How will vehicle experiences evolve with AI? | with Mel Roberts of Stellantis

Recorded live at AutoTech 2026, host John Heinlein speaks with Stellantis's Mel Roberts on how AI and small language models (SLMs) are transforming the automotive user experience. They discuss the strategic deployment of AI in vehicles, focusing on local processing for low latency, predictive maintenance, and balancing physical controls with digital UX design.

Creators and Guests

Host

John Heinlein, Ph.D.

An experienced technology and marketing leader, John brings his background from startups and established companies to Sonatus. He worked for 14 years at Arm, most recently leading Automotive Partnerships for North America where he engaged OEMs, Tier-1s, and others to deploy Arm-based solutions into automotive applications, including autonomous vehicles. His team was integral to launching the SOAFEE industry initiative for software-defined vehicles of which Sonatus is a member. Earlier he served as VP and Chief of Staff to the CEO and led a group responsible for competitive strategy. For three years, he was VP of Corporate Marketing where he led centralized outbound marketing, spanning marketing campaigns, press, events, web, digital marketing, ecosystem programs, and working closely with investor relations. Prior to Arm, John had an 11-year tenure at microprocessor startup Transmeta where he held several senior roles spanning business development, marketing, and customer success, among others. John earned his B.S. in Computer Engineering from Carnegie-Mellon University and an M.S. and Ph.D., both in Electrical Engineering, from Stanford University.

What is The Garage by Sonatus?

The Garage Podcast from Sonatus brings you conversations with thought leaders from around the vehicle technology space discussing far-reaching topics about software innovation in vehicles.

These episodes will include industry experts, whether from Sonatus, or from our amazing partners across the industry, all of whom will share their Ideas and their outlook for the most important topics in vehicle technology. Episodes span from more technical topics, to business evolution, and market trends. Join us and learn about the future of vehicle in The Garage.

Today in The Garage,

we're recording live at Auto
Tech 2026 in Novi, Michigan.

Just last night, we were
participating in the Auto

Tech Awards, including
Wards Ten Best Interiors.

And one of the things that the
interiors awards celebrated was

innovation in vehicle
experience and user

experience for the
driver and passenger.

And you're seeing an evolution
of how features get presented

to drivers, both in
terms of the displays,

voice interaction,

which controls are physical
controls for things like volume

or climate, which are soft
controls on screens that are

tactile and change over time.

And just the use and way
these things are evolving

is changing all the time
and user appetites are

changing all the time.

It's a really exciting
application of software.

And as software-defined
vehicles evolve and as we bring

more AI into vehicles,

there are exciting new ways
that we can use software and AI

to improve the user experience.

To talk about that and learn
about that important topic,

we brought on an expert,

which is Mel Roberts—our guest
today is Mel Roberts—she's

director of North America
Infotainment and Feature

Planning for Stellantis.

And in our conversation,

we talk about the user
experience in many ways,

how the user
experience is evolving,

and how software and AI can
enhance that experience now and

in the years ahead.

It's an exciting conversation,

and I think you're going to really
enjoy Mel's unique perspective.

Let's go!

Welcome to The Garage.

I'm John Heinlein, Chief
Marketing Officer with Sonatus.

We're here recording
live at Auto Tech 2026.

Mel Roberts, welcome
to the podcast.

Thank you.

You know, we've been talking for
years about having you on the podcast.

We're I'm glad our finally,

we're able to get
you on the show.

Very excited to be here. Thanks.

Start by telling us a little bit
about you and your background.

Absolutely.

So I graduated
from Michigan Tech,

and I started my career
in the medical industry.

So I was working at a hospital
as a bioelectrical engineer,

fixing medical equipment.

As you do.

I mean, it was fun.

Every day was a different
a different story.

And then after that, I
went into medical research,

found myself accidentally
doing microsurgery on

rodents for a
microcirculation model,

realized I wasn't that
good at it after a while.

And so I went back and talked
to one of my mentors from my

undergraduate career, he said,

what did you really
when you were eighteen,

what did you wanna
do when you grow up?

And I said, I wanted to
do crash test dummies.

He's like, well, you're
not even in automotive.

I'm like, that's a
valid point. Yes.

So he introduced me to someone.

