tohappen in order to use this
tool.
So with the wildfire tool as anexample, that is a seasonal
tool.
We do not need that tool in thewinter.
So you can only really build it,and perfect it, and test it in a
(32:20):
certain part of the year.
And that first day it's ready,do you think the duty officer's
sure, let the AI model decidewhere everything's going to be."
I can tell you not a chance.
And so they needed an entirefire season to work alongside
the model where they made allthe decisions based on their
experience and then looked atwhat the model said.
(32:44):
And yes, that either validatedor it didn't, and then they
waited to see what did happen 24hours from then.
Was the model right?
Or was it wrong?
Would I have made the wrongdecision had I, followed the
model?
Or did I make the wrong decisionand should have followed the
model?
And so after the season, thenthey were ready for the next
season to bring it into place.
When you're talking about ashort season that only comes up
(33:06):
once in a year, Think of howlong that cycle was to get that
to a place where it was addingvalue.
And so we have to start, we haveto get started building these
things.
We can't stop, which a lot oforganizations do that we see in
Canada, of let's educate ourboard, let's educate our
executive teams, and then we'llthink about it.
(33:30):
And we'll spend the next 18months getting our data in
order.
That's what everybody says.
Which I laugh because You, howcan you get your data in order,
unless you know how you're goingto use it?
You can start AI projects with asample of data to see if you're
going to get an accuracy that'sworthwhile.
And then you'll realize, oh I,you know what we need?
(33:51):
Is this one other piece of data.
And if we had that collected,that would make this model
incredibly powerful.
You've never been collectingthat.
We need to start collectingthat.
And they want to make all thesedata decisions and make that,
have everything perfect, by theway, data is never perfect and
then start with AI.
(34:12):
You can't do that.
That does not work.
And I, what I'm pushing, Canadaneeds to start adopting.
Stop waiting.
We've always been hesitant toadopt new technology.
The impact and the risk of notadopting AI is so massive.
In order to compete and stayproductive and efficient and,
(34:34):
frankly, to make all industriesand government across Canada
able to compete in this crazyworld, we are going to have to
get going.
Paul Bellows (34:46):
Some of the pioneers of digital government
were some of the folks in the UKand they wrote a book out of it
called Transformation at Scaleabout just how they did it.
And the core mantra out of thatis, you start by starting.
So last thought, Nicole, thishas just been a wonderful tour
of what's happening here inAlberta with the GovLab and what
you're doing at AltaML, a bit ofthe state of AI.
(35:07):
So start by starting, I'm in aprovincial government
leadership, I'm a CIO, I'm abusiness leader, I'm the federal
government, I'm in a large citythat needs to tackle AI, what's
an action that can be taken?
Nicole (35:18):
Call me.
Paul Bellows (35:19):
Clearly.
Nicole (35:21):
But I think it is just starting to take action towards
doing.
Ask yourself the question of howare we going to get something
embedded, built, whatever itmight be, so that it's in the
hands of users.
Don't focus just on theeducation component.
Think about the end state andstart working towards that.
(35:43):
Rather than just, first we'lleducate everyone.
Yes, let's educate everyone allalong the way.
But we gotta get going ongetting something in the hands
of users.
Paul Bellows (35:53):
It's time for Canada to move fast.
Nicole (35:55):
Yes.
Paul Bellows (35:56):
Thanks, Nicole.
I appreciate the time.
Nicole (35:57):
Thank you.
Paul Bellows (36:02):
Thanks so much for joining us for this
conversation.
Nicole is both a nationally andinternationally recognized
innovator and entrepreneur, butshe's also a humble expert,
clearly.
That's a compelling combination.
Some of the key themes shebrought forward that I
appreciated include Canada hasbeen investing in AI innovation
for decades, and now that therest of the world is catching
(36:25):
up, we still have a competitiveposition from which to thrive.
When you build the rightstructures for bringing problems
together with experts andtalents, you can achieve great
things.
There isn't time to waste.
Government needs to move quicklyto understand this space.
And as always, we start bystarting.
(36:46):
And finally, AI requires talent,patience, and new business
models.
Alberta's GovLab is one greatmodel, and if you want to learn
more about it, Nicole would loveto take your call.
I hope you enjoyed thisconversation with Nicole Jansen.
Please do subscribe and followthe many conversations we're
going to be releasing throughoutthe year.
(37:07):
I'd like to thank my colleagueswho work with me on this
podcast.
Kathy Watton is our showproducer and editor.
Frederick Brummer and AhmedKhalil created our theme music
and intro.
