Okoro: Banks' AI Optimism Misreads IT Sector Pulse

Look — JP Morgan and HSBC saying AI will lift the IT sector is perfectly reasonable. AI spend is real, and demand isn’t imaginary. Trade Brains is right to flag that as a theme for investors, not just tech nerds.

Here’s what nobody tells you: a solid macro story can still be a terrible investing guide.

Banks are in the business of mapping big trends to price targets. They see AI budgets going up, they see IT vendors talking about AI on earnings calls, and they connect the dots. That part is fine. The issue is what gets smoothed out or ignored when you compress that messy reality into a clean slide about “AI upside for IT.”

First point: hype vs. adoption
Big models and flashy demos don’t pay bills. We’ve seen this movie. Think back to the first cloud wave — everyone slapped “cloud” on the deck; only a subset figured out migration, security, and billing well enough to actually mint cash.

AI is running the same pattern.

Everyone has pilots. Far fewer have production deployments woven into actual workflows, with real P&L impact. For IT vendors, that’s the difference between “AI revenue” that’s really one‑off consulting and “AI revenue” that’s contracted, recurring, and expandable.

If you buy a basket of IT names because banks say “AI will drive the next leg of growth,” you’re buying an adoption assumption. You are not buying guaranteed cash flow. That gap is where investors get burned.

Second point: vendor concentration and single‑point risk
The AI stack is already heavily concentrated. A small set of cloud platforms and model builders control access, pricing, and a big chunk of the compliance story. If they raise prices, throttle features, or change terms, downstream IT firms either swallow margin hits or pass costs to customers and risk churn.

Now add regulatory friction. Data residency rules, sector‑specific AI guidelines, and region‑by‑region standards will not move in sync. JP Morgan and HSBC can model growth curves; they can’t cleanly model a sudden rule change that forces clients in one geography to halt or re‑architect deployments while others keep going.

That’s not an academic risk. Ask any fintech that had to retrofit systems after a regulator changed its stance on data storage or consent. The AI version of that will be messier.

A practical investor lens
I ran operations at a Fortune 500, and my only question about tech spend was: does this show up in revenue, cost, or risk metrics — and how fast? That lens matters more than a 40‑page AI outlook.

When you read bank research or articles that echo it, hunt for evidence of scalable adoption in the numbers vendors actually report. Are customers paying for exploratory projects or multi‑year enterprise contracts tied to hard KPIs? Is AI showing up as sticky recurring revenue, or as spiky services work that disappears once the pilot wraps?

Wake up and stop treating every “AI win” in an earnings call as equal.

You want adoption metrics: implementation timelines, churn on AI‑tagged products, the mix of services vs. software, and the incremental support cost to keep these systems live. Production AI is expensive to run, monitor, and secure. High gross margin on a PowerPoint slide can quietly erode under real operating costs.

Ask the questions banks don’t ask
Which customers are actually buying — budget owners with real authority, or innovation teams burning discretionary funds?
What share of revenue is tied to legacy maintenance versus new AI features?
How exposed is a vendor to a single cloud partner’s roadmap and pricing?
What happens to margins if model‑serving costs jump, or a regulator forces an architectural rewrite in one region?

Those aren’t gotcha questions. They’re survival questions.

Three moves that actually tighten your strategy
Don’t just “get AI exposure.” Reweight toward companies with proven deployment playbooks. Managed services firms and systems integrators that embed AI into existing workflows usually feel less valuation whiplash than pure‑play model vendors selling promise and press releases.

Second, price in regulatory and data‑rights risk like it’s a cost line, not a footnote. If a business depends on cross‑border data flows or on training data that’s legally fuzzy, assume more friction and higher compliance spend than the pitch deck implies.

Third, demand operational KPIs, not just AI‑adjacent buzzwords. Push for disclosures on time‑to‑production for AI projects, average contract value for AI‑related deals, and the percentage of revenue coming from recurring AI services versus one‑off builds. If management can’t answer, that tells you plenty.

A quick historical check
Give me a break — we’ve done this “tech megatrend” cycle so many times. During the big data rush, everyone insisted Hadoop and friends would redefine enterprise IT. The winners weren’t the flashy distribution vendors; it was the firms that turned messy data plumbing into reliable, managed services with clear outcomes. Think of how many “next big thing” tools vanished while quiet integration shops and cloud platforms compounded value.

AI will rhyme with that history. A few platforms will dominate, a lot of tool vendors will be acquired or crushed, and a narrow slice of IT firms that can translate between models, regulation, and legacy systems will quietly soak up the durable value.

So when Trade Brains relays JP Morgan and HSBC’s upbeat view on AI for the IT sector, treat it as a weather report, not a map. The storm is real; the path it takes through specific companies is where your returns — or losses — will actually come from.