For years, the prevailing wisdom in Silicon Valley held that artificial intelligence would be the great equalizer — that nimble startups armed with large language models could storm the gates of enterprise software and dethrone the incumbents who’d spent decades accumulating customers, contracts, and complexity. Leagh Turner thinks that’s exactly backward.
Turner, who took over as CEO of Coupa Software in 2023, has been making an argument that cuts against the grain of the AI hype cycle: the companies best positioned to win in the age of AI aren’t the insurgents. They’re the ones already sitting on mountains of proprietary transactional data. And she’s betting Coupa’s future on it.
In a wide-ranging interview with Fortune, Turner laid out her thesis with characteristic bluntness. “Everyone’s talking about AI like it’s some kind of blank slate,” she said. “But AI without data is just math. And we have the data.” It’s a statement that sounds simple — almost reductive. But the strategic implications are enormous, and they’re reshaping how Wall Street and enterprise buyers think about the competitive dynamics of business software.
Coupa, for the uninitiated, is a business spend management platform that processes trillions of dollars in cumulative transactional data across procurement, invoicing, expenses, and supply chain operations. The company was taken private by Thoma Bravo in a $8 billion deal in 2023, and it has been operating largely out of the public spotlight since. Turner’s comments to Fortune represent one of her most detailed public articulations of where she sees the company heading — and why she believes the AI era actually widens the moat for established SaaS players rather than narrowing it.
The core of her argument rests on a distinction that gets lost in much of the breathless AI coverage: the difference between general-purpose intelligence and domain-specific intelligence. Large language models like those from OpenAI, Anthropic, and Google are extraordinarily capable at processing and generating text. They can summarize contracts, draft emails, and even write code. But when it comes to making high-stakes business decisions — which supplier offers the best risk-adjusted pricing, whether a particular invoice pattern suggests fraud, how to optimize payment terms across a global supply chain — generic AI falls short.
What’s needed is AI trained on real-world business transactions. Billions of them. Spanning industries, geographies, and economic cycles.
That’s exactly what Coupa claims to have. According to Turner’s comments in Fortune, the company’s platform has processed over $6 trillion in cumulative business spend, generating a dataset she describes as “the world’s largest community of transactional intelligence for business spending.” The company uses this data — anonymized and aggregated — to power what it calls its Community Intelligence engine, which benchmarks pricing, identifies savings opportunities, and flags anomalies across its customer base.
Turner’s point is that no startup, no matter how clever its algorithms, can replicate this dataset from scratch. “You can’t just go build $6 trillion in spend data,” she told Fortune. “That took us 18 years.”
She’s not wrong about the difficulty. But the question is whether the data advantage translates into a durable AI advantage — or whether it’s a temporary head start that erodes as foundation models get smarter and new data sources proliferate.
The debate is playing out across enterprise software. SAP, Oracle, Salesforce, and ServiceNow have all made similar arguments about their proprietary data assets. SAP CEO Christian Klein has repeatedly emphasized that SAP sits on 87% of the world’s transaction data and that this gives it a structural advantage in building AI that actually works for business. Salesforce’s Marc Benioff has pushed the same line with his company’s Data Cloud and Einstein AI initiatives. The incumbents, in other words, are singing from the same hymnal.
And there’s evidence the market is listening. Enterprise software stocks with clear data moats have generally outperformed pure-play AI startups in terms of actual enterprise adoption. A recent analysis from Gartner noted that by 2026, over 60% of enterprises would prefer AI capabilities embedded in their existing platforms over standalone AI tools — largely because of data integration challenges and security concerns. That’s a tailwind for companies like Coupa.
But skeptics abound. Venture capitalists backing AI-native startups argue that the incumbents’ data advantages are overstated. “Legacy data is often messy, siloed, and poorly structured,” one prominent VC told a panel at a recent enterprise tech conference. “Just because you have a lot of it doesn’t mean it’s useful for training AI.” There’s also the argument that foundation models are advancing so rapidly that they’ll soon be able to infer business patterns from much smaller datasets, reducing the importance of scale.
Turner has a response to this too. She told Fortune that Coupa has invested heavily in data quality and normalization over the years, precisely because its Community Intelligence features depended on clean, structured transactional data long before the current AI boom. “We didn’t start cleaning our data when ChatGPT launched,” she said. “We’ve been doing it since the beginning because our product required it.”
This is a subtle but important point. Companies that built AI-adjacent features — recommendation engines, anomaly detection, benchmarking tools — before the generative AI wave have a structural advantage in data readiness. Their datasets were already curated for machine learning. They didn’t need a crash program to make their data AI-ready; it already was.