We're going to keep havingconversations like this.
Thanks for tuning in.
If you've got ideas for guestswe should speak to, Send an
email to the311 at northern.
(37:28):
co.
Remember, the public service isabout all of us, and when it's
done right, digital can be a keyingredient for a better world.
This has been the 311 podcast,and I'm your host, Paul Bellows.
This is the 311 Podcast, and I'm your host, Paul
Bellows.
This is a show about the peoplethat make digital work for the
public service.
If you'd like to find out more,visit northern.
co.
Today, my guest is NicoleJanssen.
Nicole is co CEO of one ofCanada's AI powerhouse
organizations, AltaML.
(00:53):
She's a national business leaderand one of our experts in AI.
Her firm is also behind theAlberta government's innovative
AI research lab, operating as anR&D hub for the big, complex
problems that the governmenthopes AI can help address.
If you're not from Canada, andmaybe even if you are, you may
not know that Canada is an AIpioneer.
(01:14):
Some of the earliest technologythat is powering this new
digital economy was conceived ofand developed in Canadian
universities.
Canada has always had a verystrong bond between higher
education, public sector, andindustry when it comes to
innovation.
AI has grown here in thiscollaborative ecosystem.
If you're feeling left behind bythe pace of change in artificial
(01:36):
intelligence at the moment, thenI think you'll find Nicole's
overview and perspectivehelpful.
She lays out what is happeningin the AI industry and where
things are going, and she alsohas great advice for government
organizations just trying to geta toehold.
Here's my conversation withNicole Janssen.
(01:57):
Nicole, first just, first name,last name, title, just a little
bit about what you do here atAltaML.
Nicole (02:01):
Nicole Janssen co founder and co CEO of AltaML.
At AltaML, we have two sides toour business.
On one side, we're a venturestudio, and we create different
ventures, and we are thetechnical co founder.
And so we support the buildingof those ventures; we now have
seven ventures that we havecreated.
(02:22):
On the other side of thebusiness, we do services with
both public and privateorganizations to build custom AI
solutions using their data andaddressing their specific
problems.
And support them in that AIjourney from wherever they're
at.
Maybe they're at the educationstage, or maybe they're further
along and they've got some usecases identified.
(02:44):
We can start wherever anorganization's at and help get
them to value.
Paul Bellows (02:50):
So AltaML is, named after the province of
Alberta, where we're sitting andspeaking today.
And we're here in the Canadiancontext and Canada has a long
history with AI.
A lot of the technology that ispopularized today and is
capturing a lot of people'simagination today was imagined
and pioneered here in Canada.
(03:10):
And you're really at the heartof what's going on in the AI
community here in Canada.
You're one of our experts interms of where you sit and what
you can see.
Maybe a little narrative ofCanada's role of, the early days
of AI and where some of thistechnology came from right here
in our backyard.
Nicole (03:25):
The interesting thing is Canada was investing in AI long
before it was sexy.
Back in the seventies, we wereheavily investing in this
technology and have been alwayssince then.
We have three of what are calledthe godfathers of AI that are in
Canada.
Located currently in Edmonton,Montreal and Toronto.
(03:48):
And they associate with thethree different AI institutes
that we have, Amii, Vector andMila, that are really at the
forefront of research.
I have to tell you the level ofresearch and the talent
generation we have here is it's,you can't compare it, because
we've been investing in it forso long in a very Canadian way,
(04:10):
humble, quietly, trucking along,just getting shit done.
And the interesting part is mostpeople, even myself, who I was
in tech, and 2018, living in acity that has one of the AI
institutes.
I had no idea that the AIinstitute existed.
(04:30):
I had no idea that we had one ofthe godfathers of AI at our
University.
I had no idea that people fromaround the world fight tooth and
nail to have a spot at theUniversity to learn about
machine learning and AI.
No idea.
I think most Canadians are inthat same camp.
We are at a place where we havethis absolutely amazing
(04:54):
foundation that today, honestly,if we tried to start today, we
could never get here.
Because of the dollars it wouldtake to get us here, because of
the hype now.
The talent that comes out ofeach of these universities that
focuses on AI is incredible.
Bar none, that is why we have anAI company here, because it is,
(05:18):
while it's one of the hottestjob titles to hire, we don't
have difficulty hiring, that jobtitle, because of the flow of
super incredible talent and therelationship that we have with
the different universitiesacross Canada.
We do a lot of internships, etcetera, and we have
relationships with some of thebest profs who will say, I know
(05:40):
AltaML offers an exceptionalexperience for an internship.
