We’re past the peak cloud hype of the 2010s and now well into the era of cloud maturity. But even as businesses have come to terms with multi-cloud and hybrid cloud models, new considerations have arisen, such as the growing push for cloud sovereignty and regional data center demand.
AI has put additional pressure on firms to figure out exactly what they’re running in the cloud – as agentic AI drives organizations to consider ‘always on’ workloads and whether this is feasible in a public cloud model.
How can leaders navigate this changing cloud landscape and what is on the horizon?
In this episode, Rory speaks to Alex Grimshaw, Chief Technology Officer (CTO) for Microsoft Azure, UK, to explore what he sees as the key trends for the cloud and how AI will shape it in the coming years.
We're now past the peak cloud hype of
the 2010s and well into the era of cloud
maturity. But even as businesses have
come to terms with multicloud and hybrid
cloud models, new considerations have
arisen, such as the growing need for
cloud sovereignty and regional data
center demand. AI has put additional
pressure on firms to figure out exactly
what they're running in the cloud. as
Aentic AI drives organizations to
consider always on workloads and whether
this is feasible in a public cloud
model. So, how can leaders navigate this
changing cloud landscape and what's on
the horizon? You're listening to the IT
Pro
[Music]
podcast. Hi, I'm Rory Bathgate. Today
I'm delighted to be speaking to Alex
Grimshaw, chief technology officer for
Microsoft Azure UK to explore what he
sees as the key trends for the cloud and
how AI will shape it in the coming
years. Alex, welcome to the show. Thanks
Roy. Great to be uh on your show and uh
yeah, looking forward to a very um
engaging conversation.
So I'm going to start with a broad
question. Um we're sort of postrush to
adopt cloud services for for a lot of
firms right now. Um how would you
describe the current era for the cloud?
Yeah, it's a very interesting time as
you as you suggest it's um the way I
would frame it is when I've previously
uh engaged with customers and certainly
partners as well over the last couple of
years. It was generally the conversation
was like how do I garner a business
value from the cloud in general like how
do I what is this cloud thing and how do
I get value from it I think certainly
more over the last I'd say 12 months the
conversations have shifted now where
it's more how do we transform our
business through the use of AI you know
and what are the steps to embracing this
technology so what are the various
stages of how we get there so for
example I'm seeing far more acceleration
of a data strategy for example so cloud
maturity for me certainly for my my
engagements whether that's from like
seale or IT leaders more from a we're
very comfortable with
IAS but now what's our approach to
becoming more pervasive or using more
sophisticated services like AI and how
do we enable or garner an internal kind
of culture and kind of process where we
can adopt these these faster moving
services like AI uh and micros service
etc in this kind of new bold world
that's certainly what I'm seeing that
shift
happen you've said the word that comes
up so frequently now which is AI of
course if we can start maybe a few years
ago what are some of the key changes
that that the adoption of generative AI
kickstarted in in the cloud either with
Azure specifically or cloud more
generally what are some of the
architectural shifts we've seen. Yeah, I
think twofold. First of all, a big part
of the conversations I generally have
are it's more about culture for me. It's
the cultural shift that I've seen change
and that's not only from IT leaders and
IT departments and the people that work
in those departments, but more so across
the entire organization, right? Because
traditionally, yeah, it should be seen
as an enabler. But as we know, a lot of
business leaders and business function
leaders would certainly see it as being
some sort of hindrance or or what are it
going to do again to us? They're going
to slow down our operation. They b they
just want their operations to continue
running, right? I've certainly seen a
shift now
where most business leaders understand
that without the IT department and a
collective unison of everyone going on
this journey where ultimately the what
is the vision utilizing AI what are the
business outcomes that we're trying to
realize like how do we get there
together as I said there was certainly a
fragmentation a few years ago where it
was something that IT people
But now this AI kind of embracement is
certainly something that the entire
organization can take the journey
together. And of course as we know
taking a journey together means it's
going to be more far more likely to be
successful as well. It's interesting to
see that evolution that you're
describing whether it's on the cultural
or the technological level um happen in
tandem with some other changes. So for
example the shift to more more hybrid
cloud and more multiloud from some
businesses do you see those as sort of
aligned journeys that are are
complimentary of one another so say when
it comes to AI and a mixture of or a
hybrid approach with some public some
private cloud is this a complimentary
approach for businesses to take it is in
in its I see two contrasting I'd say
behaviors one from those industries
which are legislatively or regulatory
constrained or mandated or governed. So
you think public sector, FSI where of
course they you know it's a more it's a
more tentative adoption of cloud in
general. So whether there's data
sovereignty requirements or data
residency requirements and of course
those two things are very different and
although clearly a lot of people use
them in um those two uh terms as the
same thing clearly they are not the same
thing but I certainly see yeah a more
tentative shift uh in terms of data
classification levels uh obviously uh
snapping to policies and regulatory
needs but then of course the advantage
that those industries is that are not
one of those regulated or legislative
industries have the opportunity I would
say to move much faster. So think about
for example digital natives or startups
or general commercial organiz
organizations like retail or
manufacturing for example they can be
far more run at their own pace in terms
of how comfortable they are in terms of
do they go all in on public cloud do
they dip their little toe into the
public cloud while remaining on prem do
they have a mixture between the two
because they're still not quite sure
about asset governance and where the
assets need to lie or should lie.
