In February 2026, Jamie Dimon made a claim that most enterprise leaders can only dream of: JP Morgan Chase’s $2 billion AI investment had “paid for itself.” Not “shows promise.” Not “is on track.” Paid for itself — $2 billion in measured benefits offsetting $2 billion in costs.
Break-even isn’t a moonshot return. But in a landscape where only 20% of enterprises can prove AI drives any revenue at all, the fact that JP Morgan can put a dollar figure on AI’s contribution — and defend it to shareholders — puts them in rare company. The question worth studying isn’t how much they spent. It’s what they measured, and how.
The Scale of the Bet
JP Morgan’s $2 billion annual AI investment sits within a $17 billion technology budget that grew to $19.8 billion in 2026 — a 10% increase year over year. The bank employs more than 2,000 AI and machine learning specialists, including 900 data scientists, 600 machine learning engineers, and 200 AI researchers. This isn’t a skunkworks team running experiments. It’s a division-level commitment that treats AI as core infrastructure alongside payment systems and risk controls.
The centerpiece is the LLM Suite, an internal platform built on models from OpenAI and Anthropic that 150,000 employees use weekly. The platform is updated every eight weeks with new enterprise data, and by late 2025, roughly half of all LLM Suite users were using it daily. At that adoption rate, AI interactions at JP Morgan generate measurement data at a scale most enterprises can’t match.
But scale alone doesn’t prove value. What makes JP Morgan’s approach worth studying is that they measure AI at the use case level — not just at the platform level. Every one of their 600-plus production AI use cases has specific metrics tied to specific business outcomes.
What They Actually Measured
JP Morgan’s AI measurement spans four categories, each connecting AI activity to a different type of business outcome:
Time recovery. The most widely cited metric: LLM Suite users report saving an average of four hours per day. At 150,000 weekly users, that’s potentially 600,000 hours per week of recovered employee time — time that gets redirected to higher-value work, client engagement, and analysis that was previously crowded out by routine tasks. Time recovery is the most accessible AI metric because it’s easy to measure and easy to understand, but it’s also the most dangerous if not connected to downstream outcomes. Four hours “saved” only creates value if those hours are deployed productively.
Cost reduction. The COiN (Contract Intelligence) platform provides the clearest cost reduction case study. Before AI, JP Morgan’s legal team manually reviewed commercial loan agreements — a process that consumed approximately 360,000 hours annually. COiN now reviews 12,000 documents in seconds rather than weeks, reducing legal operations costs by 30% and cutting compliance errors by 80%. The cost reduction is measured against a known baseline (manual review hours and error rates), making the calculation straightforward and defensible.
Revenue impact. JP Morgan’s AI trading algorithms illustrate revenue-side measurement. The bank reported that AI-driven trading systems improved win rates from 52% to 63% and saved $25 million in slippage costs. Revenue impact is harder to measure than cost reduction because attribution is more complex — markets move for many reasons, and isolating AI’s contribution requires careful methodology. But JP Morgan’s approach of measuring specific trading performance metrics (win rate, slippage) rather than aggregate revenue provides a more defensible attribution model.
Value creation mapping. Across customer personalization, trading, fraud detection, and credit decisioning, JP Morgan identified $1 to $1.5 billion in value creation from AI. This portfolio-level view is what enables the “$2 billion investment paid for itself” claim — it aggregates use case-level measurements into an enterprise-wide picture that can be presented to shareholders.
Lessons for the Rest of Us
Most enterprises aren’t JP Morgan. They don’t have 2,000 AI specialists or a $17 billion technology budget. But the measurement principles that underpin JP Morgan’s ability to claim ROI are applicable at any scale.
Measure at the use case level, not the platform level. JP Morgan doesn’t report a single “AI ROI” number derived from aggregate spending and aggregate benefits. They track 600-plus individual use cases, each with defined metrics. This granularity is what makes the portfolio-level claim credible — it’s built bottom-up from measured outcomes, not estimated top-down from spending. Even an enterprise with five AI use cases can apply this discipline: define the success metric for each use case, measure it against a baseline, and report results individually before aggregating.
Track multiple metric categories. Time recovery alone doesn’t prove ROI. Cost reduction alone doesn’t capture the full picture. Revenue impact alone is too hard to attribute without supporting data. JP Morgan tracks all four categories (time, cost, revenue, value creation) and presents them together. This multi-dimensional view is more credible to boards and CFOs than any single metric, because it demonstrates that the organization has instrumented AI measurement comprehensively.
Build governance alongside measurement. JP Morgan’s Model Risk Governance function and Firmwide Chief Data Officer aren’t separate from AI measurement — they’re integral to it. Governance forces the organization to define what each AI system does, which creates the accountability structure that measurement requires. As we’ve seen across 100-plus AI agent deployments, the enterprises with the strongest ROI data are the ones with the most rigorous governance frameworks.
Treat AI as infrastructure, not R&D. JP Morgan reclassified AI from an innovation investment to core infrastructure — the same category as payment processing and risk management. This shift has measurement implications: infrastructure has uptime, performance, and cost-efficiency metrics that are reviewed continuously, not evaluated in quarterly innovation reviews. When AI becomes infrastructure, measurement becomes operational rather than experimental.
The ServiceNow Parallel
JP Morgan isn’t the only enterprise betting big on AI measurement. ServiceNow’s AI business reached $600 million in annual contract value in 2025 and expects to exceed $1 billion by the end of 2026. Like JP Morgan, ServiceNow measures AI at the product level — tracking adoption, usage patterns, and customer value creation for each AI capability rather than reporting a single aggregate number.
The pattern is consistent across enterprises that prove AI ROI: measurement happens at the individual use case or product level, governance provides the accountability structure, and results are aggregated into a portfolio view for executive and board reporting. The enterprises stuck in pilot purgatory do the opposite — they measure at the platform level, lack governance infrastructure, and can’t connect aggregate spending to specific outcomes.
What This Means for 2026
McKinsey projects that AI could unlock $200 to $340 billion annually in value for financial services alone, and the industry is responding — more than 70% of financial institutions were using AI at scale by late 2025, up from 30% in 2023. But the gap between “using AI” and “proving AI ROI” remains wide. JP Morgan is one of the few financial institutions that can put specific dollar figures on specific AI outcomes.
The lesson isn’t that enterprises need to spend $2 billion. It’s that the measurement infrastructure JP Morgan built — use case-level tracking, baseline metrics, multi-category measurement, governance integration — is what enables the ROI claim. That infrastructure can be built at any scale, for any number of AI initiatives. The cost of building it is a fraction of the cost of running AI without it.
If your organization is investing in AI but can’t answer “what’s the return?” with specific numbers, the problem isn’t your AI. It’s your measurement. Our AI ROI framework provides the methodology, and Olakai’s platform provides the instrumentation to track AI value the way JP Morgan does — at the use case level, against baselines, across time, cost, revenue, and risk.
Ready to measure your AI like JP Morgan? Schedule a demo and we’ll show you how enterprises track AI ROI across every initiative — without needing a $17 billion technology budget.