Q&A: Microsoft CIO Tony Scott
Exec oversees "dog-fooding" of Redmond's new technology and migration to the cloud.
As Microsoft Chief Information Officer, Tony Scott is in charge of what Redmond calls the "dog-fooding" of its newest technologies. The latest dog food Microsoft is digesting these days is Windows Azure platform, the company's portfolio of cloud services. Scott joined Microsoft more than two years ago after stints as CIO of Walt Disney and Co. and CTO of General Motors. In an interview, he talked about the dog-fooding of Windows Azure, the recent Microsoft decision to consolidate its IT outsourcing with Infosys Technologies Ltd., and the challenges of running the Redmond IT organization.
Q What are some of the key initiatives on your plate these days?
A It's a very interesting time at Microsoft right now as we shift from our more traditional business model to the cloud. That has all kinds of implications, not just for customers -- who will end up buying different things from us -- but it also affects our internal systems. A lot of our internal systems were designed for the world of packaged software and traditional licenses. Now, with the cloud, we have an opportunity to sell different things -- services in particular. It's a different license model, different revenue model, a different set of things that we sell, so I have to change our internal systems to support that. But at the same time we're also putting those new systems on the cloud -- putting it on what we're going to sell.
Q From an internal perspective, how would you characterize where you are in that transition?
A I think we've done what Microsoft always has done traditionally, which is we try to dog-food our own stuff and get the bugs out and make sure the functionality is there. So we actually began porting some of our applications to the Azure platform last year, and anything new that we're starting by default goes on Azure. We have, as most companies have, a portfolio of applications that over time are existing applications. We'll move them or migrate them as the opportunity arises and as the business case makes sense.
Q What percentage of your apps are on Windows Azure at this point, and what percentage do you see being eliminated?
A Not all apps are equal. Some are really big. But to give an example of our commitment, when I took the job, I did what a lot of incoming CIOs do: talk to the CEO and ask, "What's the most important system in Microsoft; what do you care about?" And Steve [Ballmer] said licensing is important to us. It's the engine that helps us realize all of our revenue.
When I went to the technology guys inside Microsoft and asked them to rank the relative health and modern status and cost of all the systems we had, licensing was one of those things that was at the bottom of the list. And so one of the first things I started on was the re-platforming and redefining of that licensing platform. That is probably the single biggest project we're doing that will go to the cloud. It's a two- to three-year project.
Q What's the biggest difficulty in getting from here to there?
A It's not a technical challenge as much as it is a business challenge. You've got to remember these are a set of systems that started when the company was small and had a simple product -- DOS. We had a pretty simple challenge of distribution and we didn't do business in every country around the world. Every time we expanded and did business in another country or added a product, we added some complexity to the licensing. So this is something that has built up over 30 years incrementally, a building block at a time. When you do that you don't always end up with something that's the best design or the best architecture. It's time for us to refresh and create a new licensing platform for the next 30 years. We're going to put it on the cloud.
Q When you put it in Windows Azure, are you putting it in the public Windows Azure, in datacenters? Are you going to use the [newly launched Windows Azure Platform] Appliance?
A As I mentioned, we dog-food everything, so you should think of Microsoft IT as a giant lab where we do things and test them out before we go to the public. From that sense, it's separate. And mail is a great example of that. Today I run Microsoft on the same mail service that we sell to customers. But I'm in a walled-off version of that where we're doing all the beta things -- making sure they're going to work right, that it's run by the same people. So we'll do the same thing with Azure. We'll have a local instance that will be run by the same people, [the] same technology. But we'll always be doing some things in there to test out and make sure what we're ultimately offering the public is the highest quality we can. An example of that today is we run SAP, but we run it on SQL Server. The way Microsoft and SAP certify SQL Server whenever we come up with a new release or patches, we put it through our whole development cycle with SAP and ultimately run the beta version -- SQL Server in this case -- in production, and we do it for some period of time. That tells us we have a single instance; we run Microsoft globally on that. It's as high volume as you can get for SAP. That's our typical normal pattern. We do it with Windows 7; we do it with Internet Explorer; we do it with Office; we do it with all of the products that are destined for the enterprise.
Q For a long time, you put out technical documentation as the first Microsoft test customer. Now a lot of the issues IT managers are facing are really a lot more business-oriented. Have you thought about doing business guidance?
A With respect to the cloud, we're working very closely with the product groups to say, "Here's how we thought about our application portfolio; here are the conversations we're having around low, medium or high business impact; here's the framework in which we're addressing that; and here's how you have the conversation with the business partners." We're also working closely with the product groups on defining what the product is in the first place, saying in essence to them, "Here are the things that CIOs are going to care about; here's the conversations they're going to need to have with their business partners; here's what our response can be or should be in those particular cases." So it's been a fun journey from not just testing the products, which has sort of been our traditional role, but helping design the products and being the first and best customer and proving that out.
Q What do you see, as the CIO, when you move things to the cloud? Will you need as many technical people? Will they be doing different kinds of things?