And so they'll send their beststudents to us.
And we end up having just thisincredible pipeline that we
actually also leverage for allof our clients because those
interns are working on ourclient projects.
And so often they become ourclient's talent just because
(06:01):
we've had such an effectiveapproach to the internship
program.
Paul Bellows (06:06):
I want to get into the next level of some of the
work that you do here at AltaML,but I think it'd be helpful.
Large language models and theparty trick at scale that has
appeared in our culture recentlyof, robots that will talk to you
and answer questions and writeessays, et cetera, it's
engaging, it's compelling, butit's not the sum of what AI is.
AI is a whole net of things.
(06:28):
Is it fair to ask, could yougive us a short primer on how do
you think about the categoriesof AI and what is possible out
there?
I think it's helpful to skippast the thing that has
everyone's imagination right nowand remind folks there's a lot
of different things that are AIwhen we talk about that.
It's category.
Nicole (06:45):
First I have to give a little background on me.
When we launched AltML in 2018,we saw an opportunity in this
space.
My co founder is quitetechnical, I am not.
So in, in 2018 I knew that AIstood for Artificial
Intelligence, and that was theextent of my knowledge.
And often, I think that peoplelike to hear me talk about it,
(07:08):
because I have to talk about it,not in a technical way, but how
I've come to understand itthrough more of a business lens.
I laugh, actually, because Ithink of it in three categories,
but one is called and referredto often as classic AI.
That's hilarious to me, because,the capabilities we have within
(07:31):
this classic version of AI areincredible and nowhere near
adopted to the level that theyshould be.
So don't feel like you're out ofdate if you're working on a
classic project.
And things that fall in there,computer vision, natural
language processing, those typesof things.
(07:52):
Then you've got, the newerversion that came out in 2021
with ChatGPT, the generative AI.
And so you're creating, someform of content, whether it's
software code, whether it'swritten text, videos, that's
new.
And so you're building somethingthat didn't exist before,
whereas the other versions areanalyzing the data that exists,
(08:16):
learning the rules of how tomake best decisions in those
areas.
But then, now we've got theagentic starting, and this is
the new thing around taking thatgenerative piece and adding that
next component to it whereyou're saying this is what we
(08:37):
should do and now I'm going tohave all of these agents go do
it on my behalf.
So those are how I bucket thethree.
As of 2021, everybody wanted totalk about just generative AI.
As of now, everybody only wantsto talk agentic AI.
And what I see is, let's findthe tool that solves the problem
(09:00):
that you're trying to solve.
Let's not ask the question, whatcan I do with AI?
But rather ask the question ofwhat business challenges am I
facing?
And is one of these types andversions of AI the right tool to
get me to where I need to be?
Because if you're just lookingto build cool AI, you're wasting
(09:21):
your time.
Paul Bellows (09:22):
Which is why you need a shop like AltaML to help
you traverse the broad landscapeand start to actually connect
technology to business problems,right?
Nicole (09:30):
Yes.
Paul Bellows (09:30):
Yeah, so so this gets interesting I want to talk
now a little bit about thebusiness model of how you work
here at AltaML because it's oneof my it's one of my favorite
things about what you do is howyou do it.
You're not a traditionalsoftware firm.
You talked a little bit aboutyou have your venture studio,
and then you have your servicesside where you do things for
people.
Can you go a little bit, justsome examples of on the Venture
(09:52):
Studio side, what do thosepartnerships look like?
How do you work alongside yourcustomer or your partner in that
context versus the relationshipon the services side?
Nicole (10:02):
So in the Venture Studio, we are truly co founders
with the founders.
Sometimes these ideas will comefrom within our team, and we
will put a CEO in place.
But in most cases, this is afounder who's come to us with an
idea, they don't have thetechnical background, so they
(10:22):
have an exceptionalunderstanding of the industry
that they're trying to solve theproblem in, they understand that
this is a problem we're solvingand they have the expertise to
run this, they just don't havethe technical piece.
So we bring in the technicalpiece and become that co founder
at the technical level.
We also support, backend thingslike HR and finance, and those
(10:47):
things that often distract a CEOat the onset and suck up a ton
of time.
We've got a strong team that cansupport in all of that.
So that's not where that founderhas to focus, and that, we can
really just hone in on whatwe're trying to build and get it
built and get it into market.
And then on the services side wework with our clients in a way.
(11:10):
We work in a couple ways.
So one is we'll work with aprivate sector organization
where we really help them honein on what are the most
impactful use cases in whatorder, and also keeping in mind
that I would venture to say that70 percent of an AI project is
(11:30):
the change management piece.