Obviously that's based on where they
need to consume them from as well. So I
think there's very I see certainly very
different behaviors based on whether
they're in an industry which as I say is
governed versus an industry where they
don't have the same kind of shackles or
they have to appreciate a central body
or authority.
There's a a few different things to get
into there that you've raised. Um, just
to start with that feeling of still
being able to to try things out, maybe
dip your toe into into public cloud and
keep a bit on prem. In some
conversations, you are hearing an
increasing sense of urgency. You know,
um, it feels like every year now we hear
last year was the year to try things
out. This year is the year to deploy. In
your conversations with customers, is
there still that desire to to sandbox to
to to try things out? um and are
customers beginning to say be worried
about being left behind or being behind
when it comes to AI? Yeah, I think
there's certainly some sense of formal
across certain industries. If I give you
an example, so prior to my current role,
I was the CTO of HMRC.
Clearly, you will have had some dealings
with that organization. Be well aware of
what they what they uh aspirations are.
when HMRC a number of years ago realized
that they needed to embrace public cloud
not only from obviously from a a cloud
economics perspective but also from an
efficiency and um modernization agenda
perspective they you know you I think
you have two options as an organization
if you're looking at public cloud one is
you either go let's find a really small
discrete completely independent workload
load and we move that into the cloud. We
may be architecture as well. We we
clearly don't want everyone lifting and
shifting as as it is because that's not
the most efficient way to run cloud
workloads. But if you can move that and
then operate it in the cloud, it's
almost like as you say that first step
into the public cloud. It allows you to
create some governance models, some
patterns, maybe an enterprise landing
zone, all them kind of good things. Or
you go for the most
gnarly super
independent almost even having various
stack technologies as well. And you go
you know what if we can move that thing
and operate it in the cloud that means
we can move
anything. M and what was what was
obviously in retrospect looking back it
is a great it was a great uh decision
point but HMRC went for actually the the
biggest most nariest workload to try out
they moved the enterprise tax management
platform which collects over 365 billion
pounds worth of tax revenue every year
you don't get much more mission critical
than that in the UK um so they actually
went and uh we in collaboration with
obviously the customer in Capgeemini
We moved that workload to Azure. So it
was SAP based and that still operates
today.
Completely mission critical. It operates
there today. But of course what that
meant to the director generals, the
board, the rest of the organization was
look if we can move this, we can move
your business workload safely,
consistently and actually we'll get so
much more business benefits from doing
so. So go on this journey with us. as I
said it then it really has that great
effect on culture. It's those kind of
case studies that um that's a reason
that we like to discuss them on the
podcast and on the website because you
know organizations I think appreciate
those examples. One of the other points
of um anxiety or interest that you
mentioned there was data residency and
data sovereignty and cloud sovereignty
we know is something that um a rising
number of UK organizations are are
interested in. So we had um coverage
earlier this year that over half of UK
organizations now say it's critical to
their data management strategy. I wonder
if first of all you could tease out the
differences between the two that you
hinted to earlier. Yeah J I mean outside
of public sector where there's a very
defined data classification levels
um you've got official official
sensitive as a handling caveat sits
within that official classification then
you've got secret and top secret. You
know clearly secret and top secret
generally applies to certain
organizations
uh and they generally as you imagine and
you will uh respect certainly have very
different uh perspectives on what data
sovereignty means to them. Traditionally
they will keep a lot of uh their
workloads on prem and it absolutely
makes sense to do that. the the more the
official and official sensitive
uh classifications of course in public
sector. It's about the organization's
risk appetite and actually then really
understanding their assets and their
data as well. Who needs to consume it?
