\u2022 The software layer is under pressure. As AI makes it cheaper to build tools, the competitive advantage shifts from the software you buy to the data you already own.<\/p>\n
\u2022 Companies used to pay others to manage their data\u2014that\u2019s flipping. In 1985, firms spent twice as much building their own software as buying it. By 2024, they spent more than three times more on pre-packaged software. AI may reverse that trend.<\/p>\n
\u2022 HR tech company Phenom is betting on this shift. At its annual Philadelphia conference, the nearly 2,000-employee private company argued that \u201ctraditional SaaS is dying\u201d\u2014and that underutilized company data is the new asset.<\/p>\n
Twenty years ago, a British mathematician first popularized the metaphor that \u201cdata is the new oil.<\/a>\u201d<\/p>\n His point was that like oil, data is worthless in its natural state. Refinement is what adds value. This became possible only for firms well-resourced enough to execute on data, like big tech platforms, corporate retailers and financial services giants.<\/p>\n If the cost to produce software approaches zero, competitive advantage may come less from the tools companies buy and more from the data they already have.<\/p>\n Today\u2019s artificial intelligence wave, with large language models adept at serving up reasonable analysis from heaps of data, promises to reinforce the oil adage. The once-lucrative and in-demand enterprise software layer, in which tech companies rushed to service data needs, is under threat.<\/p>\n \u201cThe traditional definition of enterprise software is breaking apart,\u201d said Mahe Bayireddi, cofounder and CEO of Phenom<\/a>, an HR tech company that has grown to close to2,000 employees<\/a> and is inching toward a potential IPO.<\/p>\n At the company\u2019s annual conference inside Philadelphia\u2019s convention center in March, Bayireddi and his team laid out a vision that should unsettle any founder building enterprise software \u2014 and interest any organization sitting on underused data.\u00a0<\/p>\n As AI makes software cheaper and easier to deploy, the real value shifts to the data underneath it.<\/p>\n The implication matters well beyond HR tech. For years, enterprise software companies built lucrative businesses selling tools to manage data: applicant tracking systems, CRMs, analytics dashboards.\u00a0<\/p>\n In 1985, at the dawn of the computer age, half of early software spending was by companies for themselves, well more than double the amount spent on pre-packaged stuff, according to the Bureau of Economic Analysis. In 2024, after the rise of software-selling companies, these figures more than flipped. American companies spent close to four times as much on pre-packaged software than on making their own (48% versus 14%).<\/p>\n Now, large language models can sit on top of that data and generate insights more directly. That puts pressure on the software layer itself, and creates an opening for companies that own their data to do more with it themselves.<\/p>\n \u201cThe traditional SaaS is dying, and it will die,\u201d Bayireddi said. Phenom\u2019s pitch is that HR \u2014 often underfunded compared to sales, marketing and product \u2014 is sitting on a trove of underutilized data.<\/p>\n If that data can be structured and activated, the company argues, HR leaders can improve recruiting, retention and workforce planning without dramatically expanding headcount. The company\u2019s evolution reflects that bet.<\/p>\n The Bayireddi brothers started Phenom in 2011 in suburban Philadelphia as a more polished jobs platform. Today, it describes itself as \u201capplied AI for HR,\u201d aiming to become what executives at the conference repeatedly called a \u201cWorkOps\u201d platform \u2014 an operating system for how work gets done.\u00a0<\/p>\n \u201cGo-live is dead. It\u2019s just a milestone,\u201d said Hari Bayireddi, the company\u2019s president and Mahe\u2019s brother-cofounder. \u201cWith AI, the process is ongoing.\u201d<\/p>\n Automation at the bottom, augmentation at the top<\/p>\n Like most large tech conferences, the event doubled as a customer showcase. An HR executive from Bright Horizons, the early childhood education provider, described early internal attempts to deploy AI tools in recruiting.<\/p>\n \u201cSix months in, nobody could agree if it was succeeding because nobody could agree on what success was,\u201d she said, describing internal skepticism and fears of job loss. The company later implemented a Phenom-backed AI agent to automate reference checks for frontline workers \u2014 a narrower use case that proved easier to evaluate.<\/p>\n Other customers offered similarly quantifiable gains.<\/p>\n An Air Arabia executive said the company reduced the equivalent workload of five full-time recruiting roles \u2014 though not actual headcount \u2014 easing pressure on existing staff. At the University of Maryland, an HR leader said Phenom tools scheduled roughly half of 1,325 interviews during a test period, with about a quarter of those candidates ultimately hired.<\/p>\n Phenom executives are careful, and insistent, when discussing AI\u2019s impact on jobs.<\/p>\n \u201cLevel 1 and 2 recruiting tasks will go away, but not Level 3, which will get more valuable,\u201d Bayireddi said, arguing that routine work will be automated while higher-level, judgment-driven roles become more important. It\u2019s the \u201cJevons paradox\u201d of software<\/a> that Bayireddi has long championed<\/a>.<\/p>\n \u201cThe real world is messy,\u201d he added. \u201cContext is expensive.\u201d<\/p>\n That framing \u2014 automation at the bottom, augmentation at the top \u2014 has become a common defense across enterprise AI companies. It also leaves open the question of how many entry-level roles disappear along the way.<\/p>\n Still, the company\u2019s broader thesis reflects a shift happening well beyond HR.<\/p>\n