And long story short, I ended up

automotive and will
never look back.

I loved it, And I don't
miss the mice nor the rats.

Although sometimes I still
have stress dreams about it,

but happy to be here.

It's just such an incredible
background that people's

journey I mean, when
I began my journey,

I've been in technology
of many different kinds.

I didn't know I'd end
up working in automotive

technology, but it's a
super exciting space.

Absolutely.

We always like to get
to know our guests.

You have to tell us
a fun fact about you.

Okay.

Well, also, at that illustrious
college in Houghton, Michigan,

Michigan Tech University,
I was a college radio DJ.

Cool, was there a style
you specialized in?

Yes.

And that is the tricky
part, three hour show,

all on punk rock.

So it's like, what, a minute
and a half, two minute songs.

I hustled.

I gotta work out every
CD and vinyl because

that's what we had.

We didn't have This
was a digital music.

There was no digital then.

No. No. No. No. No.

So the whole, you know, DJ
station looked like a mess,

a tornado after I was done.

Everything was
ripped off the walls.

But I had a huge
following because we had the biggest

transmitter in Michigan.

So I had a huge following at
the local Baraga County prison

and got lots of fan mail.

And Proposals from prison?

Absolutely. No end in sight.

That is hilarious.
That is hilarious.

So I always share
a fun fact back.

My fun fact is I'm a podcaster.

That's awesome.

No.

But all joking aside, when
we started the podcast,

we it was an experiment because
we really wanted to engage,

you know, thought leaders.

And it's been a
great experience.

I really enjoyed it.

And I was not a
radio DJ in college,

but I'm jealous that
you got to do that.

And punk rock too,
that's super fun.

Yes.

So tell us now about Stellantis.

For those of our listeners who
may not know Stellantis very

well, it's a famous company.

You know all the brands that Stellantis
makes and also about your role there.

Yeah. Don't quiz me on
all fourteen. No. I'm kidding.

Fourteen brands.

Fourteen brands now with
the Stellantis merger.

And, it's pretty incredible just
watching and learning about all

of the new brands that have
come in the fold since since we

have the merger
between FCA and PSA.

And

it's been really cool
to just be a what

our CEO would call us,

a a strong global company
with strong regional roots.

So FCA, so Fiat Chrysler
Alliance, a lot of the,

you know you know,
Dodge, Chrysler, Ram,

all those Chrysler brands,
many European Fiat Fiat.

There's many others in
there. You name a few.

We're not gonna get fourteen.

Peugeot, Lancia.

Right.

Of course, I loved my Fiat
five hundred Cabriolet.

Oh my gosh. That's a cool
car. Give me that car back.

So so, anyway, many brands
under the Sonatus umbrella.

And then what's your
role for Sonatus?

Well, I am focused on
the North America brands.

So my Jeep, my Dodge,
my Ram, my Chrysler,

those are my four babies.

That is where I spend my time.

And and your your scope
of responsibility is on

infotainment and
feature planning?

Yeah.

So any feature, anything that
the customer would interact with.

Certainly, my time from my
first job of cup holders,

cup holders are still one of the most
important features to our customers.

If I bring it up,

it will ignite a
passionate response in

anybody I talk to.

Trash management to, yeah,

all the new digital
features that we can do,

things that are coming
up on the infotainment.

Safety and peace of mind is
always very important and

critical to our customer base.

On the podcast, we
we focus on software,

and especially these days,

we're spending most of
our time talking about AI.

So let's start by beginning
with exploring how OEMs are

typically using AI
in vehicles today,

and then we'll go on a journey
about what we can do in the future.

So what's your experience where
what are typical ways AI is

used in vehicles so far?

I think a lot of people are
talking about that surface

level interaction with the
customer that we're kinda

seeing on the phone.

So the enhanced conversational
personal assistant.

But I see much more
power than just that.

And so while that's cool and
that's interesting, it's fun,

and it's gimmicky, could
I sit there and tell my my

car, hey, instead of
this dashboard graphic,

I would like my dashboard
to look like I'm in

Gotham City at dusk. Right?

And then the whole mood
of the cabin will change.

That's cool.

Yeah. Right? But AI
is very powerful.

Yeah.

And we have to start talking
about the return on investment

and what is the value
for the customer.