And the technical component isnot as large.
And we also help organizationssee that maybe IT is not where
an AI project should sit.
It needs to partner with IT, butwe actually have to truly
understand the business problem,who the end users are, who will
(11:52):
be impacted by this, so that itactually gets into operation.
So we work that way.
We also work, especially in thepublic sector, in a very
collaborative way, where we say,okay, we have a model with the
Government of Alberta, forexample, it's called GovLab and
we work within that.
They allow public organizationsto collaborate with them.
(12:15):
So we've had the City ofEdmonton, we've had the City of
Calgary all involved in this andthe IP that's built can be
shared.
Why would the City of Calgarypay for something, and then the
City of Edmonton pay forsomething?
This is all taxpayer dollars.
Why don't we, one of them buildit with us, and then the other
one can leverage it.
(12:37):
And so that's how GovLab hasworked, is it's allowed that IP
to be shared.
And so that we're not all havingto pay a bunch of extra money
with our tax dollars.
Instead, let's do this together.
Paul Bellows (12:50):
So the GovLab piece is what I was really
excited to get to in ourconversation here because I
think it's a really innovativemodel and it's something that I
think public sector folks acrossNorth America should be paying
attention to as an example ofhow to bring innovation into
government.
Now innovation always, it's aword that makes anyone who's
been in government long enoughhears the word and they cringe.
(13:11):
But, I like to, I'm a bit of anetymologist, so I like to go
back to the root of things.
And it really just comes fromthe root of renewal.
So we talk about innovation andsay, how else could we do this?
And I love that you alreadycalled out, AI is so much more
often business technology thantraditional IT technology.
It's really about do weunderstand the business problem.
Can you give me an example ofwithin, I want to talk a little
(13:34):
bit about how GovLab is working.
It's the next level ofunderstanding there.
But maybe the best way to getinto that is what kinds of
problems are you able to solveat GovLab the government would
not be able to solve if theywere trying to do this on their
own, work independently from anorganization with kind of
expertise for how to solve theseproblems.
Nicole (13:52):
So a really important aspect of the GovLab model is
that this is not traditionalprocurement.
We don't go project by project.
Instead, we look at AI as awhole across the government.
And we have, we support all thedepartments in coming up with
ideas and use cases, help themsuss out what's the feasibility,
(14:16):
what's the date, do we have thedata, what's the ROI on this,
all of the different pieces, andthen they pitch it to the
governance group, who decideswhich use cases are going to get
funded for the next quarter.
And so you're not going to getfunded forever and always until
it gets into operation, you'vebeen funded to see over the next
(14:38):
quarter.
Are you worthwhile continuingafter that quarter?
So that allows us to cut offprojects that aren't showing ROI
and get rid of them.
But it also allows us to hone inon the best projects and really
put focus on those.
Some of the projects that we'veworked on, I have two favourites
that I'll share with you.
One is sexy, one is absolutelynot sexy at all.
(15:01):
So I'll start with the sexy one.
It's around wildfires.
So we can predict now, with an80 percent accuracy, 24 hours in
advance, where a wildfire willbe in Alberta.
That allows the resources to beeffectively stationed where they
need to be ready to fight thatpotential fire.
(15:21):
That also allows a significantsavings in overtime of, gosh, I
don't know, it seems like, thewhole province might go up
tomorrow, I'm going to getpeople everywhere.
This really allows the dutyofficers to use their experience
combined with the model to say,okay, We're going to focus in
these three areas, or whereverit might be, and we're going to
(15:45):
keep our resources in thoseplaces.
Because every time you moveresources for firefighting,
you're at risk.
You're moving helicopters,you're moving all sorts of
things.
Those are not things that youwant to be moving a lot.
You want to make sure they're inthe right places and ready to
fight the fires that come.
Now the interesting thing aboutthat model is we've just started
the next version of it where wecan, we're looking at the fuel
(16:08):
grid.
How many dead trees, how manydiseased trees are in these
areas and can we be doingpreventative maintenance?
So that if we get a wildfirestarted in that area, it won't
be quite as dramatic anddamaging as it would have been.
We're combining the two to bethis, information system for
duty officers to be making thebest decisions.
(16:30):
That's my sexy one.
My not so sexy one is aroundsafety codes.
So we have a lot of safety codesin the province.
Not the best reading unlessyou're wanting to fall asleep.
But there's a lot of questionsthat come in around safety codes
from builders, developers, etc.
