The ninth decade of McKinsey’s journey, from 2006 to 2016, is when computing became untethered. People stopped going to machines to access digital services and instead started carrying them; over a billion people adopted smartphones. Companies stopped owning the hardware these digital services ran on; cloud computing became the norm. And in university labs, machine learning started to work, creating the foundations for today’s AI age. Computing also came untethered from companies’ IT road maps, advancing faster than most enterprises could keep pace.
The backdrop was the worst financial crisis since the 1930s. The Great Recession, starting in December 2007 and officially lasting 18 months, hit many industries. Tighter technology budgets helped fuel demand for cloud computing and software as a service (SaaS), billed as cheaper and more flexible ways to use software and compute infrastructure. At the same time, the supply of capital for cloud start-ups surged. Just some of the cloud and software companies founded during the decade include Cloudflare, Docker, Dropbox, GitHub, HubSpot, MongoDB, Okta, Slack, Snowflake, Zendesk, and Zoom.
For chief information officers (CIOs), the decade’s SaaS boom has parallels to today. While initially cloud delivery was seen as cheaper than packaged software and servers, costs often rose at renewal. Now, CIOs are increasingly concerned about the rising costs of AI usage, with one in five companies curtailing their AI rollouts due to costs.1
The smartphone goes to work
The launch of the iPhone in June 2007 changed personal and enterprise computing forever. BlackBerry had been selling email devices since the late 1990s, but the iPhone, paired with Apple’s App Store, which opened in the summer of 2008, turned the smartphone into a platform. Google answered in October 2008 with the first Android handset and an open-source operating system that any manufacturer could use. Annual smartphone sales went from 122 million units in 2007 to nearly 1.5 billion in 2016.2
The smartphone wasn’t just a consumer phenomenon; it also changed how technology departments functioned. Employees began using their smartphones for work, whether IT wanted them to or not. Bring-your-own-device (BYOD) policies became critical to secure enterprise data. And the app-based interface of the smartphone also changed what employees expected from enterprise software, ushering in the era of easy-to-use enterprise platforms like Slack and Zoom.
Shadow AI—when employees use AI tools without their employer’s knowledge or oversight—is a BYOD problem with higher stakes. CIOs are struggling with how to encourage AI experimentation while also erecting guardrails for safe and responsible AI use.
The bumpy road to digital transformation
Amazon Web Services (AWS) began selling pay-as-you-go storage and computing capacity in 2006. Microsoft’s cloud platform, Azure, became commercially available in early 2010. For technology leaders, the emergence of cloud computing was seismic. Running a data center had long been part of the enterprise IT playbook, but now CIOs had to contend with an entirely new infrastructure model in which their data and applications lived on servers outside the company’s walls. The cloud wave also reshaped adjacent markets in security, identity, data streaming, and networking.
Though cloud migrations were often slow and complex, the investment paid off for companies that successfully completed these large-scale transformations. Netflix is a prime example of a company that leveraged a migration to create measurable value. The company began moving its IT systems and streaming infrastructure to AWS in 2008 and didn’t finish until early 2016. But a cloud-based architecture allowed Netflix to scale internationally and add tens of millions more subscribers.
Cloud migrations proved complex for reasons that had little to do with the technology—which was ready before companies were. Application dependencies were undocumented, data could not move without breaking downstream reports, and teams had skills geared toward an on-premises operating model. The same obstacles are appearing now as companies rebuild their infrastructures for and with AI. There’s no easy path to integrating agents at scale and then empowering teams to use them effectively. Entirely new technology management playbooks will be required to create value with agentic AI.
Machine learning starts to work
One of the decade’s most consequential developments was almost invisible to businesses: the rise of machine learning (ML). In 2006, computer scientist Geoffrey Hinton published a way to train multilayered neural networks, an approach now known as deep learning. Then, two more things had to arrive before ML could be used at scale. The first was data, in the form of ImageNet, released in 2009, which gave researchers millions of labeled images and later a competition to measure training progress against. The second was cheap hardware; graphics processors built for video gaming turned out to suit the mathematics of neural networks. The public saw the shift through a series of milestones. IBM’s Watson supercomputer, which used ML to weigh competing answers, beat two Jeopardy champions in early 2011. In March 2016, DeepMind, an AI lab that Google acquired in 2014, saw its AlphaGo program beat Lee Sedol, a world-champion player of the board game Go.
Inside companies, the progress toward AI was less visible. Banks began scoring credit applications with ML models and payment processors moved fraud detection from written rules to learned patterns. In most cases, the constraint was not the strength of the ML algorithm, but how to get data into the model. Legacy data sat in systems built to process transactions rather than analyze them. Thus, much of the decade’s real AI investment went into rebuilding data infrastructure. Snowflake was founded in 2012 and Databricks in 2013; both are now among the key platforms that enterprises use to run AI.
Inheriting legacy technology is an age-old problem for CIOs. The infrastructure, applications, and systems that AI needs to work within enterprises were already being assembled between 2006 and 2016. But now models have improved far faster than the data estates they depend on, which is why true AI transformation requires building new data foundations to enable agentic AI at scale.
What leaders can learn
The decade from 2006 to 2016 handed enterprises the raw materials of modern computing. Smartphones put technology in employees’ hands, cloud computing freed IT teams from running hardware and buying costly software packages, and ML laid the foundations for today’s AI age. None of these technological advances created value on their own. The companies that benefited were the ones that rewired their technology infrastructures to entirely change their ways of working. That transformation was costly and difficult, but worth it in the long run.
Three patterns from those years are repeating with AI. First, adoption is starting with employees rather than with IT, and the question for CIOs now is how to channel enthusiasm safely. Second, AI costs are growing quickly, requiring CIOs to understand the new economics of AI and how to create value at scale. And third, the roadblocks to AI scaling are not about technology itself, with models rapidly improving their performance. The enterprise snags come from slow data migrations, legacy tech stacks, and missing skills.
In the decade from 2006 to 2016, ML showed its potential value but was not yet embedded in how companies operated. The parallels to AI are clear. Today, 89 percent of companies use AI, but only 37 percent attribute at least some EBIT impact to it, essentially unchanged from 2025—showing that extracting value from a frontier technology is not easy.3 Success tomorrow will come from doing the unglamorous work of rewiring operating models and getting data ready today.
Chandrasekhar Panda is a partner in McKinsey’s Riyadh office, Henning Soller is an alumnus of the Frankfurt office, Klemens Hjartar is a senior partner in the Copenhagen office, and Sven Blumberg is a senior partner in the Düsseldorf office.
1 Dan Tinkoff, Lieven Van der Veken, Michael Chui, and Tara Balakrishnan, “The state of AI in 2026: On the road to ROI,” McKinsey, August 25, 2026.
2 “Gartner says worldwide sales of smartphones grew 7 percent in the fourth quarter of 2016,” Gartner press release, February 15, 2017.
3 Dan Tinkoff, Lieven Van der Veken, Michael Chui, and Tara Balakrishnan, “The state of AI in 2026: On the road to ROI,” McKinsey, August 25, 2026.