The AI trap: Faster solution, same problem

A survey of ~6,000 executives finds 70% of firms use AI but over 80% report zero measurable productivity impact over the past three years; only 4–6% of firms show real gains. The article argues the failure is process, not models — firms that simplified workflows first (JPMorgan on legal review) drove P&L improvements, while examples like Omega Healthcare (50% turnaround reduction, 15,000 employee hours recovered/month, 99.5% accuracy) and a Fortune 500 invoice program (45 min → 6.75 min, ~85% cycle-time reduction) show the magnitude of benefits when process redesign precedes AI. Recommendation: audit top AI initiatives now and require evidence that processes were simplified before automating, or terminate projects automating unnecessary work.

Analysis

The dominant mistake I expect the market to price repeatedly over the next 6–12 months is conflating local task acceleration with system-level productivity. When you speed a non‑value step, downstream queues and governance effort typically absorb 60–100% of the apparent time savings; that translates into flat throughput and higher variable costs even as adoption metrics ‘look’ strong. Investors who only watch adoption or NPS will miss the real P&L signal: cycle time, unit cost and exception rates.

Banks and large enterprise lenders that pair lean process redesign with automation will extract most of the economic value; redeployable headcount and reduced exception rework flow straight to EBIT. Conversely, digital incumbents that automated surface work without redesign (consumer tutoring, ticketing platforms) face structural revenue erosion as complexity distribution and market substitution shift the constraint off their optimized step. Expect outsized dispersion: winners posting mid‑teens margin tailwinds and losers facing 20–50% revenue downside over 12 months.

Near‑term catalysts to watch are CIO/COO audits, quarter‑over‑quarter cycle‑time disclosures, and vendor contract renewals where buyers insist on process metrics. Reversals happen if macro forces force hiring freezes (which forces process simplification) or regulators mandate auditability of AI decisions — either compressing or expanding the runway for automation depending on who controls process change.

Actionable read: treat AI as an amplifier of org design, not a product. The cheapest way to test management discipline is whether they can show what they removed before they show what they built; those who can will compound ROIC, those who can’t will burn cash and multiple.

Key Decisions for Investors

• Short CHGG (size 1–2% NAV) over 3–9 months: thesis is continued revenue share loss + product obsolescence. Target: 30–50% downside if Q3/Q4 guidance shows persistent churn; stop-loss +20% if management reports sustainable retargeting to higher‑value products within two quarters.
• Short KLAR (size 1% NAV) over 3–9 months or buy out‑of‑the‑money Jan‑2027 puts: thesis is routing/escalation costs and margin pressure from ticket complexity. Target: 20–40% downside if customer‑service FTE growth resumes; hedge with 10–15% notional calls if positive cash flow surprises arrive.
• Long JPM (size 2–3% NAV) over 6–12 months, or buy Jan‑2027 calls (delta ~0.40): thesis is process‑first automation in commercial lending creating measurable cycle‑time and cost improvements and redeployable legal FTEs. Expect 15–30% upside as ROE/take‑rate expands and loan throughput accelerates; downside limited by bank‑sector cyclicality — use 6–8% trailing stop.