Where does it need to lie? What are the
regulations or government policies that
dictate where that has to reside and
things like that as well. So public
sector is a bit more defined. other
organizations where it might be a
preference from a data residency
perspective because of course it would
depend on where their consumers want to
consume the data from. You know, you've
obviously got performance challenges
there, latency, what are the
independencies there to other workloads
that might not sit in that same
geography as well. So you got all of
those factors. For me, data residency
and data sovereignty generally means
from a data residency perspective that
there is a requirement there that all of
your data is resident and never leaves
the shores of
a defined boundary or geographical
boundary. Data sovereignty for me means
if during the operation of that workload
if the if there for example is a support
expectation that that support function
would not be outside of the geographical
constraints of that region again. So for
me sovereignty generally means
absolutely there was there's no eyes on
that data outside of that boundary but
data residency just for me means that
the data resides in that in that place
and is data sovereignty a concern that's
coming up in your in your conversation
with with customers right now. No,
because again I think the terminology of
what that means I think it's describing
what that means and ultimately you could
argue that it it has nothing to do with
the definition right it's all about the
what the customer needs are like what's
driving their anxiety or their need to
have data in a specific place sometimes
it's a
ultra risk averse position of course as
you can imagine and when you look at the
current backdrop of recent events then
you can understand why that would be
right. So obviously we've got the M&S
incident that happened quite recently in
the public domain. So people are more
concerned about how they protect their
assets. So I get that but that doesn't
necessarily mean oh if we host it in the
UK that means that it's more secure. It
doesn't mean that at all. It residency
sovereignty and securing your assets are
very different things. So it's uh I
think it's an educational conversation
that I generally have with customers
more than a this is something that we've
defined collectively let's let's start
from there. It's more of let's have a
conversation about what you think that
means or what your outcome is that
you're trying to deliver and then we'll
talk about that instead of definitions.
That makes a lot of sense. When it comes
to residency as well, is there the
requisite infrastructure for for all the
residency demands in the UK right now?
And what work is well Microsoft but also
the sector doing to boost availability
for things like data centers within the
UK? Oh, for quite a number of years now,
we've had two regions or two separate
regions, what we refer to as UK South
and UK West. uh we made in November,
it's November 23, we made a huge
commitment to UK government that we
would make a huge investment in UK AI
infrastructure particularly and
obviously skill a skilling commitment is
there because of course it's not it's
not only ideal just to have
infrastructure in the UK but of course
you want to be able to skill up people
so they can actually utilize that
infrastructure. So yeah, we made a huge
investment that we would invest that
over a 3year period. So again, that huge
investment should come to full fruition
by certainly the end of I'd say
[Music]
2026. And on that point of skills, um,
is this as much skills within the
organizations that are looking to
utilize these data centers as much as it
is data center operators and and
obviously people to actually build the
things.
Yeah, it's I mean, you've kind of hit on
a particular passion of mine, which is
is tech intensity. I I have a a one of
my various hats as the CTO in the UK
businesses. I lead on on learning for
the organization. So and that's three
bits for me. One is tech intensity which
we all know what that is right. It's
having the understanding to be able to
utilize our technology but also for me
as important is how understanding of how
our customers use our technology which I
call industrial empathy because I
suppose how do you write and produce
products and services if you don't
really understand how customers would
use them. So yeah, so for me tech
intensity actually means we need to be
educated as customer zero almost in our
own organization and we have a huge
amount of efforts
um and investment in that learning
process to ensure we're all really
comfortable and really conversant around
our technologies and we're able to use
them all. But then of course, yeah,
that's about translating that to
customers because what's the point in
having this great innovation in the
cloud if if people aren't feeling
confident uh and capable of of utilizing
that that new
technology on the point of the new
technology that you've just mentioned,
we spoke a little bit about AI earlier.