Does the customer need that
Gotham City dashboard? Maybe.

Maybe not.

So I'm really exploring
what are those value propositions

and where should we
strategically be using AI.

I think when many
people think about AI,

I think there's a a feeling
that maybe people really mean

LLMs or large language models.

But the reality is that's
just one small part of it.

Everyone knows
ChatGPT or Anthropic,

very topical these days.

But there's actually much more to
it than just those simple models.

And some people talk about
SLMs or small language models.

Can you tell us what
is this what is an SLM,

and and what are other ways
we should be thinking that AI

could be part of the story?

Well, it's interesting
you asked that.

And because of these
conversations I have with the

amazing supply base and the
technical experts that I meet

at conferences such as this,

they've been educating me
about the importance of an

AI orchestrator and making
strategic decisions and

highly trained small
language mod models.

And the importance for that,
as it's been explained to me,

is then you can decide.

We all know that large
language models are very data intensive,

and that's costly.

So we also know they
have a lot of artifact.

So if I, as a
customer, ask my car,

what is that icon
that's on the dashboard,

or what is the sound
that I'm hearing?

I don't want that command
to go out to an LLM that

reads a Reddit thread of
some conspiracy theory.

I want it to go to the
trusted information,

the grounded AI.

I would rather it retrieve that
information locally because it

should be small enough
to fit within our our

edge computing environment.

I want it to get that
answer with low latency,

and I want it to be accurate,

and I want it to be exactly the
correct answer when it has to

do with vehicle functions
and vehicle safety.

If I'm asking about, you know,
Taylor Swift's wedding day, okay.

You can go to the LLM for that.

You can go to Reddit
for that. I don't care.

So two or three interesting
things you said there.

One is that SLMs,
because of their small,

that's the s in SLM,

they don't have to have the
same kind of massive compute

intensive capabilities.

You know, we're talking about
building data centers near rivers so

they have enough hydroelectric
power to power the data center.

I mean, that's not a car.

So it's something that
could be in a car first.

Second, and I'm so glad
you mentioned that,

the trusted data sources, because
SLMs can be more tailored.

They can be more customized
and built on authoritative

proven sources that are known
and trusted and viewed and

reviewed by the OEM.

So I think that's
really, really great.

And finally, they can
be in the vehicle.

So latency, whether it's
response time or responsiveness

just in general, can
be much more powerful.

It's really great.

Yeah.

And, you know, as I eavesdrop on
conversations around here or at CES earlier

this year, you know, you hear
a lot about, oh, AI is scary.

AI is error prone. AI
is gonna take our jobs.

It's going to you know?

And it's all of these big
things that AI is a tool.

Every year, there's a new tool that
we're both equally concerned about

or, you know, excited about
and and both at the same time.

So for me, it's all about choosing
the right way to use this tool.

You don't come in with a
sledgehammer when you can come

in with a nail file.

I love how you said that in in what we
people think about AI as monolithic,

but the reality is the
artifacts come when you're

trying to solve very
general purpose knowledge.

When you look at
application specific AI,

the answers are very
reliable, very trusted.

And they're also they they
don't need to be complicated

over overly complicated, as
you say, the sledgehammer.

That's a great
great example. Yeah.

Another thing when
people think about LLMs,

I think they immediately connote
LLMs with voice and voice commands.

And I'm sure as part
of your responsibility,

you look a lot at voice
interactions in vehicles and

what cases people will use
voice commands for and whatnot.

What's your take on voice
and and how voice interactions in

the vehicle have evolved?

The evolution of voice
is really interesting.

Some people are doing
it better than others.

And I also see it it's almost
like a horse race where you

think someone's out in front
and then someone takes over.

And I think that in
vehicle going forward,

a multimodal experience is
gonna be great because what I

notice sometimes when I'm doing
voice interactions, you know,

in vehicle or out of vehicle,

predominantly actually
outside of vehicle,

is the regurgitation
back can be lengthy and

clunky, and I don't
have time for that.

Right?

So I would like to see because
we have these beautiful screens

in the car, because we've
embedded lighting and other,

you know, audio
cues, visual cues,

let's have a great multimodal
experience where if I

ask for the weather,

maybe it displays on the screen
instead of reading me each day

where I've forgotten what
happened on Monday by the time

they get to Thursday.

