The Next AI Winners... - 8/2

Wall Street Buys "The Future." It's Better to Own the Toll Booths.             

A quick note from Behind the Markets

Wall Street keeps selling AI like it's free money.

But the real enterprise buyers don't care about demos. They care about liability.

And the minute AI starts touching payments, healthcare, critical infrastructure, or defense workflows… compliance isn't optional. It's the product.

It's Sunday — a good morning to study the three toll booths on that road: the governance layer, the identity layer, and the fraud layer. All three, conveniently, are on sale.

1) Enterprise AI Is Colliding With Reality: Governance Isn't a "Feature," It's the Business

The consumer AI story is fun. The enterprise AI story is expensive.

Why? Because the minute an AI model makes decisions inside a regulated workflow, the questions get nasty: Who signed off? What data was used? Can we reproduce the output? Can we audit it? Who pays when it breaks?

This is the wedge Wall Street underestimates. It's not just "compute." It's policy, monitoring, and documentation. And there's a 115-year-old company whose entire AI strategy is built on exactly that wedge.

Company: IBM (SYM: IBM)
Big Blue's quiet reinvention — the governance-first AI platform for banks, hospitals, and governments, trading at a fraction of the multiples its flashier rivals command.

Forget the IBM you think you know. The current version has staked its AI future on precisely this thesis: its niche is industrial-grade AI that handles sensitive data with rigorous compliance — capturing share in financial services and healthcare, where data integrity isn't a feature, it's a legal requirement — anchored by the watsonx platform and its governance tooling. And the demand is contracted, not conceptual: a $12.5 billion generative-AI book of business — locked-in consulting and license agreements from Fortune 500 companies building responsible-AI governance frameworks — plus the distribution coup of the summer: IBM embedded watsonx, Red Hat, and its AI automation stack directly into ServiceNow's enterprise AI platform — used by more than 8,100 enterprises processing trillions of dollars of workflow transactions annually. The trust layer, wholesaled. Wedbush's Dan Ives lifted his target to $350 at Outperform on exactly this — "what IBM has built, not slideware."

Now the honest file, and it's substantial. This stock has whipsawed all year on the AI debate — an AI-disruption scare wiped 13% off it in February, an Accenture guidance cut in June dragged it from $332-era highs down to $249 before a partnership-driven rebound — because the bear case is real: the same AI wave IBM sells could cannibalize its consulting arm and its mainframe-era COBOL moat. Two structural caveats: over 80% of that $12.5 billion AI book is consulting revenue, not high-margin software — the mix shift is the bet — and consulting growth has lagged. IBM reported mixed second-quarter 2026 results on July 22, 2026, posting revenue of $17.16 billion (up 1% year-over-year) and an adjusted earnings per share of $2.93. Total revenue slightly missed Wall Street consensus expectations, leading the company to lower its full-year constant-currency revenue growth forecast. What doesn't change either way: this is the cheapest large-cap pure-play on AI governance in existence, with a dividend while you wait.

Bottom line: The AI winners won't just build models. They'll build trust layers that regulators and CFOs can live with.

2) The Quiet AI Trade: Identity, Permissions, and Data Lineage

Most AI breaches won't look like Hollywood. They'll look like: a chatbot pulling the wrong internal doc. A model leaking customer data. A vendor training on something they shouldn't.

So the underfollowed spend isn't "another model." It's identity and access management, permissions and policy engines, data lineage, and monitoring for misuse.

One housekeeping note for regular readers: we've covered the access-management layer before (Okta, last month). Today's name sits one layer deeper — governance of identities — and it comes with a twist the whole market is only starting to price: the fastest-growing "employees" at big companies aren't human.

Company: SailPoint (SYM: SAIL)
The identity-governance specialist — now selling security badges for AI agents — trading 50% below its post-IPO high.

SailPoint's product answers these questions for a living: its platform lets organizations establish, control, and automate the policies that define who — and what — can access which data, across employee identities, non-employee identities, and machine identities, in service of security posture and regulatory compliance. The business is compounding — fiscal 2026 revenue grew 24% to $1.07 billion, free cash flow turned sharply positive (up 386% in its latest reported quarter), and the model is now cloud-subscription — and the AI twist is already shipping: Agentic Fabric, launched in May, manages security for AI agents and other non-human identities, extended by the completed acquisition of Tel Aviv's Entro to advance its agentic-identity vision. When your company deploys a thousand AI agents, every one needs permissions, monitoring, and an audit trail. That's a per-identity toll — and identities are about to multiply.

