Monday, January 28, 2019

Who, if anyone, is in charge of multi-cloud business optimization?

Credit: Wikimedia Commons
The next BriefingsDirect composable cloud strategies interview explores how changes in business organization and culture demand a new approach to leadership over such functions as hybrid and multi-cloud procurement and optimization.

We’ll now hear from an IT industry analyst about the forces reshaping the consumption of hybrid cloud services and why the model around procurement must be accompanied by an updated organizational approach -- perhaps even a new office or category of officer in the business category.

Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. 

Here to help explore who -- or what -- should be in charge of spurring effective change in how companies acquire, use, and refine their new breeds of IT is John Abbott, Vice President of Infrastructure and Co-Founder of The 451 Group. The discussion is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.

Here are some excerpts:

Gardner: What has changed about the way that IT is being consumed in companies? Is there some gulf between how IT was acquired and the way it is being acquired now?

Abbott: I think there is, and it’s because of the rate of technology change. The whole cloud model is up over traditional IT and is being modeled in a way that we probably didn’t foresee just 10 years ago. So, CAPEX to OPEX, operational agility, complexity, and costs have all been big factors.

Abbott
But now, it’s not just cloud, it's multi-cloud as well. People are beginning to say, “We can’t rely on one cloud if we are responsible citizens and want to keep our IT up and running.” There may be other reasons for going to multi-cloud as well, such as cost and suitability for particular applications. So that’s added further complexity to the cloud model.

Also, on-premises deployments continue to remain a critical function. You can’t just get rid of your existing infrastructure investments that you have made over many, many years. So, all of that has upended everything. The cloud model is basically simple, but it's getting more complex to implement as we speak.

Gardner: Not surprisingly, costs have run away from organizations that haven’t been able to be on top of a complex mixture of IT infrastructure-as-a-service (IaaS), platform-as-a-service (PaaS), and software-as-a-service (SaaS). So, this is becoming an economic imperative. It seems to me that if you don't control this, your runaway costs will start to control you.

Abbott: Yes. You need to look at the cloud models of consumption, because that really is the way of the future. Cloud models can significantly reduce cost, but only if you control it. Instant sizes, time slices, time increments, and things like that all have a huge effect on the total cost of cloud services.

Also, if you have multiple people in an organization ordering particular services from their credit cards, that gets out of control as well. So you have to gain control over your spending on cloud. And with services complexity -- I think Amazon Web Services (AWS) alone has hundreds of price points -- things are really hard to keep track of.
Gain New Insights Into
Managing the Next Wave
Of IT Disruption
Gardner: When we are thinking about who -- or what -- has the chops to know enough about the technology, understand the economic implications, be in a position to forecast cost, budget appropriately, and work with the powers that be who are in charge of enterprise financial functions -- that's not your typical IT director or administrator.

IT Admin role evolves in cloud 

Abbott: No. The new generation of generalist IT administrators – the people who grew up with virtualization -- don't necessarily look at the specifics of a storage platform, or compute platform, or a networking service. They look at it on a much higher level, and those virtualization admins are the ones I see as probably being the key to all of this.

But they need tools that can help them gain command of this. They need, effectively, a single pane of glass -- or at least a single control point -- for these multiple services, both on-premises and in the cloud.

Also, as the data centers become more distributed, going toward the edge, that adds even further complexity. The admins will need new tools to do all of that, even if they don't need to know the specifics of every platform.


Gardner: I have been interested and intrigued by what Hewlett Packard Enterprise (HPE) has been doing with such products as HPE OneSphere, which, to your point, provides more tools, visibility, automation, and composability around infrastructure, cloud, and multi-cloud.

But then, I wonder, who actually will best exploit these tools? Who is the target consumer, either as an individual or a group, in a large enterprise? Or is this person or group yet to be determined?

Abbott: I think they are evolving. There are skill shortages, obviously, for managing specialist equipment, and organizations can’t replace some of those older admin types. So, they are building up a new level of expertise that is more generalist. It’s those newer people coming up, who are used to the mobile world, who are used to consumer products a bit more, that we will see taking over.

We are going toward everything-as-a-service and cloud consumption models. People have greater expectations on what they can get out of a system as well.

Also, you want the right resources to be applied to your application. The best, most cost-effective resources; it might be in the cloud, it might be a particular cloud service from AWS or from Microsoft Azure or from Google Cloud Platform, or it might be a specific in-house platform that you have. No one is likely to have of all that specific knowledge in the future, so it needs to be automated.

We are going toward everything-as-a-service and cloud consumption models. People have greater expectations on what they can get out of a system as well.
We are looking at the developers and the systems architects to pull that together with the help of new automation tools, management consoles, and control plans, such as HPE OneSphere and HPE OneView. That will pull it together so that the admin people don’t need to worry so much. A lot of it will be automated.