A I think, at the end of the day, it's still a little too early to tell, but there are some things we see that are different. We're also looking closely at our methodology and saying, "How does that change vis-ˆ-vis doing development in the cloud? What things changed there; what things changed in our financial management of applications; how do we have to think about that?" So, as an example, most companies capitalize a lot of their early investment and not the full development investment expense when they're developing a new application. A lot of that is hardware, but they also tend to capitalize on the initial investment in development. Well, now, when your development is a service that you're paying as an expense, and your test environment and even your production environment is an expense, that changes the financial profile of a project. It's still the same project, but it might look different in terms of how you budget for it, how you plan for it and so on. I don't see the net resources necessarily changing; I think that it will allow us to do more work within the same budget profile of what we've had before. I've talked to a lot of CIOs over the last six months and one of the common themes I hear from them is, "Right now, we've got the business backlog of work that our business wants us to do that we just can't address with today's resource pool; it's the biggest backlog we've ever had." I and others are looking to the cloud to say, "How can we do more? How can we attack more of that backlog than we would using conventional resource planning?" So, for us, some of that efficiency has come from provisioning. We're doing one more test pass than we did before because we've got the time to be able to do it.
We're also seeing another interesting phenomenon, which is we measure sev-1 [severity one] defects where there's a business impact where a system goes down or there's some revenue impact or something. We won't go live with a known sev-1 defect, but sometimes these things crop up in production after you go live. So, we measure those as a quality measure. What we've seen with the cloud -- and it was a trend we saw starting with virtualization -- is a 50 percent reduction in our sev-1 defects in production. As the CIO, I've worked hard to get 10 percent reduction before, through software engineering practices and those kinds of things. But we're seeing these kinds of reductions as a result of the standardization and commoditization of that infrastructure layer -- the operating system, the storage technology and so on. Because it's the same now for every application you go deploy -- it's not up to the whims of this developer or that infrastructure guy to go be creative in terms of something he did. And that translates to higher quality, so for us that's real. Whenever you have a sev-1 defect in production, you're now pulling your most precious developers and resources in operations and you are diverting them from what they're normally doing to get this thing fixed, and that interrupt-driven world is not conducive to smooth project execution. For us, that's a big benefit.
Q What led to your decision to consolidate your IT outsourcing arrangements with Infosys?
A We had traditionally a decentralized IT organization, and that in turn led to a very decentralized sourcing model with respect to our suppliers. Over the last several years we'd been centralizing and consolidating. There was also a tendency for someone in a sales region to influence the selection of a supplier by saying, "We're doing a deal with XYZ company in the marketplace, and it would be good if we could give a little business to them from Microsoft." That was the historical way it was done. As we centralized, we saw an opportunity to not only consolidate all of that into a smaller number of vendors but [to] change the economics for both us and our suppliers. They came to us saying, "We're spending a lot of money pursuing all these little deals, and the cost of pursuit is getting higher and higher, and we don't want to do that anymore." At the end, we consolidated all of it. We can take some of the savings ourselves, some of it falls to our suppliers, and they in turn can make an investment in Microsoft, learning our process and getting familiar with IT.
Q Why Infosys?
A We did a competitive bakeoff with all the key suppliers that we deal with, and they had the winning combination of cost and services and approach that resonated well with our valuation team.
Q Do you hear a lot of criticism about the fact that this can be seen as the offshoring of U.S. jobs?
A No, because we actually didn't move any more jobs offshore or onshore than we had before. It really was a consolidation. At the end of the day it's a much more efficient model for us, and, frankly, the things we're using Infosys for are the things every other IT organization on the planet uses outsourcers for. It's help desk, desk-side service and support, and those kinds of things.
Q What is the Unisys role in this contract?
A There was a certain set of services that Infosys said they preferred to partner with Unisys on. For example, we have these physical help desk facilities -- we call them Tech Links. It's a desk; it has its own branding, and Microsoft employees can walk up to the desk and hand their PC to a Tech Link professional who will then install software for them or configure something they might not have the skills to do themselves. They can diagnose hardware problems. [For] that service, Infosys said they wanted to partner with Unisys. For us, we get the best of both worlds. We get the combined talents of both companies.
Q Do you plan to continue running your systems on SAP, or do you see migrating to the Microsoft ERP and financial software?
A What we have is a two-tier ERP model that's part of our ERP strategy. So we use SAP at the core; it's the financial system of record for the company. But we have a lot of different businesses around Microsoft, so some of the local ERP processing we'll do with the Dynamics product set, and we're using it widely for CRM and for other kinds of functions that Dynamics products can support. Whether it's regionally or for a specific business, our pattern is to use the Dynamics for that at the local level and have SAP be the consolidation level.
Q Is it feasible to move off of SAP at some point?
A I think it could be. It's not in our current road map. We consider SAP to be a great partner; they sell a lot of SQL licenses for us every year, especially when you consider the cost alternatives.
Q How do you see moving the SAP systems to the cloud?
A It's a conversation we're having with SAP, and I can't predict where it's going to go. One of the things we've done through our Duet work with SAP, which we recently showcased at their Sapphire show, uses SharePoint and familiar Microsoft interfaces to front-end SAP. Every Microsoft employee accesses SAP today; they just don't know it. They see a familiar Microsoft interface when they're doing expense reports or other kinds of activities, but for them it's just a natural part of their desktops. Unlike a lot of other places, we don't have complaints about the SAP GUI. For them, it's just the familiar thing. We think a lot of our other customers want to do the same kind of thing. The interest in this using SharePoint and Microsoft as a front-end to SAP is something that's getting a lot of traction and a lot of customer interest.