Coupa has been rolling out new AI-powered features at a steady clip. The company’s AI-driven procurement recommendations now factor in supplier risk scores, ESG ratings, geopolitical exposure, and historical performance data. Its invoice processing system uses machine learning to detect duplicate payments and fraudulent submissions with what Turner claims is significantly higher accuracy than manual review or rule-based systems. And its supply chain modules now include predictive analytics that can flag potential disruptions weeks before they materialize.
None of this is unique to Coupa, of course. Every major enterprise software company is racing to embed AI into its products. What Turner argues is different is the depth and specificity of the data powering these features. A generic AI model might be able to tell you that a particular supplier is based in a region prone to natural disasters. Coupa’s model can tell you that this specific supplier has a 23% higher rate of late deliveries during Q4, that their pricing is 7% above the community benchmark, and that three alternative suppliers with better track records are available in adjacent regions. That level of specificity, Turner contends, only comes from years of accumulated transactional data.
The private equity angle adds another dimension. Since Thoma Bravo took Coupa private, the company has had the luxury of investing without the quarter-to-quarter scrutiny of public markets. Turner has used that breathing room to accelerate R&D spending on AI, according to her comments, while also pushing operational efficiency — the classic private equity playbook of growing the top line while expanding margins. The question is whether Coupa will return to public markets, and if so, whether its AI story will command the kind of premium multiples that investors are currently awarding to AI-forward enterprise companies.
There are hints that an IPO is on the horizon. Thoma Bravo has been actively exiting positions in its portfolio, and Coupa’s revenue trajectory — while not publicly disclosed — is believed to be strong based on industry estimates and customer expansion data tracked by firms like Gartner and IDC. Turner didn’t address IPO timing in her Fortune interview, but her willingness to make a high-profile public case for Coupa’s AI strategy suggests the company is at least beginning to lay the groundwork for a return to public markets.
The broader implications of Turner’s thesis extend well beyond Coupa. If she’s right — if proprietary transactional data is the real moat in enterprise AI — then the competitive map of business software looks very different than the one many AI evangelists have been drawing. It means the winners won’t be the companies with the best models. They’ll be the companies with the best data. And in enterprise software, the companies with the best data are almost always the ones that have been around the longest.
This is uncomfortable for the startup world. It suggests that the AI revolution in enterprise software might not be a revolution at all — but rather an acceleration of existing advantages. The rich get richer. The incumbents get more entrenched. The startups, for all their speed and innovation, find themselves locked out of the data they need to compete.
Not everyone buys this framing. Some industry analysts point out that the most transformative AI applications in business haven’t come from incumbent platforms but from new entrants that reimagined entire workflows. Companies like Glean, which is building AI-powered enterprise search, or Harvey, which is applying AI to legal work, are creating value precisely because they’re not constrained by legacy architectures and existing product roadmaps. Their argument: incumbents have data, but they also have technical debt, organizational inertia, and a tendency to bolt AI onto existing products rather than rethinking them from the ground up.
Turner acknowledged this tension in her Fortune interview, noting that Coupa has been rebuilding portions of its platform specifically to take advantage of modern AI architectures. “We’re not just putting a chatbot on top of our existing product,” she said. “We’re rearchitecting how decisions get made inside the platform.” She pointed to Coupa’s new agentic AI capabilities — autonomous agents that can execute procurement workflows, negotiate with suppliers, and approve routine transactions without human intervention — as evidence that the company is thinking beyond simple feature additions.
Agentic AI is the buzzword of the moment in enterprise tech, and every major vendor is racing to claim leadership. But Turner’s version of the pitch is more grounded than most. She argues that autonomous agents are only as good as the data and business logic they’re trained on, and that Coupa’s 18 years of transactional history give its agents a decision-making foundation that can’t be replicated by a startup training on synthetic data or limited customer deployments.
It’s a compelling argument. Whether it holds up will depend on execution — always the hardest part. Coupa needs to prove that its AI features deliver measurable ROI to customers, that its data advantage translates into better outcomes than competitors can offer, and that it can move fast enough to stay ahead of both the startup swarm and the other incumbents circling the same opportunity.
The enterprise AI wars are far from settled. But Turner has staked out a clear position: in a world where everyone has access to the same foundation models, data is the differentiator. And incumbents, for all their supposed disadvantages, might just have the strongest hand.
Time will tell if she’s right. But the argument deserves more serious consideration than it’s getting from an industry still intoxicated by the promise of AI-native disruption. Sometimes the boring answer — we’ve been collecting this data for two decades, and now it’s incredibly valuable — turns out to be the right one.
WebProNews is an iEntry Publication