Gardner: Are we getting to a point where we will look for an outsourced approach to overall cloud operations, the new IT procurement function? Would a systems integrator, or even a vendor in a neutral position, be able to assert themselves on best making these decisions? What do you think comes next when it comes to companies that can't quite pull this off by themselves?

People and AI partnership prowess

Abbott: The role of partners is very important. A lot of the vertically oriented systems integrators and value-added resellers, as we used to call them, with specific application expertise are probably the people in the best position.

We saw recently at HPE Discover the announced acquisition of BlueData, which allows you to configure in your infrastructure a particular pool for things like big data and analytics applications. And that’s sort of application-led.

The experts in data analysis and in artificial intelligence (AI), the data scientists coming up, are the people that will drive this. And they need partners with expertise in vertical sectors to help them pull it together.

Gardner: In the past when there has been a skills vacuum, not only have we seen a systems integration or a professional services role step up, we have also seen technology try to rise to the occasion and solve complexity.

Where do you think the concept of AIOps, or using AI and machine learning (ML) to help better identify IT inefficiencies, will fit in? Will it help make predictions or recommendations as to how you run your IT?
Gain New Insights Into
Managing the Next Wave
Of IT Disruption
Abbott: There is a huge potential there. I don’t think we have actually seen that really play out yet. But IT tools are in a great position to gather a huge amount of data from sensors and from usage data, logs, and everything like that and pull that together, see what the patterns are, and recommend and optimize for that in the future.

I have seen some startups doing system tuning, for example. Experts who optimize the performance of a server usually have a particular area of expertise, and they can't really go beyond that because it's huge in itself. There are around 100 “knobs” on a server that you can tweak to up the speed. I think you can only do that in an automated fashion now. And we have seen some startups use AI modeling, for instance, to pull those things together. That will certainly be very important in the future.

https://451research.com

Gardner: It seems to me a case of the cobbler’s children having no shoes. The IT department doesn’t seem to be on the forefront of using big data to solve their problems.

Abbott: I know. It's really surprising because they are the people best able to do that. But we are seeing some AI coming together. Again, at the recent HPE Discover conference, HPE InfoSight made news as a tool that’s starting to do that analysis more. It came from the Nimble acquisition and began as a storage-specific product. Now it’s broadening out, and it seems they are going to be using it quite a lot in the future.

Gardner: Perhaps we have been looking for a new officer or office of leadership to solve multi-cloud IT complexity, but maybe it's going to be a case of the machines running the machines.

Faith in future automation 

Abbott: A lot of automation will be happening in the future, but that takes trust. We have seen AI waves [of interest] over the years, of course, but the new wave of AI still has a trust issue. It takes a bit of faith for users to hand over control.

But as we have talked about, with multi-cloud, the edge, and things like microservices and containers -- where you split up applications into smaller parts -- all of that adds to the complexity and requires a higher level of automation that we haven’t really quite got to yet but are going toward.

Gardner: What recommendations can we conjure for enterprises today to start them on the right path? I’m thinking about the economics of IT consumption, perhaps getting more of a level playing field or a common denominator in terms of how one acquires an operating basis using different finance models. We have heard about the use of these plans by HPE, HPE GreenLake Flex Capacity, for example.

I wrote a research paper on essentials of edge-to-cloud and hybrid management. We recommend a proactive cloud strategy. Think out where to put your workloads and how to distribute them across different clouds.
What steps would you recommend that organizations take to at least get them on the path toward finding a better way to procure, run, and optimize their IT?

Abbott: I actually recently wrote a research paper for HPE on the eight essentials of edge-to-cloud and hybrid IT management. The first thing we recommended was a proactive cloud strategy. Think out your cloud strategy, of where to put your workloads and how to distribute them around to different clouds, if that’s what you think is necessary.

Then modernize your existing technology. Try and use automation tools on that traditional stuff and simplify it with hyperconverged and/or composable infrastructure so that you have more flexibility about your resources.

Make the internal stuff more like a cloud. Take out some of that complexity. It's has to be quick to implement. You can’t spend six months doing this, or something like that.
Gain New Insights Into
Managing the Next Wave
Of IT Disruption
Some of these tools we are seeing, like HPE OneView and HPE OneSphere, for example, are a better bet than some of the traditional huge management frameworks that we used to struggle with.

Make sure it's future-proof. You have to be able to use operating system and virtualization advances [like containers] that we are used to now, as well as public cloud and open APIs. This helps accelerate things that are coming into the systems infrastructure space.

Then strive for everything-as-a-service, so use cloud consumption models. You want analytics, as we said earlier, to help understand what's going on and where you can best distribute workloads -- from the cloud to the edge or on-premises, because it's a hybrid world and that’s what we really need.