If I am asking the
vehicle to turn up

the I'm I'm cold.

Right.

You know, make me warmer.

Maybe it's a subtle ambient
lighting switch that just gives

me the red that says,
yes, I heard you,

and I'm making it warmer.

And blue would be
make it colder.

So combining those
multimodal cues, I think,

is really interesting.

It lets the user know
that the vehicle is

responding to you but
in a less obtrusive way.

And what I noticed even just
coming down here today with all

the detours to get to Novi
from where I'm from is

my GPS is competing
with my SiriusXM,

and I was not pleased
because it's talking over each other.

Right.

But there were things
I needed to know,

and I was glad that she told
me that there was a cop ahead,

or an emergency
vehicle or a detour.

I also missed the punch line on
on the show I was listening to.

Shout out to Howard Stern.

Don't cut that.

No.

And and I think that it's so
interesting because voice,

of course, has
become more popular,

but there's also
moderation to everything.

If you do everything with
voice, it can be overwhelming.

And and I think everyone's
personal preferences are different.

Some people love to be talked
to and they would they don't

wanna be distracted with UI,

but I think it's all too
easy for everything to to be

singularly voice
and be overwhelming.

You know, we were here last
night at the Auto Tech Awards,

and they always do Wards
Ten Best Interiors.

And a number of
vehicles won won awards,

including a a Jeep was one
of the ones that won an

award last night.

But it talked about
it was very funny.

The judges said the focus
and the shift these days is

actually back towards a
balance between a changing

sort of screens that had
that can reconfigure and hard

controls for audio or
climate or whatever it is.

And they were really looking,

and they talked about
this repeatedly,

a balance between kind of
application specific single

function controls and
screens that can reconfigure.

It's so important,

and we have to keep an eye on
that and what the customers

need and how they interact.

And if you remember,

the one thing that they shouted
out when Jeep got their award

was rear audio controls,
which is a huge,

huge discussion always.

And people that haven't
had it, they're like,

but once you have it,

you do not wanna let
go of those rear audio controls.

And I think it's a little bit
about a power struggle in my

own car because my fiancé
will try to turn down my

punk rock subtly, and I'm
turning it up on up on the

manual button.

He's turning it down like when
he thinks I'm not looking.

So but for the fact that Big
Brother's turning down my punk

rock, the the volume controls,

physical volume controls being
dedicated...v ery important.

And there are certain buttons
like that that we need to

remain hard buttons.

Right. Right. And you
mentioned a couple of those.

I can't remember how many times
they mentioned how much they

were happy with the placement
of the cup holders last night.

Yes.

Because that's not the
subject of the podcast,

but how crazy is it that that's a
really big focus, but it matters.

It does matter.

And I was kind of ruminating on
the fact that I think the next

big evolution in cup holders
is actually going to be cup

holders and the other perennial
unmet need of the cabin.

Phone holder....OK,
maybe I'll give you that.

I think I think there's
enough solutions to that,

but I am thinking...
purse storage.

Purse storage might not
be your top concern,

but it's many
people's top concern.

It's a perennial unmet need.

And the technology that I believe
that's going to solve this Yeah.

Is steer by wire,
believe it or not.

I think getting all of
those mechanical monuments

out of my cabin
and getting Yeah.

More leg room and more space.

That space back to prioritize
the space for what I needed,

which is more cup holders.

I need all the
drinks and my purse.

Well, there you go.

There's a first there's a first
for The Garage, purse holders.

You mentioned a number of
topics around personalization.

And personalization has really
been changing a lot in these

days with some vehicles
being hyper personalized,

and we're seeing some interest in
that in the survey we recently did.

I'm sure we'll talk about
that a little later.

What's your take on
personalization and and where

we are as an industry?

I have seen that one
ebb and flow as well.

When we were taking a
look at and benchmarking

infotainment maybe three
to five years back,

it wanted to be the
industry was endlessly,

I'm gonna configure this.

I'm gonna swipe this over from
this screen to that screen,

and I want my widgets to be
exactly what I want them and

and the same shape and
the size and customize.

And I think what we found
is those are underutilized

features and sometimes
considered digital clutter.