The honest file: the stock closed at $15.89 this week — down about 27% over the past year and roughly half its post-re-IPO high near $26 — because the market is digesting real issues: still GAAP-unprofitable (a $270 million net loss last fiscal year, largely buyout-era amortization), FY27 guidance that came in below street hopes, with Microsoft and Okta squeezing the lower end of the market, a Rosenblatt Neutral at $16 as the dissent, and the structural overhang that private-equity sponsor Thoma Bravo still owns the majority of the company — future share sales are a matter of when, not if (our Karman lesson from yesterday, in software form). Against that: 24 analysts average a Buy with a $18.90 target — about 19% above the price — with bulls as high as $25, and a management target of more than $2.1 billion in ARR by fiscal 2029. Speculative-growth sizing; the sponsor overhang and losses are the price of the theme.

Bottom line: If you control access and auditability, you become the toll collector on enterprise AI — especially when the "employees" are software.

3) The CFO Checklist: AI Budgets Are Getting Rewritten Around Measurable ROI

The AI hype phase is ending. Now comes the procurement phase.

CFOs will fund: tools that reduce fraud and errors, automation that cuts costs without breaking operations, systems that reduce compliance risk. They will not fund "strategic" pilots that never scale.

Here's the test to run on every AI-adjacent name you own, including today's three. Ask: Can the customer's CFO prove the ROI in a spreadsheet? IBM's pitch is auditable governance. SailPoint's pitch is provable access control. And the next section's company can show a bank, to the dollar, how much fraud it prevented last quarter. That's what "auditable outcomes" means. If a company's AI story requires the word "transformation" instead of a number, it fails the checklist — no matter how good the demo looks.

Bottom line: The best AI businesses will look boring because their value is provable.

4) Watch the Next Battleground: AI + Payments

Payments is where AI risk becomes immediate. Fraud is real-time. Chargebacks are real. Regulators are real.

Any company that can measurably reduce fraud losses — without adding friction — has massive pricing power. This is an underappreciated angle because it doesn't sound like "AI." It sounds like plumbing.

That's exactly why it's investable. And the deepest plumbing in the entire payments system just went on its biggest sale in a decade.

Company: Fair Isaac (SYM: FICO)
The company whose AI has guarded your credit card since before "AI" was a marketing word — cut in half by a regulatory fight that has nothing to do with its fraud business.

Everyone knows FICO for the credit score. Fewer know this version: FICO's Falcon Fraud Manager protects more than 4 billion payment cards worldwide, using artificial intelligence and patented fraud-detection methods to monitor transactions in real time — it has been the industry standard for decades, and it keeps expanding (Swiss banking giant PostFinance just widened its Falcon deployment across 2.4 million customers and millions of daily transactions). Here's the detail that ties today's whole issue in a bow: among FICO's newly granted patents is a real-time algorithm for detecting concept drift in machine-learning models — core technology for its Responsible AI strategy and Falcon. "Monitoring for drift" was literally a bullet in Section 2. FICO has patents on it.

Now the crash, and the honest anatomy of it: this was one of the great compounders in the S&P 500 — roughly $100 a share in 2016 to an all-time high near $2,217 by early 2025 — and it has since been nearly cut in half, with the stock down 48% over the past year to a recent close near $1,044, even as the 20-analyst average target of roughly $1,565 sits 50% above the price. The cause: concerns about VantageScore adoption after regulators opened conforming mortgages to lender score choice, pricing-pressure fears, and AI-related software weakness. That mortgage-scores fight is the real, structural bear case — the Scores segment is the cash engine, and competition there is existential-adjacent — which is why fair-value opinions span an almost absurd range ($1,205 to $2,628). The bull rebuttal: incremental scores cost almost nothing to produce, so pricing flows nearly entirely to the bottom line, and the long-term thesis — including the fraud and decisioning software — remains intact. At roughly 17–18x forward earnings — a decade-low multiple for this franchise — you're being paid to sit through the regulatory fight and getting the Falcon fraud toll booth thrown in. One caution: the fight could genuinely go badly. FICO reported its third quarter fiscal 2026 results on July 29, so those results are already priced in.

Bottom line: AI is becoming a security product. The winners will be the ones who save real dollars — and can prove it, transaction by transaction.

Before You Go

Wall Street loves to buy "the future."

You're better off buying the toll booths.

Enterprise AI will be forced into governance, monitoring, and compliance — and that spend is stickier than the hype. Today's three collect at different gates: the governance platform banks already trust, the badge office for a coming army of AI agents, and the fraud brain that's guarded your wallet for decades. All three are marked down, and I've told you exactly why — a consulting-mix question, a sponsor overhang and red ink, a regulatory knife-fight. Discounts always come with reasons. Your job, this Sunday, is deciding which reasons are temporary.