And then make sure you can control your spending and utilization of those services, because otherwise they will get out of control and you won't save any money at all. Lastly, be ready to extend your control beyond the data center to the edge as things get more distributed. A lot of the computing will increasingly happen close to the edge.

Computing close to the edge

Abbott: Yes. That's has to be something you start working on now. If you have software-defined infrastructure, that's going to be easier to distribute than if you are still wedded to particular systems, as the old, traditional model was.

Gardner: We have talked about what companies should do. What about what they shouldn't do? Do you just turn off the spigot and say no more cloud services until you get control?

It seems to me that that would stifle innovation, and developers would be particularly angry or put off by that. Is there a way of finding a balance between creative innovation that uses cloud services, but within the confines of an economic and governance model that provides oversight, cost controls, and security and risk controls?

Abbott: The best way is to use some of these new tools as bridging tools. So, with hybrid management tools, you can keep your existing mission-critical applications running and make sure that they aren't disrupted. Then, gradually you can move over the bits that make sense onto the newer models of cloud and distributed edge.
Gain New Insights Into
Managing the Next Wave
Of IT Disruption
You don't do it in one big bang. You don’t lift-and-shift from one to another, or react, as some people have, to reverse back from cloud if it has not worked out. It's about keeping both worlds going in a controlled way. You must make sure you measure what you are doing, and you know what the consequences are, so it doesn't get out of control.


Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. Sponsor: Hewlett Packard Enterprise.


You may also be interested in:

Wednesday, January 23, 2019

A discussion with IT analyst Martin Hingley on the culmination of 30 years of IT management maturity

The next BriefingsDirect hybrid IT strategies interview explores how new maturity in the management and composition of multiple facets of IT -- from cloud to bare-metal, from serverless to legacy systems -- amount to a culmination of 30 years of IT evolution.

We’ll hear now from an IT industry analyst about why – for perhaps the first time -- we’re able to gain an uber-view over all of IT operations. And we’ll explore how increased automation over complexity such as hybrid and multicloud deployments sets the stage for artificial intelligence (AI) in IT operations, or AIOps.

It may mean finally mastering IT heterogeneity and giving businesses the means to truly manage how they govern and sustain all of their digital business assets.

Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. 

Here to help us define the new state of total IT management is Martin Hingley, President and Market Analyst at ITCandor Limited, based in Oxford, UK. The discussion is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.

Here are some excerpts:

Gardner: Looking back at IT operations, it seems that we have added a lot of disparate and hard-to-manage systems – separately and in combination -- over the past 30 years. Now, with infrastructure delivered as services and via hybrid deployment models, we might need to actually conquer the IT heterogeneity complexity beast – or at least master it, if not completely slay it.

Do you agree that we’re entering a new era in the evolution of IT operations and approaching the need to solve management comprehensively, over all of IT?

Hingley: I have been an IT industry analyst for 35 years, and it’s always been the same. Each generation of systems comes in and takes over from the last, which has always left operators with the problem of trying to manage the new with the old.

Hingley
A big shift was the client/server model in the late 1980s and early 1990s, with the influx of PC servers and the wonderful joy of having all these new systems. The problem was that you couldn’t manage them under the same regime. And we have seen a continuous development of that problem over time.

It’s also a different problem depending on the size of organization. Small- to medium-sized (SMB) companies can at least get by with bundled systems that work fine and use Microsoft operating systems. But the larger organizations generate a huge mixture of resources.

Cloud hasn’t helped. Cloud is very different from your internal IT stuff -- the way you program it, the way you develop applications. It has a wonderful cost proposition; at least initially. It has a scalability proposition. But now, of course, these companies have to deal with all of this [heterogeneity].

Now, it would be wonderful if we get to a place where we can look at all of these resources. A starting point is to think about things as a service catalog, at the center of your corporate apps. And people are beginning that as a theory, even if it doesn’t sit in everybody’s brain.

So, you start to be able to compose all of this stuff. I like what Hewlett Packard Enterprise (HPE) is doing [with composable infrastructure]. … We are now getting to the point where you can do it, if you are clever. Some people will, but it’s a difficult, complex subject.

Gardner: The idea of everything-as-a-service gives you the opportunity to bring in new tools. Because organizations are trying to transform themselves digitally -- and the cloud has forced them to think about operations and development in tandem -- they must identify the most efficient mix of cloud and on-premises deployments.

They also have to adjust to a lack of skills by automating and trying to boil out the complexity. So, as you say, it’s difficult.

But if 25 percent of companies master this, doesn’t that put them in a position of being dominant? Don’t they gain an advantage over the people who don’t?