They don't necessarily want
big customers to have to go

and reprogram their
whole a superuser might.

But not everyday customers
wanna sit there and hyper

personalize their entire screen.

They want us as the
experts in UX and UI to

say, this is the right
information to display in the

right place in the right time.

Right.

And that to me is the key.

And when we can do that and
utilize these new tools with

respect to AI and customization

and connected data, real time
alerts going to the right

location, And also
just I'm a huge fan of

the HUD, the heads up display
screen and getting information

prioritized over there.

So I think that will be a better
level of personalization for us.

And then the the second aspect
I would think that is an

emerging and unmet need
that will start to grow is

climate because I think
we have the tools now.

It's just a little
bit expensive. Right?

We have the tools. We
can scan your body heat.

We could say where
should we direct the air.

Oh, automatic Yeah.

You know, we know how to do
it. We can motorize a vent.

We can scan your body,
but it's expensive.

So I'm curious to see
what AI can do to help us

redeploy those oscillation
patterns and make it more like

a custom Dyson experience
within your cabin.

It's so unique.

I never heard that
anyone suggest the idea because,

of course, you have
you can set, you know,

automatic temperature,
it changes the fan.

But the reality is different passengers,
men, women, tall, short, you know,

may have different kind of body
experience for when they're so

really adjusting it even more
than we think of as auto today.

Yeah.

Is that's very powerful.

Well well, that and sitting
just sitting right here. Right?

Like, got the lights on me.

It's like I'm a little I got
some hot spots. Like, you know?

Like, I would like some
subtle breeze gently

direct it...mental

note we will add a subtle
breeze to the future.

The hair blowing back.

Yeah.

You know, and the other thing
you mentioned was about thinking

about the UX design.

And and we work with
a number of different user experience,

user interface designers in our
company and other companies,

and it's an
important profession.

Those experts really think
through about how to present

information in a smart way.

And I know for my vehicle,

I'm I love how it was clearly
very thoughtfully designed,

and it's not very personalized.

Well, there's some there's some
careful personalization about

some information I can
choose to show or not.

But fundamentally, it's
just thoughtfully designed.

I get in rental cars sometimes,

and you have the ability to
really move screens around in

kind of, I think, in
me, frivolous ways.

And a lot of the screens
are hard to read.

And as a rental car,

it's kind of almost unsafe
because you wanna make sure

that the key
information is viewable.

So I really celebrate the
thoughtful UX design coupled

with tasteful, sensible
customization for maybe

people's personal for
example, entertainment or not,

for example, or maps
or not, for example.

Yeah. I think that's
the right way to go.

And I also think that with
increasing levels of autonomy,

that becomes even
more important.

So I love to deep dive these
things with the thoughtful

experts at our company
because like you said,

it it it's very carefully
designed, every aspect,

and they have words and
terminology and and rationale

as to every little piece of it.

But for me as the end user,

what I really need to know is
if my car is in one of these

higher levels of autonomy.

It's not that I see brake
lights ahead of me or I saw

a car weaving in
and out in front.

Did my car see it?

And that's the only way I'm
gonna truly be able to let go

and say, yes, I will
let you have autonomy in

this situation.

You have to make people feel
comfortable that the car is

seeing what you're seeing.

Yeah. I completely
agree with you.

My car has that capability when
I'm in sort of lane keeping

mode, lane lane assist mode,
kind of cruise control,

And the car in front slows down.

It's incredibly reassuring
to display on the dash that

shows "I see that car ahead,
don't worry, I got you,

and I'm going to be slowing
down in a in a careful measured

way" as opposed to I need
to take over control.

And that's probably my single
most important feature to me to

know that it sees
the car in front.

And it does.

I agree with you completely.

Now we recently did a
survey with with Omdia.

In fact, this show,

Informa is is affiliated with
Omdia and inform the Informa

Group.

We recently did a survey of
OEMs and tier one professionals.

And one of the questions
we asked was about

best use cases of AI.

Aside from
autonomous, of course,

but aside from autonomous,

one of the most popular
use cases of AI was for

preventative maintenance
and smart diagnostics.

And the observation from the
survey was that not only is it

a great use of AI,

but it's also a strong
opportunity to create either

customer loyalty or
potentially aftermarket

revenue source.