And they're often waiting onthose questions to move forward.
(16:54):
And so the wait time was oftenweeks, sometimes up to a month,
waiting to find out the responseto these questions because it
was a lot of information to gothrough.
And they had to have highlyskilled engineers.
providing these responses.
We've created a model, agenerative AI model, that
(17:16):
produces that response, that canonly pull data from the existing
safety code, so you're notgetting the, the issues of
ChatGPT, where you have no ideaof what it's saying is true.
Instead, we're saying, you canonly provide examples from this
content, and it gives the sourceof this, which safety code it's
(17:36):
in, what clause it is, etcetera,drafts the email.
We've got a human in the loopbecause that's an important part
of responsible AI to take alook.
Is this the right response?
And so now response time is twominutes instead of many weeks.
And those individuals who arehighly trained engineers, are
now actually getting to dosomething in their work that's a
(17:59):
little bit higher value and whatthey were actually trained to
do.
And so that has been a big winfor that department, and we see
that opportunity there to takethat and expand it across really
anywhere that has legislation orregulation that needs a
response.
Because if you think about howmany things and categories we
(18:21):
have of that in government, howmany places could this be an
impactful model?
So not so sexy, but super, superimpactful.
Paul Bellows (18:30):
Useful though.
Nicole (18:32):
Yes, very useful.
Paul Bellows (18:33):
So useful, I mean it's just, if you think of that
as a paradigm of government hasdata, the data is complex, it
needs interpretation, or search,or discovery, and you've got
very high paid people, whichmeans you can't have that many
of them, because you just can'thave everyone sitting benched,
and when things get busy, itslows down business, and having
done some commercial real estatedevelopment in the past, just
(18:54):
through offices opening things,every day you're waiting for an
answer, costs go up.
Nicole (18:59):
Yes.
Paul Bellows (18:59):
Openings are delayed, I love this, that's a
brilliant example.
Anyone who's ever had to getsomething done alongside
government would appreciategovernment having access to
those tools.
Nicole (19:09):
Yes.
Paul Bellows (19:10):
So I want to just a couple of questions one thing
you said that I love it I justwant to put some punctuation
behind is I love you talkedabout procurement and how this
is a different model ofprocurement.
You talked about the abilitywhen something's not working we
can stop, which is remarkable ingovernment procurement.
Usually, you do a public RFP,you buy the thing that you need,
(19:32):
and then you've bought it, andyou have to see it through to
the end, right?
It's hard to stop once you'vebought something and struck a
contract.
This is a structure in which badideas can fail fast.
And without public visibilityand people getting embarrassed,
it's just, we were all sure thiswas a good idea.
We were wrong.
We can be wrong without spendinga lot of time and money or
(19:53):
having to build something thatwe know people will never use.
That's just one of my favoriteanecdotes you've pulled out here
so far.
Tell me a little bit more, theGovernment of Alberta has said
we're going to put somepermanent, time based funding in
place, to create a certaincapacity for AI work, on a
quarterly basis.
You say, how are we going tospend that capacity?
What problems are we going tosolve?
And you have to continuallyjustify the value that's getting
(20:15):
produced, am I understandingthat correctly?
Nicole (20:16):
Yes, definitely.
Paul Bellows (20:18):
And have you seen this happening anywhere else in
Canada, the U.
S.?
Is this a novel thing happeninghere in Alberta, or is this
something more uptake withingovernments?
Nicole (20:26):
I haven't seen it elsewhere, but I can tell you
that every government that Italk to is excited and
interested about it.
We're talking with a bunch ofprovinces and we have talked to
some states in the US as well,where there's, they see how
difficult it is to do AI ifyou're not either buying it off
(20:51):
the shelf, which often is notgoing to get you what you need.
Sometimes it is, and then youshould just do that.
But a lot of times it's not andthen they look at their existing
procurement.
And they think this isimpossible.
How, because you can't, youdon't know what the end state is
truly going to look like with AIuntil you get started.
(21:12):
And you actually see is the datagoing to give us the accuracy we
need?
And then if it isn't, is thereanother way to solve this
problem with a differentapproach?
You have to have that agility.
And so because that hasn'texisted really before in
traditional procurement, It'sjust so incredibly challenging.
And so I would say that, I wouldsay a year from now, we will
(21:36):
have, you'll see a whole bunchmore governments starting to
deploy this model.
Hopefully with us.
Paul Bellows (21:42):
They're all going to hear this podcast.
Yeah, absolutely.
So next question then.