Hingley: Yes, but my warning from history is this. With mainframes, we thought we had it all sorted out. We didn’t. We soon had client/server, and then mini-computers with those UNIX systems, all with their own virtualizations and all that wonderful stuff. You could isolate the data in one partition from application data from a different application. We had all of that, and then along comes the x86 server.
How to Remove Complexity
From Multi-cloud
And Hybrid IT

It’s an architectural issue rather than a technology issue. Now we have cloud, which is very different from the on-premises stuff. My warning is let’s not try and lock things down with technology. Let’s think about it as architecture. If we can do that, maybe we can accommodate neuromorphic and photonic and quantum computing within this regime in the future. Remember, the people who really thought they had it worked out in previous generations found out that they really hadn’t. Things moved on.

Gardner: And these technology and architectural transitions have occurred more frequently and accelerated in impact, right?

Beyond the cloud, IT is life

Hingley: I have been thinking about this quite a lot. It’s a weird thing to say, but I don’t think “cloud” is a good name anymore. I mean, if you are a software company, you’d be an idiot if you didn’t make your products available as a service.

Every company in the world uses the cloud at some level. Basically there is no longer choice about whether we use a cloud. All those companies that thought they didn’t, when people actually looked, found they were using the cloud a lot in different departments across the organization. So it’s a challenge, yet things constantly change.

If you look 20 years in the future, every single physical device we use will have some level of compute built into it. I don’t think people like you and I are going to be paid lots of money for talking about IT as if it were a separate issue.

It is the world economy, it just is; so, it becomes about how well you manage everything together.

If you look 20 years in the future, every single physical device we use will have some level of compute built into it.  ... It becomes the world economy. It becomes about how well you manage everything together.
As this evolves, there will be genuinely new things … to manage this. It is possible to manage your resources in a coherent way, and to sit over the top of the heterogeneous resources and to manage them.

Gardner: A tandem trend to composability is that more-and-more data becomes available. At the edge, smart homes, smart cities, and also smarter data centers. So, we’re talking about data from every device in the data center through the network to the end devices, and back again. We can even determine how the users consume the services better and better.

We have a plethora of IT ops data that we’re only starting to mine for improving how IT manages itself. And as we gain a better trail of all of that data, we can apply machine learning (ML) capabilities, to see the trends, optimize, and become more intelligent about automation. Perhaps we let the machines run the machines. At least that’s the vision.

Do you think that this data capability has pushed us to a new point of manageability? 

Data’s exploding, now what? 

Hingley: A jetliner flying across the Atlantic creates 5TB of data; each one. And how many fly across the Atlantic every day? Basically you need techniques to pick out the valuable bits of data, and you can’t do it with people. You have to use AI and ML.

The other side is, of course, that data can be dangerous. We see with the European Union (EU) passing the General Data Protection Regulation (GDPR), saying it’s a citizens’ right within the EU to have privacy protected and data associated with them protected. So, we have all sorts of interesting things going on.

The data is exploding. People aren’t filtering it properly. And then we have potential things like autonomous cars, which are going to create massive amounts of data. Think about the security implications, somebody hacking into your system while you are doing 70 miles an hour on a motorway.

I always use the parable of the seeds. Remember that some seeds fall on fallow ground, some fall in the middle of the field. For me, data is like that. You need to work out which bits of it you need to use, you need to filter it in order to get some reasonable stuff out of it, and then you need to make sure that whatever you are doing is legal. I mean, it’s got to be fun.
How to Remove Complexity
From Multi-cloud
And Hybrid IT
Gardner: If businesses are tasked with this massive and growing data management problem, it seems to me they ought to get their IT house in order. That means across a vast heterogeneity of systems, deployments, and data types. That should happen in order to master the data equation for your lines of business applications and services.

How important is it then for AIOps -- applying AI principles to the operations of your data centers – to emerge sooner rather than later?

You can handle the truth 

Hingley: You have to do it. If you look at GDPR or Sarbanes-Oxley before that, the challenge is that you need a single version of the truth. Lots of IT organizations don’t have a single version of the truth.

If they are subpoenaed to supply every email that it has the word “Monte Carlo” in it, they couldn’t do it. There are probably 25 copies of all the emails. There’s no way of organizing it. So data governance is hugely important, it’s not nice to have, it’s essential to have. Under new regulations coming, and it’s not just EU, GDPR is being adopted in lots of countries.


It’s essential to get your own house in order. And there’s so much data in your organization that you are going to have to use AI and ML to be able to manage it. And it has to go into IT Ops. I don’t think it’s a choice, I don’t think many people are there yet. I think it’s nonetheless a must do.

Gardner: We’ve heard recently from HPE about the concept of a Composable Cloud, and that includes elevating software-defined networking (SDN) to a manageability benefit. This helps create a common approach to the deployment of cloud, multi-cloud, and hybrid-cloud.