Um but I feel that we do have to we do
have to dig in a little bit on um what
this transformation is that's that's
going on and and and what deploying AI
in the cloud actually looks like. So two
years ago we were talking about LLMs
primarily when we were discussing
generative AI we were discussing AI
workloads. Now attention has largely
shifted to agentic AI. So we heard just
recently that Microsoft expects 1.3
billion AI agents to be operational by
2028. So how does all of this impact the
cloud and what needs to be done to
support all of this deployment?
Yeah, it's interesting isn't it? If I if
I touch on my conversations with
customers, more recent conversations,
I think a lot of organizations really
have a long way to go still to really
understand the true value of AIDM and
gen
AI. Some AI simply means co-pilot type
functionality or it means a human
interface to an AI entity or a more
informed chatbot. Even in some places I
think there's a very rich tapestry of
functionality and autonomy that a agent
AI can bring to bear and it's certainly
obviously a trans a significant
transformation enabler for many
organizations. I mean, I think about
like C to C, B to C, B2B. Um, because
like enabling autonomous AI agents is
going to really supercharge like human
productivity
because, and I know there's some
concerns and maybe we can address some
of them later, but I think that this is
such a revolutionary time. I think back
to how like the internet uh massively
changed the entire retail sector for
example in that it created so many new
roles because it it released almost the
repetitive burden and some of the
backend orchestration from how people
operated and then this resulted in
creating brand new roles. So like
digital marketers, web web designers,
content creators, etc. So I think we're
going to have a
similar piece here in terms of how we
absolutely define and revolutionize
certain industries. There's definitely a
lot of conversations happening around
how much of this can be used to um you
know reduce toil and and help automate
tasks and and kind of fulfill that
automation picture um that we've been
discussing for some years now. Before we
get into some of those challenges which
you you hinted at or or potential
concerns around AI, I've also had some
interesting discussions recently around
how specifically AI agents present a
different picture when it comes to
compute demand compared to um other
workloads primarily because these are
workloads that need constant inference
um and and this can have an effect on
the kind of pricing when it comes to to
running AI agents in the cloud. Are
these conversations already happening
with customers? Um, and how do you see,
you know, AI agents versus say an LLM
workload as as different in the cloud?
Yeah, it's it's an interesting point. As
I said, I think I think there's such a
maturity curve that needs to take place.
Yeah. For example, when you think about
the areas in a
business that can be utilized by if we
if we kind of just bucket AI together in
one as one big kind of set of
capabilities like if you
think four distinct places for me that
where AI can have benefits is like
reshaping business processes. So I think
about if we've got humanborn
workflows. So humans have created
workflows which essentially are business
processes within organizations. Now
because they're human
created obviously there there's certain
inefficiencies there because there's a
lot of things that humans don't do very
well. So I certainly see Agentic models
and the orchestration there of certainly
having a big part in this almost space
where almost like a fire and forget
almost position where you want to do
something. It's a business process off
it goes and actually the agents are
doing all the work in the background
rather than people which of course then
releases us to do more creative things
and more things that are more I would
say critical thinking and that's
certainly something let's come back to
because I've got a real passion about
this critical thinking that we need to
continue to have rather than AI
replacing some of that. Let's come back
to that. And of course then we've got
things like reinventing customer
engagement. So I talked about HMRC
before, you know, one of the biggest
contact centers in Europe, HMRC have
got. So when I think about the
introduction of multimodal interactions
for example, having agents that are just
listening almost on all these different
types of channels and when I think about
another passionate area of mine is
neurodyiversity and inclusion and again
that's a huge area, right?
multimodal has such vast benefits for
those people that want to communicate
with the organizations in very different
ways. And then business model innovation
of course or bending the curve on
innovation. AI is going to have make
huge strides in certain industries like
drug discovery or digital manufacturing.
You know, every industry is ripe for
some sort of business model innovation
or just general innovation in in some
form. And then when I think about
employees themselves is this is about
enriching the employee experience. So
not only increasing everyone's
productivity but also when you're
attracting talent and retaining talent
to the market. You know, again, you
would have a certain stereotype of
certain organizations, maybe some in the
public sector where, you know, if you
came to work for that organization, you
wouldn't get to use the latest and
greatest tech and you'd be given a brick
of a laptop and it wouldn't necessarily
be running the latest tech.
Organizations by embracing AI can be
seen as employers of choice and very
attractive to people entering the
workforce as well. So that there's lots
in
there. I just think that um I still feel
there's some around customers where
they're thinking right where do I start?
I'll start that I'm still finding people
are thinking about how do I use this
co-pilot thing as almost that first
step. How do I get confidence in what AI
is because I want to see what it does,
what business value it brings. Is it
transparent? Do I understand how it's
deriving outcomes and content? Yes. And
then of course then you you'll think
about the little pilots who actually
using machine learning or AI where it's
about okay let's see what this can do in
an isolated incident where it's not
about human interaction. It's about it
actually doing a task or a thing in
itself. And then growing from that
you'll get interdep departmental AI
where actually the agents then come in
and start to do departmental
integrations and tasking as well. And
then ultimately if you've t again coming
back to the cultural piece if it's all
everyone's on that bus then of course
the entire organization really embraces
AI and you've got a culture of trying
trying new things with AI and actually
everyone ultimately trusting AI and the
value that it brings. So it's
interesting that point you made there
around some of these small pilot schemes
that maybe those more hesitant firms are
running. Are these more about improving
trust in AI and understanding of AI or
are they about demonstrating potential
return on investment when it comes to
AI? Yeah, all of the above. Certainly
the the you've got to have some measures
of success or metrics in there.
Obviously, one of those is the ROI one
is is this do we have what what form of
skilling do we require to make this
scalable and repeatable if it's
successful? Have we got skills across
the organization to utilize and build
and operate this kind of workload using
these more pervasive services? And then
of course, yeah, you've got you've got
all the metrics that you would normally
associate with any pilot of any form
using technology or any other new
innovation where what business value
does this ultimately bring? And actually
you you touched on it earlier from a
cloud economics perspective do these
services that we're consuming whether
they are uh LLMs or they are
organizations where they're even triing
creation of their own LLMs or even SLMs
and then ultimately you've got aentic
stuff uh and maybe smaller consumable
chunks of more pervasive services. It's
like at what again what maturity level
do they have existing knowledge and
patterns and governance and enterprise
architectural principles and solution
architectural principles and data
architectural principles and all the
others to actually embrace and almost
through
osmosis pull this into their existing
operation certainly from the enterprise
architecture perspective and then how
scalable is this thing you know did it
take an army to build this little pilot
that we're not get seeing the return.
It's obviously not going to convince
anyone from a trust perspective as you
say. It's not going to potentially win
over the the security naysayers, but
ultimately it's got to be successful for
all stakeholders involved and all
stakeholders have to have their own
metrics of success as well for that
particular pilot. You mentioned security
naysayers just now and and earlier you
you hinted at you know there were at
least initially some concerns around the
technology itself. Earlier on you know a
year two years ago there were a lot of
conversations happening around the
potential data security concerns of AI
particularly AI in the in the public
cloud. Are these still concerns that you
hear from customers and what are some of
the the main hurdles as far as customers
see it to their adoption? Yeah, I'm I'm
not hearing uh any concerns from my
customers. I mean, obviously some of the
things I might have heard uh or seen in
the press certainly point to some
concerns around and I'm not sure where
this is been born out of to be fair that
somehow tools like Copilot can access
data that you yourself couldn't actually
access. I mean it doesn't it doesn't
make sense because if you think about
how copilot works it's essentially an
inheritance model where you are
inheriting the the data governance and
access rights that you already have to
the ex underlying data in the
organization anyway. So it has no
ability to bypass them things because
for one it's operating in the context of
you your user authority. It doesn't have
some s central Uber admin rights where
it can just do whatever it feels like
because that I mean who would release a
product that would do that anywhere?
Yeah. But again it's just inheritance.
So if you've got a robust data
governance model in
place, any any any
uh AI agent user interface should not be
able
to view over any data and certainly not
consume any data or um retrieve any data
that's outside the bounds of you as an
individ individual using that. So in in
your context, it shouldn't have any
access to any data that you wouldn't
have access to anyway. All it's doing
essentially is just as we know LLMs are
fantastic at at language. So ultimately
it's you know it's the use of language
but they still have the same access to
the data. It's their interpretation of
what data we require that's the the
value here not the fact that they can
very weirdly like circumvent security
governance.
So another concern um around the
potential impact of AI has come from
Microsoft itself actually. Um earlier in
the year there was a report on how
generative AI may impact um the critical
thinking skills of employees. Um
Microsoft and Carnegie Melon University
found that there's a potential for um
critical thinking skills to be impacted
by AI tools if you don't take the right
approach. um what would you say about
this and what do you think the right
approach is to prevent this from
happening? Yeah, it's a great question.
Yeah, you're right. The research of
Microsoft, we released a paper that kind
of challenged
the almost the idea that the co-pilot
type technologies were almost dumbing
down our ability to critically think
about scenarios because a lot of people
may utilize a tool like that and just
take for granted what the response is.
There's no, it's not augmenting human
thought. It's almost replacing human
thought, which is clearly not the idea
here. What I've been doing internally,
and I mentioned before, I'm one of our
learning leaders. I've been running
workshops with groups of our employees
to say, look, you know, when you think
about the use of things like C-pilot,
there's two areas I want you to really
think about. One is filtering out the
noise, the the volume of emails,
chats, teams messages, outside noise. It
can be so overwhelming at times.
Basically, if you if you come back from
leave, for example, for a couple of
weeks, you will come back and there's a
huge amount of things to go through. AI
is great at sifting through that. Maybe
prioritizing things, but actually, you
know, at the end of the day, what's a
priority for you might not be a priority
for the co-pilot. So, you've got to
bring a level of critical thinking to
what it thinks is important and what you
should focus on first. So, kind of bring
those two things together. And then what
I've talked about is critical thinking
in general. It's that I think maybe it's
a cultural thing, but I constantly
challenge any output from AI, not
because it's prudent to do so
because, you know, not it's nothing to
do with trust. It's just it can only
relay what it sees based on the content
and the the prompt that you've gave it.
If you've given it a terrible prompt,
you cannot just expect uh some magic
source of content that comes out of
that. So yeah, internally I've been
running workshops for us to think about
critical thinking as a topic and
utilizing a combination of critical
thinking and output from AI but actually
the the end result should be a
collaboration between the two. That's
definitely something that organizations
will have to contend with in the coming
years and it'll be interesting to see
the the the strategies that people
deploy to uh ensure that they are as you
say collaborating with AI rather than
letting it replace those functions
entirely. On that note and looking at
the cloud more broadly, what do you
expect in terms of transformation in the
next two to three years? What are some
changes that that organizations can can
look forward to? It's it's a super
exciting time to be alive as a
technologist. I'm sure you feel exactly
the same way reporting on technology as
well. It's I mean it's just amazing
right? So I you know I think about yeah
AI geni agentic AI even quantum right
quantum has been around for over 40
years now as a concept but I still feel
that it's very specialized use cases for
quantum but I feel almost in a similar
vein that it's all about trust and that
adoption will be purely based on trust.
It will also be based on I think the
velocity of the adoption of AI based
services will will be largely dependent
on trust. It'll be based on regulations
as I said consumer trust and
organizational trust and I think almost
it will also be critical around like the
normalization of societal expectations
of what AI delivers. I think it just
needs to become something that's I think
many people are still fearful of. As
long as it continues to be
explainable, I think that we will see a
steady adoption of of this amazing
technology. And I just think over the
next two or three years, I don't think
we'll see a massive spike in usage. I
just think that everyone's on their own
journey. And as I said, culture is as
important as the adoption of the actual
technology.
Well, unfortunately that's all we have
time for today, but Alex, thanks so much
for joining us. Great. Thanks for your
time, Rory. As always, you can find
links to all of the topics we've spoken
about today in the show notes and even
more on our website at itpro.com. You
can also follow us on LinkedIn as well
as subscribe to our daily newsletter.
Don't forget to subscribe to the IT Pro
podcast wherever you find podcasts. And
if you're enjoying the show, let us know
by leaving a rating or review. We'll be
back next week with more from the world
of IT, but until then, goodbye.
[Music]