In 1999, Hewlett-Packard spun off its testing and measurement business to form Agilent Technologies, which at the time was the largest initial public offering in Silicon Valley history. \u00a0\nFor ten years, Agilent had two primary businesses: one focused on life sciences technologies and chemical analysis solutions, and the other on electronic instruments. In 2014, with a high-growth life sciences business on its hands, the executive team at Agilent spun off the electronics business to form Keysight Technologies. Dan Krantz, who led the Agilent-Keysight IT separation, became CIO of Keysight in 2017.\n[ Learn the 10 old-school IT principles that still rule and the 12 'best practices' IT should avoid at all costs. | Get the latest insights by signing up for our CIO newsletter. ]\nKeysight pursued a three-pronged strategy to shift the focus from product sales to industry solution sales, increase R&D spend, and acquire new companies to grow the company from $2.8 billion to $4.6 billion in annual revenue.\nThis new high-growth business required a new set of business capabilities, including digital engagement technologies, a more dynamic supply chain, and the ability to support a product line that was shifting from warranty-based hardware to subscription-based software. \u00a0\nKrantz knew that stabilizing the new IT environment would just be table stakes. To provide a new set of business capabilities and support the tremendous growth ahead, he and his team would have to transform IT. \u00a0\nCloud-first, composable architecture\n\u201cYou might think that a high-tech company would run on modern IT,\u201d says Krantz, \u201cBut technology companies need to modernize IT just like everyone else. We may be offering the leading quantum computing capabilities in the market, but the underlying IT infrastructure can still be 15 years old.\u201d\nKrantz\u2019s target architecture would require moving from traditional onsite monolithic systems to a cloud-first composable architecture bucketed into three areas: systems of engagement, which included both sales transactions and customer support; systems of record; and systems of insight.\u00a0\nFor systems of engagement, \u201cour preference is to subscribe to a configurable cloud architecture, but if there is nothing that meets our needs, we'll develop our own solution and build the integrations between them,\u201d says Krantz.\nPricing, for example, required a custom solution. Keysight has more than 8,000 product families, each with up to 10,000 configurations, which requires sophisticated pricing algorithms. \u201cWe wrote the pricing engine with AWS microservices, so whether its presales, sales, post-sales, or even internal operations, every platform makes API calls to our custom-developed pricing engine,\u201d Krantz says.\nFor systems of record, the team combines its traditional ERP with new cloud-enabled solutions to support new business capabilities. For example, when the sales process shifted from one-time sales to subscriptions, Keysight needed a new way to recognize revenue.\u00a0 \u201cThe financial flows and manufacturing data from our 78 legal entities around the world run through one on-premises ERP,\u201d says Krantz. \u201cSo, rather than try to modernize our ERP, we bolted a cloud-based revenue recognition system onto our legacy ERP. That is the benefit a composable architecture.\u201d\nHow to get the data out\nKrantz and his team are now starting to focus on the third layer of the architecture, systems of insight. \u201cOnce we\u2019ve modernized our systems of engagement and systems of record, how do we get data out of them? That\u2019s where we are now,\u201d says Krantz. \u201cWe\u2019ve been replacing the reporting tools that were designed for a quarterly calendar and tied into our old monolithic systems, and moving them to a cloud platform.\u201d For this, Krantz is working on the visualization layer and partnering with a start-up analytics company because, \u201cWe like to work with smaller, hungry players who are eager to listen to a large complex customer and quickly iterate on new features.\u201d\nOne key step in improving analytics has been shifting from ETL (extract, transform and load), where you extract systems data, transform it, and then load it into your data warehouse, to ELT, where you extract the data out of your system of record, load it into a modern in-memory analytics system, and then transform it. \u201cWith traditional ETL, the transformation step really slowed us down,\u201d says Krantz. \u201cShifting the process enables us to move faster. Don\u2019t transform the data before you load it.\u201d\nBut architecture is just one piece of Keysight Technologies\u2019 IT transformation. \u201cWhen you have large monolithic systems, you spend six months on requirements, then a year building, another six months stabilizing, and then two years later, you start over,\u201d says Krantz. \u201cWith composable architecture, you can work independently, iteratively, and quickly on all these little building blocks.\u201d\nAdopting a growth mindset\nIterating on building blocks requires a new way of working, so Krantz and his team are on a journey to be Agile at scale, with nimble, cross-functional teams, and product owners. This is a major shift from the heavily outsourced IT model that Krantz inherited.\u00a0 \u201cWhen I joined Keysight, only 15 percent of the team was internal,\u201d says Krantz. \u201cOur team managed the IT suppliers who did the work, whether it was day-to-day support or project delivery. IT played the role of a broker.\u201d\nTo evangelize the idea of growing internal talent, Krantz introduced his leaders to the neuroscience concept of \u201cgrowth mindset,\u201d which emphasizes brain elasticity and learning potential. But the shift didn't happen overnight, and Krantz took the same iterative approach to talent development that he took to IT modernization. \u201cOur insourcing sequencing largely followed our architecture,\u201d says Krantz. \u201cAs we modernized our systems of engagement, we insourced that talent. We did not want to stand up a new digital product with outsourcers. When we transitioned from our legacy MPLS (Multiprotocol Label Switching) to software defined networking, we decided to run our network with a much higher mix of internal network engineering talent.\u201d Today, Keysight IT is 55 percent internal.\nLessons learned\nWhile Krantz acknowledges that the journey to an Agile model and cloud-composable architecture is never over, he can share some lessons learned:\n\nKrantz advises that CIOs modernizing their architecture \u201cfigure out a way for your new systems to coexist with your current systems,\u201d he says. \u201cA plan that includes knife-edge cutovers from old to new is a recipe for perpetual project delays. You will never get your new system to work exactly like your old one, so your business partners will never be ready for go-live.\u201d \nCalling new solutions \u201cdigital products\u201d rather than projects, was key to providing maximum value. \u201cWhen you sell a product in the marketplace, you don't say, \u2018I'm done. I will never add another feature or enhancement,\u2019\u201d Krantz says. \u201cCustomers will want enhancements indefinitely, which is why we called our new revenue recognition solution a \u2018revenue digital product.\u2019\u201d\nAutomate testing as early in your transformation as possible. \u201cIn our traditional model, 60 percent of our time was spent on testing and 40 percent on actual development. That doesn\u2019t scale,\u201d he says. \u201cFor us, the big \u2018A-ha!\u2019 came when we automated our test cycle so that every piece of code we push triggers a rerun of automated testing scripts, which ideally, should always be running. Testing cannot be manual or platform-specific, since in a composable architecture, a business process can cut across multiple systems.\u201d\n\nIT strategic planning comes next\nWith his target architecture in sight, and his team largely insourced, Krantz is now setting his sights on improving the strategic planning process for IT. \u201cIn our traditional approach, we would ask our mid-level business partners what they needed from IT,\u201d he says. \u201cThe demand would outweigh our resources, so we would ask the executive team to prioritize. We wasted a lot of time.\u201d\nRecognizing that the process was inefficient, Krantz and the rest of the executive team have made a change. Before Keysight\u2019s P&L leaders commit a strategic plan to the board, they meet with Krantz and his team to discuss the IT resources their plans will require, and those resources go into the IT strategic plan. \u201cShifting our bottom-up planning process to a top-down approach is something I wish I had started earlier,\u201d Krantz says.\nFor most CIOs, the \u201cwhat\u201d of a modern IT organization is pretty clear:\u00a0 a cloud-enabled, microservices-based architecture with platforms, APIs, and agile development that allow for fast, iterative releases. By focusing as much on the \u201chow\u201d\u2014iteration, insourcing, and strategic planning\u2014Krantz and his team are well on their way to modernizing technology in an 80-year-old high-tech company.