It’s essential that you get your house in order. And there's so much data in your organization that you are going to have to use AI and ML to be able to manage it. And it has to go into IT Ops.
Is this the right direction to go? Should companies be thinking about a common denominator to help sort through the complexity and build a single, comprehensive approach to management of this vast heterogeneity?

Hingley: I like what HPE is doing, in particular the mixing of the different resources. You also have the HPE GreenLake model underneath, so you can pay for only what you use. By the way, I have been an analyst for 35 years, if every time the industry started talking about the need to move from CAPEX to OPEX had actually shifted, we would have been at 200 percent OPEX by now.

In the bad times, we move toward OPEX. In the good times, we secretly creep back toward CAPEX because it has financial advantages. You have to be able to mix all of these together, as HPE is doing.

Moreover, in terms of the architecture, the network fabric approach, the software-defined approach, the API connections, these are essential to move forward. You have to get beyond point products. I hope that HPE -- and maybe couple of other vendors -- will propose something that’s very useful and that helps people sort this new world out.
How to Remove Complexity
From Multi-cloud
And Hybrid IT

Wednesday, January 9, 2019

How global HCM provider ADP mines an ocean of employee data for improved talent management

The next BriefingsDirect big data analytics and artificial intelligence (AI) strategies discussion explores how human capital management (HCM) services provider ADP unlocks new business insights from vast data resources.

With more than 40 million employee records to both protect and mine, ADP is in a unique position to leverage its business data network for unprecedented intelligence on employee trends, risks, and productivity. ADP is entering a bold new era in talent management by deploying advanced infrastructure to support data assimilation and refinement of a vast, secure data lake as foundations for machine learning (ML).

Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. 

Unpack how advances in infrastructure, data access, and AI combine to produce a step-change in human capital analytics with panelists Marc Rind, Vice President of Product Development and Chief Data Scientist at ADP Analytics and Big Data, and Dr. Eng Lim Goh, Vice President and Chief Technology Officer for High Performance Computing and Artificial Intelligence at Hewlett Packard Enterprise (HPE). The discussion is moderated by Dana Gardner, Principal Analyst at Interarbor Solutions.



Here are some excerpts:

Gardner: Marc, what's unique about this point in time that allows organizations such as ADP to begin to do entirely new and powerful things with its vast data?

Rind: What’s changed today is the capability to take data -- and not just data that you originally collect for a certain purpose, I am talking about the “data exhaust” -- and to start using that data for purposes that are not the original intention you had when you started collecting it.

Rind
We pay one in six full-time employees in the US, so you can imagine the data that we have around the country, and around the world of work. But it's not just data about how they get paid -- it's how they are structured, what kind of teams are they in, advances, bonuses, the types of hours that they work, and everything across the talent landscape. It's data that we have been able to collect, curate, normalize, and then aggregate and anonymize to start leveraging to build some truly fascinating insights that our clients are able to leverage.

Gardner: It's been astonishing to me that companies like yours are now saying they want all of the data they can get their hands on -- not just structured data, but all kinds of content, and bringing in third-party data. It's really “the more, the merrier” when it comes to the capability to gather entirely new insights.

The vision of data insight

Rind: Yes, absolutely. Also there have been advances in methodologies to handle this data -- like you said, unstructured data, non-normalized data, taking data from across hundreds of thousands of our clients, all having their own way that they define, categorize, and classify their workforces.

Learn How IT Best Supports 

The Greatest Data Challenges

Now we are able to make sense of all of that -- across the board -- by using various approaches to normalize, so that we can start building insights across the board. That’s something extremely exciting for us to be able to leverage.

Gardner: Dr. Goh, it's only been recently that we have been able to handle such vast amounts of data in a simplified way and at a manageable cost. What are partners like HPE bringing to the table to support these data platforms and approaches that enable organizations like ADP to make analytics actionable?

Goh: As Marc mentioned, these are massive amounts of data, not just the data you intend to keep, but also the data exhaust. He also mentioned the need to curate it. So the idea for us in terms of data strategy with our partners and customers is, one, to retain data as much as you can.

Goh
Secondly, we ensure that you have the tools to curate it, because there is no point having massive amounts of data over decades – and when you need them to train a machine –  and you don’t know where all of the data is. You need to curate it from the beginning, and if you have not, start curating your data now.

The third area is to federate. So retain, curate, and federate. Why is the third part, to federate, important? As many huge enterprises evolve and grow, a lot of the data starts to get siloed. Marc mentioned a data lake. This is one way to federate, whereby you can cut across the silos so that you can train the machine more intelligently.

We at HPE build the tools to provide for the retention, curation, and federation of all of that data.

Gardner: Is this something you are seeing in many different industries? Where are people leveraging ML, AI, and this new powerful infrastructure? 

Goh: It all begins with what I call the shift. The use of these technologies emerged when industries shifted from when prediction decisions were made using rules and scientific law-based models.