So it's stable funding, thegovernment can control the
spend, see this is the amountwe're willing to commit.
We get as much work done withinthat as we can.
We produce models, we produceintellectual property.
Who owns that, at the end of theday?
Nicole (21:59):
This is another interesting part of the model,
is that the government owns theIP.
But the Government of Alberta,for them, it's very important
that they're a leader in thecommercialization of AI because
Canada really has struggled inthat area.
And so they have allowed us tohave the ability to license that
(22:20):
IP and build products on top ofit and commercialize those as
long as they get a share in theprofits.
Their goal is over time, andwhen you're building products,
that doesn't happen overnight.
So we don't have any fully inmarket yet.
But once we build thoseproducts, and we have a few that
we think are just about there,then their AI program will start
(22:43):
generating revenue.
And their hope is, over time, itbecomes self sustaining.
And in fact, they're not puttingany money into AI, other than
the money they're getting backfrom it.
And so that is a reallyinteresting approach to AI that
I think is, you can't expectthose returns immediately, but
with the long term plan, thatcan be really advantageous.
(23:06):
Then the other piece that wasreally important that I haven't
talked about is the talentdevelopment, because with with
the public sector, let's just behonest, it's not the first stop
for your newest grads coming outof tech with the hottest, job
title.
But a startup or a scale up likeus is and so that's why we have
(23:28):
no problems attracting teammembers for our internship.
Then they work on these projectsin government and realize, wow,
there is some incredibleproblems that government can
solve in AI because the datathat is here is incredibly
valuable and can really solvevery complex very interesting
(23:51):
and impactful problems.
So, every person that theGovernment of Alberta has hired
in this area has now comethrough GovLab as an intern.
And so this has become theirtalent pipeline.
And then whoever doesn't becomea Government of Alberta or an
AltaML staff person goes intothe ecosystem.
(24:12):
And is someone who doesn't justhave the academic background,
but they have real worldexperience developing these
solutions and getting them intooperation.
Which is valuable to our, to ourmarket.
Paul Bellows (24:25):
One of the things that I think, as a practitioner
in the digital software spaceworking alongside government,
one of the things that I'vealways found is the most
fascinating is just the sheerscale of what government does
and does every single day.
The scale of the data, the scaleof the operations, the scale of
the staff, the organization, thebudgets.
Government is a large, behemoth,which I think often makes it a
(24:48):
target for why does it cost somuch, that'll be a separate
conversation that I don't liketo weigh in on, I'm not
qualified.
But, to say, hey, working withgovernment, you get access to
data, but also to do things thatimpact real people every day,
like it's useful what you get tocreate when you do good things
with government.
I think that's really excitingas a talent pipeline.
And again, some of the talentthe government wants to hire,
(25:11):
you need to build a relationshipbecause what the market's
offering for salaries andcompensation, in start ups and,
venture backed organizations,versus what government can offer
in compensation, they're justdifferent categories.
But sometimes you fall in lovewith the work, and maybe a lower
comp, but more stability, morepatience of an organization,
longer timescales, and theability to have impact, I think,
(25:32):
can be really powerful for youngpeople.
who look around and say, I wouldlike to contribute to the
betterment of the world ratherthan just pad my own bank
account.
What, how can I work?
I think that's really exciting.
Nicole (25:42):
We're doing a project with Alberta Cancer Care right
now that's a part of the GovLab.
And we're working on how do weget individuals who've been
diagnosed with cancer quickeraccess to oncologists?
How do we remove the wait timeand start reducing it?
Now, obviously, eventually,we'll roll that work out,
hopefully, across the healthcaresystem, but we're starting with
(26:05):
cancer.
Now, imagine coming into aninternship and getting to work
alongside that, thinking, mywork is impacting the wait time
of somebody with canceraccessing the system.
That's incredible.
Paul Bellows (26:21):
What
Nicole (26:21):
a thing worth getting out of bed for in the morning, I
love it.
That's brilliant.
Something that maybe peopledon't understand clearly is what
is the skill set you're hiringfor?
You're talking about internsfor, and again, for folks who
don't really get into the weedsof what AI is and the types of
technology, what kinds of skillsets are you able to bring to
government?
Paul Bellows (26:41):
Are these design people, coders, math people, all
of the above?
Can you talk a little bit aboutthe kind of talent that would
sit within GovLab?
Nicole (26:50):
We would have software engineers we would have machine
learning engineers but we alsohave both internships and a lot
of roles in the business side.
Some of the most valuable peopleare those who can listen to the
business side and translate tothe technical side.