What's your take on that?

And do you see that
as a useful use of AI?

And do you believe that's helpful
for your drivers and customers?

I really do. I think I've
been testing it for years.

It always tests really,
really well various use cases.

I think it's very similar to
what we were calling, you know,

dynamic road view with respect
to the visual feedback of the

car knowing.

So the car knowing other things
with respect to tire wear,

payload indication, you
know, things of that nature,

that gives the customer
the peace of mind.

And that's what they
want first and foremost.

They don't need their car
to be as much as, you know,

I love an infotainment feature,

they don't need it to be a repeat
of their cell phone on wheels.

I think the return on
investment for AI in those

applications is really powerful
because when we looked at

predictive maintenance in
little buckets, it's like, okay.

I'm gonna put a new sensor
on the the brake pads.

I'm gonna put a new sensor here.

I'm gonna new put
a new sensor there.

So now I have these new
sensors that are costly.

I have to read those sensors.

I have to do something
with that data. Right.

And then sometimes
those sensors fail.

Right.

And so people are bringing
their cars to the shop,

and they're like, there was
nothing wrong with the car,

but the sensor failed.

You know?

And you hear that all the time,

and and they're paying
to replace that.

And I think with AI,

we can have virtual sensors

that are, you know, the
digital twin that is going to

process this data.

And without the added hardware
and the added risk of putting

that hardware in that
could potentially fail,

we could get a lot
of these use cases.

So it almost skips this whole
level of where we were trying

to deploy these things via one
at a time sensor solutions.

We might be able to jump that
whole scenario and start to

bring in predictive maintenance
applications much faster

with these types of analyses.

And that goes back to
what we're speaking about.

And and I think I shouldn't
speak to it too much because I

don't have the
depth in this area.

But if I understand correctly,
the SLM versus the LLM,

those highly trained
grounded AI models

that the experts in the
field will develop specifically for

those use cases,
and we deploy those.

So instead of thinking
of as AI as, like,

salt that we're sprinkling
over everything,

let's take those little grains
and decide where are we going

to best deploy them.

Right.

And in my mind, the predictive
maintenance is the key area

for true value to our customer.

We agree.

And we're hearing that
from our customers as well,

that predictive maintenance
and and smart diagnostics,

I think those are
sort of cousins.

So I lump those together,
is so important.

And our AI Technician product
we're we're exploring right now

and showing in the demo,

literally just behind this wall,

is is showing how you can use
a lot of existing data sources to

give a better experience.

You also mentioned something
that's so important and I think

under-appreciated is we don't have
to put new sensors in the vehicle.

A lot of the genius here and a
lot of the smarts is taking the

data we actually already have
and looking at it in new ways.

I kind of liken it to a
needle in the haystack.

The needle's already there.

If we just know to to look
for it and where to find it,

we can get the useful
information instead of having

to to put dedicated sensors
for different applications.

So it needn't be a cost burden.
It could be a cost benefit.

Absolutely.

And I also think that
analyzing the data has also

become a huge
juggernaut in terms of

we we thought we had to
collect all the data.

Now now we have all the data
where we're gonna store it,

how are we gonna process it?

And the interesting new
solutions and the new

technology is really
being strategic in,

I'm gonna collect this data
for this time because of this

reason, because AI is helping
me define those boundaries,

And then I can
collect less data.

I can collect more
relevant data,

and I can make
decisions accordingly.

So I think that is
another powerful use case

instead of this kind of a few
years ago mentality where it's

like, let's collect everything.

Yeah. Yeah. I agree.

I like it to a fishing net sort
of just trolling the ocean and

you get the fish,

but you also get fish you don't want
and lots of broken bottles or whatever.

Just really zooming in on
the things you're looking to

capture, I think
is so important.

Much more sensible with costs,

much more sensible with cloud
upload bandwidth and storage

and all of those things.

This has been such a wide
ranging conversation.

I'm so glad we were able to
finally get together and talk.

You've brought together your
perspective on your customers

is so invaluable.

And thank you for
sharing it with us.

Thanks for having me.

If you like what you're
seeing in this episode,

please like and subscribe
to see more like it,

both from shows like Auto Tech
and from around the world.

Thanks very much,

and we'll see you again
in another episode very soon.