Then came a recent reemergence of ML, where instead of being based on laws and rules, you evolve your model more from historical data. So data becomes important here, because the intelligence of your model is dependent on the quantity and quality of the data you have. And by using this approach you are seeing many new use cases emerge, of using the ML approach on historical data.

One example would be farming. Instead of spraying the entire crop field, they just squirt specifically at the weeds and avoid the crops.

Gardner: This powerful ML example is specific to a vertical industry, but talent management insights can be used by almost any business. Marc, what has been the challenge to generate talent management insights based on historical data?

Rind: It’s fascinating because Dr. Goh’s example pertains to talent management, too. Everyone that we work with in the HCM space is looking to gain an advantage when it comes to finding, keeping, and retaining their best talent.

We look at a vast amount of employment data. From that, we can identify people who ended up leaving an organization voluntarily versus those who stayed and grew, why they were able to grow, based on new opportunities, promotions, different methods of work, and by being on different teams. Similar to the agriculture example, we have been able to use the historical data to find patterns, and then identify those who are the “crops” and determine what to do to keep them happier for longer retention.

It’s fascinating because Dr. Goh’s example pertains to talent management, too. Everyone that we work with in the HCM space is looking to gain an advantage when it comes to finding, keeping, and retaining their best talent.
This is a big shift in the talent management space. We are leveraging vast data -- but not presenting too much data to an HCM professional. We spend a lot of time handling it on their behalf so the HCM professional and even managers can have the insights pushed to them, rather than be bombarded with too much data.

At the end of the day, we are using AI to say, “Hey, here are the people you should go speak with. Or this manager has a lot of high-risk employees. Or this is a critical job role that you might see higher than expected turnover with.” We can point the managers in that direction and allow them to figure out what to do about it. And that's a big shift in simplifying analysis, and at the same time keeping the data secure.

Data that directs, doesn’t distract 

Goh: What Marc described is very similar to what our customers are doing by converting their call center voice recordings into text. They then anonymize it but gain the ability to figure out the sentiment of their customers.

The sentiment analysis of the text -- after converting from a voice recording – helps them better understand churn. In the telco industry, for example, they are very concerned about churn, which means a customer leaving you for another vendor.

Yes, it’s very similar. First you go through a massive amount of historical data, and then use smart tools to convert the data to make it useable, and then a different set of tools analyzes it all -- to gain such insights as the sentiment of your customers.

Gardner: When I began recording use case discussions around big data, AI, and ML, I would talk to organizations like refineries or chemical plants. They were delighted if they could gain a half-percent or a full percent of improvement. That alone meant billions of dollars to them.

But you all are talking about the high-impact improvement for employees and talent. It seems to me that this isn’t just shaving off a rounding number of improvement. Marc, this type of analysis can make or break a company's future.

So let's look at the stakes here. When we talk about improving talent management, this isn’t trivial. This could mean major improvement for any cdanStaveMen66ompany.

Learn How IT Best Supports 

The Greatest Data Challenges

Rind: Every company. Any leader of an organization will tell you that their most important resource is the people that work for the company. And that value is not an easy thing to measure.

We are not talking about how much more we can save on our materials, or how to be smarter in electricity savings. You are talking about people. At the end of the day, they are not a resource as much as they are human beings. You want to figure out what makes them tick, gain insight into where people need to be growing, and where you should spend the human time with them.

Where the AI comes in is to provide that direction and offer suggestions and recommendations on how to keep those people there, happy and productive.

Another part of keeping people productive is in automating the processes necessary for managers. We still have a lot of users punching clocks, managing time, and approving pay cards and processing payroll. And there are a lot of manual things that go on and there is still a lot of paperwork

We are using AI to simplify and make recommendations to handle a lot of those pieces, so the HR professional can be focused on the human part -- to help grow careers rather than be stuck processing paperwork and running reports.

Cost-effective AI, ML has arrived 

Gardner: We’re now seeing AI and ML have a major impact on one of the most important resources and assets a company can have, human capital. At the same time, we’re seeing the cost and complexity of IT infrastructure that support AI go down thanks to things like hyperconverged infrastructure (HCI), lower cost of storage, capability to create whole data centers that can be mirrored, backed up, and protected -- as well as ongoing improvements in composable infrastructure.

Are we at the point where the benefits of ML and AI are going up while the cost and composability of the underlying infrastructure are going down?

Goh: Absolutely. That’s the reason we have a reemergence of AI through machine learning of historical data. These methods were already available decades ago, but the infrastructure was just too costly to amass enough data for the machine to be intelligent. You just couldn’t get enough compute power to go through that data for the machine to be intelligent. It wasn’t until now that the various infrastructure required came down in cost, and therefore you see this reemergence of ML.