I refer to them as ourtranslators, but they're,
(27:12):
project delivery, they'reproduct owners, they're those
individuals who might not be theperson doing the code.
But they can help dive down intoexactly the right problem to
start solving that's going tobring value.
And then communicate that to thetechnical folks so they can dive
in right away.
Because if you just handtechnical folks a bunch of data
(27:34):
and say, What cool stuff can youbuild with this?
I'm going to be honest, a lot oftimes there's not a lot of value
there.
There's a lot of cool things!But it might not be the most
valuable thing to theorganization.
And so those individuals arereally important too.
The skill set that thoseindividuals need to have is an
awareness of the capabilities ofthe technology and then being
(27:58):
able to really dive in and digdeep into understanding a
business problem.
Paul Bellows (28:04):
I think one of the most important questions that
many technologies fail to answeris, what should we do?
There's a problem.
Do we understand it?
Have we analyzed it?
Have we gotten to the root ofthat problem?
And then, yeah, we have a lot oftechnology, but those are just
tools.
They only ever tell us how, notwhat and not why.
Yeah, the people who canconnect, a problem space to a
(28:25):
technology space and create abridge, that's powerful.
I'd like to think that's whatI'd use.
I'm glad to hear that's usefulin the world.
I want to talk just for one moreminute, coming back to where we
started.
So historically, and there's somany parallels with everything
Canada has ever done, which iswe're really good at resources.
We have all these wonderfulresources and we've always
(28:46):
relied on other people tocommercialize those resources to
some degree, that it's the, ourgreat critique is, and then we
ship it off and someone elsebuilds the house out of it.
Someone else upgrades our energyand does things, and it sounds
like there's a bit of thathappening with AI.
We've pioneered a lot of this.
We've invented a lot of it.
We've got these deep wells ofexpertise here.
We've gotten used to havingthese people here.
We have these ecosystems thatproduce more of them.
(29:08):
So we have all of this, but wehaven't always harnessed it to
produce value, wealth,solutions, whatever you wish to
have at the end of the day.
So as you sit where you sit,working alongside government, In
a moment where I think Canada'sbeing challenged to level up in
2025 here and get to the nextlevel.
What do you see as some of theproblems that we need to solve
(29:29):
in the business ecosystem, inthe technology ecosystem, in our
public service ecosystem?
What do you think we need toconfront as governments,
provincially, municipally,federally, right across Canada
right now?
How do you, see that?
Nicole (29:42):
So I think that the two biggest challenges are both
commercialization and adoption.
On the commercialization side itbecomes tough when your
investors and your customersaren't here.
It's very easy to raise capitalin, in AI in the US.
It's not so easy in Canada.
(30:02):
It's also, to give perspective,our sales cycle.
In Canada is 18 months, in theU.
S.
it's four.
I am a proud Canadian and wewill always be, a Canadian
company.
But, we also, do work in theU.S.
because that's a big differenceto a business, right?
(30:24):
You can't ignore those facts.
And how do we as as a countrydecide who we want to be and
what we want to do, really,frankly, for everything, but
specifically in AI, what are thelayers that we want to be
exceptional at?
Because there's a massive supplychain to AI.
There's the infrastructure,there's the foundational models,
(30:48):
there's the application layer,there's all of these different
pieces.
Are we trying to be excellent atall of them or some of them?
What's our brand going to be?
And I, I think the federalgovernment is really tired of
hearing from me continuallysaying our brand needs to be
that we're the global leader ofresponsible AI.
And if you buy AI in Canada, youknow it's ethical and built with
(31:10):
responsible AI principles.
That's a brand that aligns verywell with who we are.
But, that's a side.
I have to tell everyone I evertalk to about that because
someone will start listening.
But, so that's the kind ofcommercialization side.
But then the adoption side alsorelates to it because it's about
(31:31):
the customer side.
A piece of commercialization.
With that 18 month sales cycleversus four months, with AI, the
impacts it starts having on anorganization are very quick.
Once it's in operation, itbegins to make an impact
immediately.
But the life cycle of getting tooperation, isn't short with AI,
(31:56):
it is an investment of timebecause you can build this
really cool thing that solvesthis really impactful problem,
but there's change that has tohappen in order to use this
tool.
So with the wildfire tool as anexample, that is a seasonal
tool.
We do not need that tool in thewinter.
So you can only really build it,and perfect it, and test it in a
(32:20):
certain part of the year.
And that first day it's ready,do you think the duty officer's
sure, let the AI model decidewhere everything's going to be."
I can tell you not a chance.
And so they needed an entirefire season to work alongside
the model where they made allthe decisions based on their
experience and then looked atwhat the model said.
(32:44):
And yes, that either validatedor it didn't, and then they
waited to see what did happen 24hours from then.
Was the model right?
Or was it wrong?
Would I have made the wrongdecision had I, followed the
model?
Or did I make the wrong decisionand should have followed the
model?
And so after the season, thenthey were ready for the next
season to bring it into place.
When you're talking about ashort season that only comes up
(33:06):
once in a year, Think of howlong that cycle was to get that
to a place where it was addingvalue.
And so we have to start, we haveto get started building these
things.
We can't stop, which a lot oforganizations do that we see in
Canada, of let's educate ourboard, let's educate our
executive teams, and then we'llthink about it.
(33:30):
And we'll spend the next 18months getting our data in
order.
That's what everybody says.
Which I laugh because You, howcan you get your data in order,
unless you know how you're goingto use it?
You can start AI projects with asample of data to see if you're
going to get an accuracy that'sworthwhile.
And then you'll realize, oh I,you know what we need?
(33:51):
Is this one other piece of data.
And if we had that collected,that would make this model
incredibly powerful.
You've never been collectingthat.
We need to start collectingthat.
And they want to make all thesedata decisions and make that,
have everything perfect, by theway, data is never perfect and
then start with AI.
(34:12):
You can't do that.
That does not work.
And I, what I'm pushing, Canadaneeds to start adopting.
Stop waiting.
We've always been hesitant toadopt new technology.
The impact and the risk of notadopting AI is so massive.
In order to compete and stayproductive and efficient and,
(34:34):
frankly, to make all industriesand government across Canada
able to compete in this crazyworld, we are going to have to
get going.
Paul Bellows (34:46):
Some of the pioneers of digital government
were some of the folks in the UKand they wrote a book out of it
called Transformation at Scaleabout just how they did it.
And the core mantra out of thatis, you start by starting.
So last thought, Nicole, thishas just been a wonderful tour
of what's happening here inAlberta with the GovLab and what
you're doing at AltaML, a bit ofthe state of AI.
(35:07):
So start by starting, I'm in aprovincial government
leadership, I'm a CIO, I'm abusiness leader, I'm the federal
government, I'm in a large citythat needs to tackle AI, what's
an action that can be taken?
Nicole (35:18):
Call me.
Paul Bellows (35:19):
Clearly.
Nicole (35:21):
But I think it is just starting to take action towards
doing.
Ask yourself the question of howare we going to get something
embedded, built, whatever itmight be, so that it's in the
hands of users.
Don't focus just on theeducation component.
Think about the end state andstart working towards that.
(35:43):
Rather than just, first we'lleducate everyone.
Yes, let's educate everyone allalong the way.
But we gotta get going ongetting something in the hands
of users.
Paul Bellows (35:53):
It's time for Canada to move fast.
Nicole (35:55):
Yes.
Paul Bellows (35:56):
Thanks, Nicole.
I appreciate the time.
Nicole (35:57):
Thank you.
Paul Bellows (36:02):
Thanks so much for joining us for this
conversation.
Nicole is both a nationally andinternationally recognized
innovator and entrepreneur, butshe's also a humble expert,
clearly.
That's a compelling combination.
Some of the key themes shebrought forward that I
appreciated include Canada hasbeen investing in AI innovation
for decades, and now that therest of the world is catching
(36:25):
up, we still have a competitiveposition from which to thrive.
When you build the rightstructures for bringing problems
together with experts andtalents, you can achieve great
things.
There isn't time to waste.
Government needs to move quicklyto understand this space.
And as always, we start bystarting.
(36:46):
And finally, AI requires talent,patience, and new business
models.
Alberta's GovLab is one greatmodel, and if you want to learn
more about it, Nicole would loveto take your call.
I hope you enjoyed thisconversation with Nicole Jansen.
Please do subscribe and followthe many conversations we're
going to be releasing throughoutthe year.
(37:07):
I'd like to thank my colleagueswho work with me on this
podcast.
Kathy Watton is our showproducer and editor.
Frederick Brummer and AhmedKhalil created our theme music
and intro.
We're going to keep havingconversations like this.
Thanks for tuning in.
If you've got ideas for guestswe should speak to, Send an
email to the311 at northern.
(37:28):
co.
Remember, the public service isabout all of us, and when it's
done right, digital can be a keyingredient for a better world.
This has been the 311 podcast,and I'm your host, Paul Bellows.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.