I’ve been a CEO for 25 years. The AI hype and hysteria is getting old

The article argues that AI should be evaluated on a company-by-company ROI basis rather than treated as an existential threat or universal mandate. The author says Capitolis is investing hundreds of thousands of dollars annually in AI, but has not yet seen returns, though engineering productivity could improve by about 25% over time across roughly 100 developers. Overall, the piece is a cautionary commentary on AI hype, with limited immediate market impact.

Analysis

The market is still pricing AI too much like a universal productivity shock and too little like a selective cost-out tool. That mismatch matters: the first-order beneficiaries are the infrastructure and workflow layers that monetize experimentation now, while most operating companies will only see incremental margin help after a long integration cycle. In other words, the near-term P&L impact is likely to be concentrated in vendors selling picks-and-shovels, not broadly reflected in end-demand or revenue acceleration across corporate users.

For DASH, the second-order read is positive but modest: AI can reduce support cost, improve dispute handling, and raise conversion in customer ops, but the franchise is still driven far more by marketplace density, logistics reliability, and local economics than by model quality. The real competitive risk is not that AI destroys the business, but that it narrows the gap between incumbents and smaller platforms on service automation, which caps differentiation and compresses take-rate elasticity over time. That makes AI more of a margin lever than a growth catalyst here.

The contrarian point is that consensus is overestimating the speed at which boards can translate AI enthusiasm into measurable ROI. Procurement friction, data governance, and change-management costs will likely push meaningful payback out by 12-24 months for most non-software businesses, which means current spend may be a drag before it becomes a benefit. The trap for investors is extrapolating vendor hype into near-term earnings revisions; the better setup is to own companies that sell AI adoption, not those that merely announce it.

Catalyst-wise, watch for the next two earnings cycles: if management teams start quantifying AI savings in headcount growth, support costs, or engineering throughput, multiples on exposed software/infrastructure names can expand again; if not, the market may punish AI capex as undisciplined spend. The reversal risk is simple: any sign that AI is delivering 20%+ productivity in customer support or coding at scale would pull forward margin inflection and force a re-rating across software, fintech, and internet names with heavy operating leverage.

Key Decisions for Investors

• Maintain a tactical long bias in AI infrastructure beneficiaries over AI adopters for the next 6-12 months; favor names monetizing spend immediately rather than promised productivity gains.
• For DASH, avoid paying up for a full AI re-rating; use any post-earnings strength to trim or hedge because the likely benefit is incremental margin, not a step-change in growth.
• Pair trade idea: long AI workflow / tooling providers, short a basket of non-essential software names that are advertising AI features without clear payback; target 12-18 months, because the market will start discriminating on realized ROI.
• Consider a call spread on a market leader in automation/software infrastructure into the next two earnings seasons, funded by a put spread on an operating company where AI spend is likely to compress near-term margins.
• If management commentary starts to quantify AI-driven productivity gains in the 15%-25% range, cover shorts quickly; that is the level where the market will begin to pull forward 2026-2027 margin upside.