If one were to ask why in the last few years there has been a surge to AI, it would be lower cost of compute capability. We have reached a certain point where it is cost-effective enough to amass the data. Also because of the Internet, the data has become more easily accessible in the last few years.

Gardner: Marc, please tell us about ADP. People might be familiar with your brand through payroll processing, but there's a lot more to it.

Find, manage, and keep talent 

Rind: At ADP, or Automatic Data Processing, data is our middle name. We’ve been working at a global scale for 70 years, now with $12 billion in revenue and supporting over 600,000 businesses -- ranging from multinational corporations to three-person small businesses. We process $2 trillion in payroll and taxes, running about 40 million employee records per month. The amount of data we have been collecting is across the board, not just payroll.

Talent management is a huge thing now in the world of work -- to find and keep the best resources. Moving forward, there is a need to understand innovative engagement of that workforce, to understand the new world of pay and micro-pay, and new models where people are paid almost immediately.

The contingent workforce means a work market where people are moving away from traditional jobs. So there are lots of different areas within the world of payroll processing and talent management. It has really gotten exciting.

This could mean major improvement for any company. Where the artificial intelligence comes in is to provide that direction and offer suggestions and recommendations on how to keep those people there happy and productive.
With all of this -- optimizing your workforce – also brings better understanding of where to save the organization from lost dollars. Because of the amounts of data, we can inform a client not just on, “Okay, this is what your cost of turnover is based on who is leaving and how long it takes them to get productive again, and the cost of recruiting.”

We can also show how your HCM compares against others in your field. It's one thing to share some information. It’s another to give an insight on how others have figured this out or are handling this better. You gain the potential to save more by learning about other methods out there that you should explore to improve talent retention.

Once you begin generating cost savings for an organization -- be it in identifying people who are leaving, getting them on-boarded better, or reducing cost from overtime – it shows the power of the insights and of having that kind of data. And that’s not just about your own organization, but it’s in how you compare to your peers.

So that’s very exciting for us.

All-access data analytics

Goh: Yes, we are very keen to get such reports on intelligence with regards to our talent. It’s become very difficult to hire and retain data scientists focused on ML and AI. These reports can be helpful in hiring and to understand if they are satisfied in their jobs.

Rind: That’s where we see the future of work, and the future of pay, going. We have the organization, the clients, and the managers -- but at the end, it’s also data insights for the employees. We are in a new world of transparency around data. People understand more, they are more accepting of information as long as they are not bombarded with it.

As an employee, your partner in your career growth and your happiness at work is your employer. That’s the best partnership, where the employer understands how to put you into the right place to be more productive and knows what makes you tick. There should be understanding of the employees’ strengths, to make sure they use those strengths every day, and anticipate what makes them happier and more productive employees.

Those conversations start to happen because of the data transparency. It’s really very exciting. We think this data is going to help guide the employees, managers, and human resources (HR) professionals across the organizations.

Learn How IT Best Supports 

The Greatest Data Challenges

Gardner: ADP is now in a position where your value-added analysis services are at the level where boards of directors and C-suite executives will be getting the insights. Did that require a rethinking of ADP’s role and philosophy?

Rind: Through our journey we discovered that providing insights to the HR professional is one thing. But we realized that to fully unleash and unlock the value in the data, we needed to get it into the hands of the managers and executives in the C-suite.

And the best way to do that was to build ADP’s mobile app. It’s been in the top three of the most downloaded applications from the business section on the iTunes Store. People initially got this application to check their paystub and manage their deductions, et cetera. But now, that application is starting to push up to the managers, to the executives, insights about their organization and what's going on.

A key part was to understand the management persona. They are busy running their organizations, and they don’t have the time to pore through the data like a data scientist might to find the insights.

So we built our engine to find and highlight the most important critical data points based on their statistical significance. Do you have an outlier? Are you in the bottom 10 percent as an organization in such areas as new hire attrition? Finding those insights and pushing them to the manager and executive gets them these headlines.

Next, as they interact with the application, we gain intelligence about what's important to that manager and executive. We can then then push out the insights related to what's most important to them. And that's where we see these value-added services going. An executive is going to care about some things differently than a supervisor or a line manager might.

We can generate the insights based on their own data when they need it through the application versus them having to go in and get it. I think that push model is a big win for us, and we are seeing a lot of excitement from our clients as they are start using the app.

Gardner: Dr. Goh, are you seeing other companies extend their business models and rethinking who and what they are due to these new analytics opportunities?

Data makes all the difference

Goh: Yes, yes, absolutely. The industry has shifted from one where your differentiated asset was your method and filed patent, to one where your differentiated asset is the data. Data becomes your defensible asset, because from that data you can build intelligent systems to make better decisions and better predictions. So you see that trend.

