- 6 minutes ago
AI capability is advancing fast, but enterprise impact still depends on execution. Embedding AI into real operations—across workflows, enterprise systems, and decision-making—remains complex, especially in environments shaped by legacy infrastructure, fragmented data, and growing demands around security, governance, and trust. As companies move beyond experimentation, what separates successful deployment from stalled initiatives? Where is AI already delivering measurable operational value? And what does it take to scale systems that are not just powerful, but reliable, secure, and built to last?
Category
🤖
TechTranscript
00:00So please be ready to welcome Anna Paula Assis from IBM and Alexandre Fredy from Orisha in conversation with Jad
00:10Shimali from EY.
00:35Hello, everyone. Great to be back at VivaTech. A few things have changed since last year.
00:41One of the main things is that technology is no longer the constraint.
00:46It's clearly that technology is a few steps ahead of where organizations are.
00:53AI has moved from use cases to being a bit more integrated into the DNA of organizations.
01:00And today we're going to talk exactly about this.
01:05How are companies moving from specific use cases to AI at scale?
01:10How companies are getting the right ROI out of their AI investments and what's next to come?
01:16So I'm joined today by two great colleagues or friends or becoming friends.
01:23I'm joined by Anna Paula Assis.
01:27She's the EMEA and AsiaPAC IBM lead.
01:30IBM, obviously, they don't need an introduction from AI to hybrid cloud to consulting, working with large scale organizations to
01:39get them to their future state.
01:41And I'm also joined by Alexandre Fredy from Orisha, embedding AI into core workflows, systems, and processes, and getting straight
01:53into execution.
01:55So today we're hoping to bridge exactly that, AI strategy, deployment, all the way down to execution.
02:03And hopefully we're going to bridge a few things.
02:05We're going to move from hype to evidence-scaled AI.
02:10We're going to talk about a few things.
02:13So with that said, I think I'm going to start with you, Anna.
02:18So many AI initiatives are failing to deliver on the returns that are expected.
02:26We see many use cases.
02:28Some are delivering not at scale.
02:30Some are not delivering.
02:32And their investments are kicking off again.
02:34And in particular, one of the key challenges that companies are facing is how to integrate AI and where does
02:42the value of AI reside within a process?
02:45Does it reside within a function?
02:47Does it reside in between functions?
02:50What's the cost of AI?
02:51You can't get to returns without truly understanding the cost of AI.
02:55So maybe with that, Anna, what are you seeing?
02:59And then I'll move on to you, Alexandre.
03:03But give us your take of what's happening and what are you seeing?
03:07Well, first of all, Jad, very happy to be here with you and with Alexandre.
03:11I hope we can have a productive conversation.
03:13And yes, I mean, AI is really a watershed moment.
03:16We can see that the organizations that are advancing fast are becoming the winners in this space.
03:24And the gap between those that are fast and adopting AI is becoming wider and wider, right, from those that
03:32are not doing that in the same consequential way.
03:35So how do I define obtaining the return on the investment on AI?
03:41It's not about doing isolated use cases.
03:44It's not about applying AI for individual productivity.
03:48It's really infusing AI in the way that the company operates.
03:52It's understanding your workflows, reimagining your workflows, and using AI to transform those, right?
03:58And the companies that are achieving that level of implementation, the stakes are very high because you can achieve up
04:06to 40% of productivity gains.
04:09So normally I like to use an analogy with electricity, right?
04:13When electricity started, the first application was a light bulb.
04:17So you would put light bulbs in factories that would make them cleaner, that would make them safer to operate.
04:24But that wouldn't change the way that the factories operated.
04:27It was really when the assembly line was established that we managed to really transform and increase productivity significantly.
04:35So it's about revisiting the process.
04:38And why do we see companies today struggling with that?
04:43Data in silos.
04:45Operational processes that are completely fragmented.
04:48And AI, for you to have that level of benefits to the organization, it has to cut across multiple disciplines
04:56in the organization, right?
04:58So you have to have a much more integrated view of your processes to operate.
05:03Now, everybody has tremendous expectations with AI.
05:07So from the studies that we have conducted with our clients, 80% of the clients expect that AI is
05:14going to significantly improve the top line of the business.
05:18But only 20% of them can tell exactly where and how this is going to happen.
05:25So I think the challenge that companies are facing right now is really how I map my processes, how I
05:31understand to infuse AI in those processes, and how I have a more integrated view of my data state.
05:40Great perspective.
05:43Alexandre, does that resonate with you?
05:45How do you see it?
05:46Yeah, clearly, I think the execution gap, I think everyone is seeing it.
05:50We are all frustrated by what we are delivering, even within our companies, even if we have good partners to
05:57work on.
05:58And I think that if we look at the execution gap, the main reason comes from external reason, and that's
06:05what Anna-Polla mentioned.
06:06Most of the organizations try to layer AI on top of an unchanged operating model.
06:12So it doesn't generate any transformational change.
06:16It just generates, you know, incremental change.
06:18And if you look also at the past experience before AI, companies have had lots of complexities in the way
06:24they are working, in the way they are delivering.
06:26So AI is not magic.
06:28I think the good question for a company is not where you should add AI, but clearly, if we had
06:35to renew the way we are working, do we rebuild the operations the same way?
06:42And very often, the answer is no.
06:44And coming back to your question about where does the value come from within the organization or between the organization,
06:51I'm quite convinced, like you, that most of the value comes across the organization.
06:57But having said that, I think it's very important to start within the organization for several reasons.
07:03First, it's easier, you know, it's easier to ask your CHRO and say, OK, you should improve all the processes,
07:11all the HR processes, because you have one person deciding.
07:15Second, I think it's a mandatory step.
07:17If you want to build a winning team in AI, a winning organization, you should have the best players on
07:23the pitch.
07:23We were just discussing the World Cup just before the session.
07:27And having the best player on the pitch in AI is making sure that HR is AI first, finance is
07:32AI first, sales, marketing, and so on.
07:36And I think that also thinking about the short-term value, you will experience the value.
07:44The frustration in AI comes from the fact that everyone is expecting value, but if you want to tackle the
07:48biggest problem in the company, that will never happen a long time ago.
07:52So that's great.
07:54You do need to have the best player on the pitch, but it's a team sport as well.
07:58At the end of the day, you only need to drive the connectivity.
08:00I do love many of the examples you mentioned, including the fact that only 20% of the companies have
08:07really linked their AI programs to growth, and specifically which element of growth.
08:12What we do see as well is that you have less than 15% of the companies have linked AI
08:18programs to their financial statements in general.
08:21So it's unclear where the value is going to come from.
08:24It's unclear how the cost structure is going to be impacted.
08:30And by default, it becomes a bit more difficult to move from an output-driven program to outcomes.
08:37Correct.
08:37Really, I mean, when companies get the true benefit is then when they focus on business outcomes rather than specific
08:46use cases and specific output KPIs.
08:49And maybe if you could share some examples about companies where you've seen this working and how it impacted their
08:57operating model or what did they do to drive those significant changes.
09:00Yeah, I'll start with IBM itself because we really considered AI as a key transformation element of our own business
09:09model, right?
09:10So we have started with a vision.
09:14We needed to generate cash for the company to invest in innovation, to invest in M&A, to invest in
09:21R&D, right?
09:22So how we were going to do that by elevating our productivity levels.
09:27And in order to do it, we mapped a lot of the processes that we had in the organization.
09:32We used the support of companies to give us the benchmarks.
09:36Okay, what good looks like for you to run your HR processes, your procurement processes, your financial organization.
09:43Once we had that target model, we started really to bring the different functions in the organization to work together
09:51in what we called MVPs at the time.
09:54Minimum viable products, prototypes of how AI was going to transform our business model.
10:00Originally, we started with a target to generate about $2 billion of productivity savings within three years of the project.
10:07We exited last year with $4.5 billion.
10:10So I think the work that we have done really demonstrates what good looks like, what is the potential.
10:18One of the areas that we put a lot of focus to the point that you were making before was
10:22in human resources.
10:24So we implemented an agent that today handles 94% of all the questions that come to our HR department
10:32in an automated way.
10:34And we did that with also increasing the employee experience.
10:38So our NPS that was around 20 is now at about 74.
10:43So significant improvement also in the satisfaction.
10:46So I think the lesson here is not only you can drive productivity, but it can also improve the services
10:53that you are providing.
10:54And for us, that had another effect, since it affected the entire population of IBM, our almost 300,000 employees,
11:04they understood what AI meant for them in their day-to-day jobs, in their day-to-day lives.
11:10And that created this flywheel effect where everybody in the organization started to play with the technology to see how
11:20that was going to transform the way that they're running their businesses.
11:23So I think that this approach is important because you have to bring people along.
11:28You have to ensure that people are embracing the transformation and that they see the benefits for that to happen.
11:36That is what is going to create the bigger impact, is when people really start to use AI to transform
11:42the way that they are doing their business.
11:44That's great.
11:44The upskilling is a big part of the programs as well.
11:48Just making sure that the employees are part of the journey and becoming better, more productive, happier, with a broader
11:56perspective.
11:57And they don't become a force of resistance to the transformation.
12:00They actually become catalyzers of the process.
12:04That's a great point.
12:05Alexandre, where have you seen AI getting steeped into workflows and working?
12:11And a few examples from your perspective.
12:12Yeah, I think as Ana Paula mentioned, first, we need to remove the barriers.
12:17And if we think about Orisha, we have the same barriers about cultural resistance, but also a fragmented environment.
12:23You know, different data, fragmented customer experience.
12:27So it's difficult.
12:29But for the cultural part, you can work on it.
12:33And what I think that now we are seeing in the market also some new category of professionals.
12:39People who can have operational judgments, business acumen, but also AI mastery.
12:45And the idea for a CEO like me is to identify that type of people.
12:51And I would say to give them full responsibility and let them do.
12:55Because they will understand everything around the life cycle.
12:59And if we come back a bit more about Orisha, so we have a transversal AI that we launched in
13:05January that is called Scout.
13:07But Scout is really specific to the different markets.
13:10So if you are speaking, for instance, about real estate, Scout is there to do some deep draftings to assist
13:20the real estate agents with live home staging.
13:23So, Jade, if you want to acquire a house in Paris, you have directly, you know, the vision of what
13:29it could be.
13:29So, at the end of the day, it helps conversion.
13:31It facilitates the life.
13:33It's time savings.
13:35In healthcare, it's totally different.
13:37Use cases, it's medical prescription.
13:39It's the capacity to anticipate illness based on X-ray.
13:44And to give you a third example, we are working also in a vineries.
13:48And in a vineries or vineyards, the idea is more to sort the groups on the bench.
13:54So, at the end of the day, it's AI, but dedicated AI that we co-design with our clients.
13:59And in terms of ROI, it's obvious because it's completely embedded in the value chain.
14:05That's great.
14:06The couple of thoughts, by the way, that you triggered.
14:10Anapala, I love your example about electricity.
14:12And when electricity started, it didn't change the way operations worked.
14:17It just gave better visibility to how operations worked.
14:21We talked about AI and how to get value out of AI.
14:26We haven't talked about the cost element of AI.
14:29Another example that might be relevant from an electricity perspective, that whenever we talk about the cost of AI, some
14:36of our minds go to tokens and the cost of tokens.
14:39However, token optimization, for me, is as if you're optimizing your electricity bill in a factory.
14:47You're completely, potentially, if you focus on the electricity bill, you're missing many of the other cost elements.
14:53From AI would be subscription, would be governance, would be risk, would be rework, change management, risk of failure.
15:01And so the cost elements of AI, many of them are visible, but many of them are invisible.
15:08So when we talk about AI programs, shifting from a view of what the cost umbrella would be to more
15:18better understanding the value system out of which cost is obviously an element is something that organizations are moving more
15:27and more towards the more they understand about the cost implication of AI.
15:31So maybe just pivoting a bit more about the value chain and how do you shift the mindsets from spending
15:38on AI to managing it as a series of value system, part of it workflows, but part of it much
15:47broader, much broader than that.
15:49So not just a portfolio of use cases, but something much more comprehensive.
15:54So, Alexandre, I'm going to start with you this time.
15:56How does that play out from where you sit?
16:00I think it's a very difficult question because if you look at where we are now at Orisha, I think
16:07it's the question that gives me the most headache.
16:10Because I don't know anyone, my investors in the market, who can tell me what will be the cost of
16:17AI three years from now, but even six months from now.
16:21So, of course, it's a concern.
16:23It's a concern for the CFO, but it's a concern for everyone in terms of value regression versus cost.
16:27So, we are just transitioning to a world where we were focusing on adoption, and now we are moving to
16:33a world of value creation.
16:35And I think that the way to frame that, little by little, I think that we will consider agents a
16:42bit like employees.
16:43So, when you look at employees, you wouldn't tolerate two employees doing exactly the same thing.
16:50As of today, in my organization, I have tons of agents doing exactly the same thing.
16:54And we don't see really the cost right now, but we will see in the future.
16:57So, we need to prepare a bit of the organization thinking like that.
17:01And if we speak more specifically about software developments, I think in 2025, even 2026, I wouldn't say it's not
17:09a topic.
17:10But last year, the share of wallet of the cost of AI compared to the wages of the software developers,
17:18it was 2% at Orisha.
17:20This year, it will be 5%.
17:23But I have software developers who consume 40%, 50%.
17:27So, we know the trajectory, and we know that these models were subsidized.
17:31So, it's a big concern thinking about, okay, we know that the value is behind, but we need to move
17:36from adoption to value creation.
17:38And I think it's clearly we are on it, but it's difficult because we are not mastering the cost side
17:43of the model.
17:44So, we just need to make sure that everyone has understood it.
17:47And the way we are working on it at Orisha is that now we try to put discipline in a
17:53way we are using agentic.
17:55So, even for the software developer, you should know day by day.
17:58And it's not because you consume 40% of your wages on AI that you consume well.
18:04So, it's also another topic.
18:06That's great.
18:06I mean, discipline is a key point there, obviously.
18:10And cost, some of the things you said, by the way, give me mild anxiety, which I do fully agree
18:15with.
18:16The fact that cost and six months from now are going to be very different from what they are now.
18:21Yet, we're signing off today on programs that we're going to be living with six to 12 months from now.
18:27So, costs are a big part of the equation.
18:32Visibility, transparency, discipline are obviously key.
18:35How does that translate to the enterprise level?
18:39So, I heard a great expression today.
18:42Somebody told me we are moving from the FOMO moment, the fear of missing out, to the fear of messing
18:48up, to the FOMO moment, right?
18:51So, and I think that we approach this from three very important layers.
18:57The governance layer, so having visibility of how the agents are operating.
19:03I actually, I think that this is the biggest nightmare that CEOs have today.
19:08I was actually having lunch with the CEO of a big insurance company a few weeks ago.
19:12And he said, before, I was super worried about cybersecurity.
19:16Now, I'm super worried that an agent is going to take an action that is happening somewhere in my organization.
19:23And I don't have visibility of that.
19:25So, governance is a very important element.
19:27And the orchestration of the multiple agents that you have in your environment.
19:32On the topic of application development, the way that we have approached this is really creating a platform that allows
19:40you to, I would say, broker between models.
19:43So, that you can define the best model that can be applied for the certain function that you are executing.
19:50This platform is called Bob, by the way, right?
19:53So, it's supporting all of our developers community to make better decisions on how to allocate their token spending for
20:02maximum productivity.
20:03While at the same time, managing the costs and ensuring that you have visibility of how the platform is operating.
20:09So, I think it's those elements that you mentioned are going to become more and more important for companies to
20:16build an AI stack that is responsible.
20:20And that addresses all the challenges that we are getting from our clients.
20:23That's great.
20:24Great perspectives.
20:26I did hear, obviously, FOMO, fear of missing out, you have now you introduced FOMO, the fear of messing up.
20:35And some companies are maybe thinking that they're going to get to a JOMO stage, a joy of missing out.
20:42Then hopefully, that's not going to be the case.
20:43No, that's not going to happen.
20:45That's not going to happen.
20:45We need to avoid that to happen, right?
20:47Absolutely.
20:48So, we need to make sure we avoid this happening.
20:50Which leads me to the next part of the conversation.
20:52Obviously, we're scaling, we're seeing some value, but how do we scale responsibly?
20:58It's a big part of the onus as on many of the organizations that are currently transforming.
21:08It's showing up in different parts of society, different parts of the world geographically.
21:15Scaling in a trusted way, in a responsible way is a massive part of the equation.
21:21So, Anna Paul, I'm going to start with you.
21:24How do you approach that scaling with balance between doing it responsibly, making sure you have the right level of
21:32oversight, while at the same time it's happening, innovation is happening at pace?
21:36Yeah.
21:37So, first, we are looking, the way that we are approaching at IBM is really looking at defining the platforms.
21:44So, the platforms are, I would say, very centralized in the company.
21:48The architecture that we operate in is very centralized.
21:52And we delegate to the functions, the processes, and how they want to implement that, right?
21:57But they need to be compliant with the standards that we define as a company.
22:02That is point number one.
22:04Point number two, there is tremendous, I would say, tone from the top, right?
22:08So, the direction of how you want to implement AI in the company is very well defined to align with
22:14our culture, to align with our strategic objectives, to align with our financial goals.
22:20And finally, education, education, education, right?
22:25So, the more that you can skill your organization with this technology, and every year we run what we call
22:32AI challenges in the organization, where we address our entire population to test how AI can transform their functions.
22:43It's very, very important, because they start to understand, really, the impact that it can create to the business, but
22:49at the same time, the benefits that they can get from it.
22:52But very much the governance coming from the top, I think, is very important.
22:56And absolutely.
22:57And education, to your point, is critical.
22:59We have 90% of our people have gone through AI training, and more than 200,000, so more than
23:0550%, are going through advanced AI training.
23:08And it becomes much more steeped into the DNA of an individual.
23:12Right.
23:13But I would love to get your perspective, Alexandre.
23:15Yeah, I think I will try to use the same framework, so an easy one, one, two, three.
23:19The first point is the same than Anapola.
23:21I think you need to think AI as a core part of the technology.
23:25Very often, we see that it's something on top.
23:29Looking at the security, the data, the governance, it's really the core of the technological stack.
23:34The second topic I see is more around observability and explainability.
23:41The specific of AI is that it's probabilistic and not deterministic.
23:46So it means that at two different moments, it can give you two different answers.
23:51So for a company with people moving for the trust, you need to generate for your clients.
23:56You need to have the trustability to know exactly what happened at which time.
24:01And the third one, I think it's a one, and maybe we underestimate that one, is a one, real ability.
24:08Now, you know, AI is a bit like the oxygen we inspire.
24:13And we experienced two weeks ago, you know, a downtime on the AI LLM.
24:19And my developers were almost on strike.
24:21So it's something that we really need to work to have, you know, a fallback mechanism to make sure that
24:28we have always an option to be up and running.
24:30Because it's really the value of the company which stops if the AI stops as well.
24:36Absolutely.
24:36I'd love to take this point about the value of the company.
24:40In 30 seconds or a minute or less, next year when we meet again, life is going to be very
24:47different.
24:47So the next 12 months are going to be pivotal.
24:51What do you think leaders should focus or what should the traits of a leader be over the next 12
24:56months to really lead to sustained enterprise value improvement?
25:01What would be your advice, starting with you, Alexandre?
25:05I think that for me the key word, and that's the key word we are using internally, you know, is
25:10courage.
25:12It's courage because it's really difficult.
25:15Courage, first, to start from the white page.
25:18We are just implementing, you know, a new lead to cash organization within Orisha.
25:24And we all have the easiest pattern to think the new processes the way we were doing before.
25:31How to force the organization to think two years from now with the AI capabilities.
25:36And that's hard.
25:37So courage to start from the white page.
25:40Second, courage to challenge the talent workforce.
25:44We see that you need to accompany, you need to train, you need to educate and educate, educate, educate.
25:51At some point, you need to make sure that you have the best players on the pitch.
25:55And, you know, now that the onboarding is much faster with AI.
25:59So calibrating the way you onboard people internally, the way you attract talent externally is also a must-do and
26:08requires courage.
26:09And last but not least, I would say courage to lead from the front, to lead.
26:14And it's the CEO topic.
26:16And it's difficult because you need to experience the value.
26:19You need to be credible when you talk about AI.
26:21So we are spending lots of time, you know, to foster AI within the leadership team to make sure that
26:27everyone will be credible when they will discuss with the software development teams.
26:31So definitely it's not a topic for the CPTO, for the CHRO.
26:35It's a corporate topic and the CEO.
26:36Absolutely.
26:38Yeah, I think don't underestimate the importance of cultural transformation in this journey, right?
26:46And I think we put a lot of focus on the technology.
26:49We put a lot of focus on the hard metrics, right?
26:52Productivity gains, financial impact.
26:54But at the end of the day, if people really don't buy in, if people don't really understand what's in
27:01it for them and how they can actually get the benefits of this transformation, it's going to be very hard
27:08to accelerate adoption going forward.
27:10So put a lot of emphasis on culture.
27:13Don't underestimate the importance of it.
27:16That's a great point.
27:16Courage, culture, bringing others along and be agile is the one that I would add.
27:22The next 12 months are very unpredictable.
27:25We know that technology is going to be changing at pace and the world is going to be changing at
27:29pace.
27:30So lead from the front, bring others along and make sure that you drive the right culture.
27:34So with that said, thank you very much.
27:36It was great.
27:37Thank you, everyone.
27:38Great seeing you.
Comments