In order for this trend to continue, the infrastructure must be there to continually reduce cost, so you can handle the growing amounts of data and not have the cost become unmanageable. This is why HPE has gone with the edge-to-cloud hybrid approach, where the customer can implement this amassing of data in a curated and federated way. They can handle it in the most cost-effective way, depending on their operating or capital budgets.

Gardner: Marc, you have elevated your brand and value through trends analysis around pay equity or turnover trends, and gaining more executive insights around talent management. But that wouldn't have been possible unless you were able to gain the right technology.

What do you have under the hood? And what choices have you made to support this at the best cost?

Rind: We build everything in our own development shop. We collect all the data on our Cloudera [big data lake] platform. We use various frameworks to build the insights and then push those applications out through our ADP Data Cloud.

We have everything open via a RESTful API, so those insights can permeate throughout the entire ADP ecosystem -- everyone from a practitioner getting insights as they on-board a new employee and on out to the recruiting process. So having that open API is a critical part of all of this.

Gardner: Dr. Goh, one of the things I have seen in the market is that the investments that companies like ADP make in the infrastructure to support big data analytics and AI sets in motion a virtuous adoption benefit. The investments to process the data leads to an improvement in analytics, which then brings in more interest in consumption of those analytics, which leads to the need for more data and more analytics.

It seems to me like it’s a gift that keeps giving and it grows in value over time.

Steps in the data journey 

Goh: We group our customers on this AI journey into three different groups: Early, started, and advanced. About 70 percent of our customers are in the early phase, about 20 percent in the started phase, where they have already started on the project, and about 10 percent are in the advanced phase.

The advanced-phase customers are like the automotive customers who are already on autonomous vehicles but would like us to come in and help them with infrastructure to deal with the massive amounts of data.

But the majority of our customers are in the early phase. When we engage with them, the immediate discussion is about how to get started. For example, “Let’s pick a low-hanging fruit that has an outcome that’s measurable; that would be interesting.”

We work with the customer to decide on an outcome to aim for, for the ML project. Then we talk about gaining access to the data. Do they have sufficient data? If so, does it take a long time to clean it out and normalize it, so you can consume it?

After that phase, we start a proof of concept (POC) for that low-hanging fruit outcome -- and hopefully it turns out well. From there the early customer can approach their management for solid funding to get them started on an operational project.

We are using AI to simplify and make recommendations to handle a lot of those pieces, so the HR professional can be focused on the human part -- to help grow careers rather than be stuck processing paperwork and running reports.
That’s typically how we do it. It always starts with the outcome, and what we are aiming for this machine to be trained at. Once they have gone through the learning phase, what is it they are trying to achieve, and would that achievement be meaningful for the company? A low-hanging fruit POC doesn’t have to be that complex.

Gardner: Marc, any words of wisdom looking back with 20/20 hindsight? When it comes to the investments around big data lakes, AI, and analytics, what would you tell those just getting started?

Rind: Much to Dr. Goh’s point, picking a manageable project is a very important idea. Go for something that is tangible, and that you have the data for. It's always important to get a win instead of boiling the ocean, to prove value upfront.

A lot of large organizations -- instead of building data lakes, they end up with a bunch of data puddles. Large companies can suffer from different groups building their own.

We have committed to localizing all of the data into a single data lake. The reason is that you can quickly connect data that you would never have thought to connect before. So understanding what the sales and the service process is, and how that might impact or inform the product or vice versa, is only possible if you start putting all of your data together. Once you get it together, just work on connecting it up. That's key to opening up the value across your organization. 

Connecting the data dots 

Goh: It helps you connect more dots.

Gardner: The common denominator here is that there is going to be more and more data. We’re starting to see the Internet of Things (IoT) and the Industrial Internet of Things (IIoT) bring in even more data.

Even relevant to talent management, there are more ways of gathering even more data about what people are doing, how they are working, what their efficacy is in the field, especially across organizational boundaries like contingent workforces, being able to measure what they are doing and then pay them accordingly.

Marc, do you see ever more data coming online to then need to be measured about how people work?

Rind: Absolutely! There is no way around it. There are still a lot of disconnected points of data, for sure. The connection points are going to just continue to be made possible, so you get a 360-degree view of the world at work. From that you can understand better how they are working, how to make them more productive and engaged, and bringing flexibility to allow them to work the way they want. But only by connecting up data across the board and pulling it all together would that be possible.

Learn How IT Best Supports 

The Greatest Data Challenges

Gardner: We haven’t even scratched the surface of incentivization trends. The more data allows you to incentivize people on a micro basis in near-real time, is such an interesting new chapter. We will have to wait for another day, another podcast, to get into all of that.

Listen to the podcast. Find it on iTunes. Read a full transcript or download a copy. Sponsor: Hewlett Packard Enterprise.

You may also be interested in: