Jacob Sheldon

An NBER working paper surveyed nearly 6,000 CEOs and CFOs across four countries. Not a vendor study. Central bank researchers designed it specifically to measure whether AI deployment is translating to results. The answer, mostly, is no. Not yet. 63% of finance departments have deployed AI. 21% report measurable value. The RAND Corporation puts the broader AI project failure rate at 80%. MIT says 95% of GenAI pilots never reach production. Adoption and impact are running on completely different timelines. Almost everyone is deploying. Almost nobody can prove it's working at the firm level. The gap seems to come down to measurement. Most organizations track whether they're using AI. Very few track whether the work actually got better. Those are different questions with completely different answers.

Stanford used ADP payroll data from millions of workers to measure AI's labor market impact. Not a survey. Payroll records. Workers aged 22-25 in AI-exposed jobs: 13% relative employment decline since late 2022. Workers over 30 in the same roles: stable or growing. Big tech cut entry-level hiring 25%. UK tech cut graduate roles 46%. The pipeline problem here is well documented. What's less discussed is what a solution actually looks like. IBM might be the most interesting signal. In February they announced they're tripling entry-level hiring in 2026, but completely redesigning what those roles are. Less routine coding, more customer interaction and AI oversight. Their CHRO's bet: "The companies three to five years from now that are going to be the most successful are those companies that doubled down on entry-level hiring in this environment." The ATM precedent is worth remembering. ATMs cut tellers per branch from 21 to 13, but didn't kill the job. They changed it. Tellers went from counting cash to relationship banking. The role was redesigned, not preserved. That's probably the pattern here too. The junior roles don't come back in their old form. They get rebuilt around AI oversight and judgment-building. The question is how fast.

AI tokens got 1,000x cheaper in 3 years. $37 per million tokens in 2022 to under $0.40 today. Enterprise AI spending surged 320% anyway. Microsoft's CEO called it when DeepSeek dropped: "Jevons Paradox strikes again." Make a resource cheaper, people use vastly more of it. Total spend goes up. The math that tends to catch founders off guard: a single agentic workflow can trigger 10-20 model calls behind the scenes. A $0.15-per-execution pipeline looks great until you're processing 500,000 requests a day. Inference is 85% of the enterprise AI budget now. Not training. Running. The companies getting this right aren't chasing the cheapest model. They're treating inference cost as an architecture decision, not an optimization you get to later.

Gartner predicts 40% of enterprise applications will have embedded AI agents by the end of 2026. Up from less than 5% in 2025. That's not a gradual adoption curve. That's a phase change. Think about what that means practically. Every major enterprise software category (ERP, CRM, HRIS, finance, procurement, legal, IT service management) is racing to ship some version of "AI agents" this year. The ones that don't will look outdated by Q4. For software companies, this creates a 3-to-6 month window. Either you have a credible agent strategy by mid-year, or you're explaining to customers why your competitor does and you don't. For startups, the opportunity is in the gap between "we shipped an AI agent" and "our AI agent actually works reliably in production." That gap is enormous right now. The tools, infrastructure, and expertise needed to make agentic AI work at enterprise scale barely exist yet. The best time to build the picks and shovels for the agent gold rush was six months ago. The second best time is now.

The most useful signal for AI founders right now isn't what VCs are funding. It's what they stopped funding. The dead categories in early 2026: thin wrappers on LLM APIs with a chat interface. Vertical SaaS that just bolted on AI without a technical moat. Anything where unit economics don't survive inference costs eating the margin. What's still fundable: AI-native infrastructure. Vertical SaaS with proprietary data moats. "Systems of action" that complete tasks, not just provide information. Anything deeply embedded in mission-critical workflows where switching costs are existential. The bar shifted. "We added AI" is not a pitch anymore. The pitch is "we have data nobody else has" or "we're so embedded that ripping us out would shut down your operations." This connects to the SaaSpocalypse narrative from a couple weeks ago. The $2 trillion wipeout in software stocks wasn't about AI killing SaaS. It was about investors realizing which SaaS companies actually have defensible positions and which ones were coasting on switching costs that AI just eliminated. If you're building right now, the question isn't "does my product use AI?" It's "would my product still win if every competitor also had AI?"

Gartner predicts 90% of finance functions will deploy at least one AI solution by the end of this year. But less than 10% will actually reduce headcount. Wolters Kluwer's data backs this up from a different angle: agentic AI adoption among finance leaders is set to jump 6x this year, from 6% to 44%. The money is pouring in. The tools are being deployed. But the org charts aren't changing. Why? Because automation is eliminating data entry and reconciliation work while simultaneously creating new work: monitoring AI outputs, managing exceptions, auditing automated workflows, and doing the strategic analysis that humans were too busy for before. The Klarna lesson applies here. Replacing humans sounds good in a press release. It doesn't survive contact with reality in complex, high-stakes domains where errors have real consequences. For founders building in fintech and finance, the opportunity isn't "replace the accountant." It's "give the accountant superpowers and charge for the platform." Very different product, very different go-to-market.

Here's a failure mode in multi-agent AI systems that almost nobody is talking about: memory poisoning. Galileo AI studied 1,642 execution traces across production multi-agent systems. They found that a single compromised or hallucinating agent poisoned 87% of downstream decision-making within 4 hours. The mechanism is simple and terrifying. Agent A hallucinates a fact. Stores it in shared memory. Agent B retrieves that "fact" as verified truth. Agents C, D, and E build on it. Within hours, the entire system is operating on fabricated information and none of the standard monitoring tools catch it. Real example: an inventory agent invents a nonexistent SKU. That triggers pricing, stocking, and shipping workflows for a phantom product. Everything looks normal in the dashboards. Traditional software monitoring (uptime, latency, error rates) tells you nothing about whether your agents are making correct decisions. The failure rate in production multi-agent systems ranges from 41% to 87%, according to these studies. Enterprises are racing to deploy multi-agent architectures while the fundamental reliability problem is unsolved. The monitoring and observability tools for agentic systems are going to be a very big market. galileo.ai/blog/multi-age…

Last summer, a controlled study by METR found that experienced developers using AI coding tools were 19% slower than without them. And the developers themselves estimated they were 20% faster. A nearly 40-point perception gap. That's a striking result. But here's what I keep thinking about: the study used Cursor Pro with Claude 3.5 Sonnet, which was the frontier in early 2025. The models and tooling have improved dramatically since then. Agentic coding workflows barely existed at the time. Context windows were a fraction of what they are now. The cautionary finding still matters. If you're measuring AI adoption rates instead of actual output quality, you might be tracking a feeling rather than a result. PR sizes went up 150%. Bug counts up 9%. Review time ballooned 91%. Those are real costs that get hidden when you only measure speed of first draft. But I'd bet the results look very different if you reran this study today. The tools have gotten substantially better at understanding large codebases, maintaining context across sessions, and catching their own mistakes before a human has to. The gap between "AI generates code" and "AI generates correct code" is closing fast. The real takeaway isn't "AI doesn't help." It's "measure the right things, and expect rapid improvement." The teams that built good evaluation frameworks in 2025 are the ones seeing real gains now. metr.org/blog/2025-07-1…

BNY Mellon just deployed 20,000 AI agents across their global operations. Not a pilot. Not a proof of concept. 20,000 agents running in production across 125 use cases. This is one of the largest AI deployments in financial services. BNY didn't hand this to a consulting firm and wait 18 months. They built an internal platform called Eliza, partnered with OpenAI, and rolled it out with the goal of "AI for everyone, everywhere, and in everything." What stands out to me is the contrast with Deloitte's data showing only 11% of enterprises have agentic AI in production. BNY is in that 11%. Most of their competitors are still running pilots. The pattern I keep seeing: the companies moving fastest on AI aren't buying the most tools. They're building internal platforms that let every team deploy AI for their own workflows. The platform approach scales in a way that buying point solutions doesn't. For fintech founders, this is worth paying attention to. The biggest financial institutions are building internal AI capabilities at speed. I think that actually creates more opportunity for startups, not less. When a bank has 20,000 agents running, they need better data infrastructure, better compliance tooling, better monitoring. That's all greenfield. openai.com/index/bny/ #AI #fintech #banking #startups

The "SaaSpocalypse" narrative hit Wall Street hard in February. Software stocks lost roughly $2 trillion in market cap. Price-to-sales ratios compressed from 9x to 6x. Some investors are acting like SaaS is finished. I think the panic is overdone, but the underlying shift is real. Here's what's actually happening: the cost to build software dropped dramatically because of AI coding tools. That means the bar for "is this SaaS product worth paying for?" just went up. If a competent team can rebuild 80% of your product's functionality in a week, your remaining value has to come from data, integrations, reliability, or some other moat that's hard to replicate. The SaaS companies that survive this will be the ones solving problems that are genuinely hard, not the ones that were just "good enough" because building an alternative was too expensive. That filter is going to produce better software for everyone. This is healthy. Painful for some incumbents, but healthy for the ecosystem.

Deloitte just published their State of AI 2026 report. The number that jumped out at me: only 11% of enterprises have agentic AI running in production. 38% are piloting. 30% are still just exploring. Meanwhile, 74% of organizations say they plan to deploy autonomous AI agents within the next two years. And only 21% have any governance framework in place for those systems. This is the execution gap that nobody in the AI hype cycle wants to talk about. The technology is ready. The pilots work. But getting from "cool demo" to "running in production at scale" requires solving boring problems like data infrastructure, permissions, compliance, and audit trails. Most organizations aren't there yet. The opportunity for founders: the companies stuck in pilot mode need help getting to production. The tools and services that bridge that gap are going to be very valuable over the next 18 months. Not the AI models themselves, but everything around them that makes them actually usable in an enterprise context. deloitte.com/us/en/about/pr…

Klarna replaced 700 customer service agents with AI. The CEO bragged about it publicly. The AI handled 75% of all customer chats and Klarna saved millions. Then quality dropped. Customer satisfaction fell. Repeat inquiries went up. And now Klarna is hiring humans back. This is the most useful AI case study out there right now, because it shows both sides honestly. The AI worked incredibly well for straightforward requests. It failed on anything requiring empathy, nuance, or complex problem-solving. The lesson isn't "AI doesn't work for customer service." It clearly does, for a specific slice of it. The lesson is that going 100% AI on anything customer-facing is a mistake if you haven't figured out where the handoff to a human needs to happen. The companies getting this right are running hybrid models where AI handles volume and humans handle complexity. That's not a compromise. That's the actual answer. bloomberg.com/news/articles/…

OpenAI just raised $110 billion. Amazon put in $50B, SoftBank $30B, Nvidia $30B. It's the largest private funding round in history and values the company at $730 billion. The scale of this is hard to wrap your head around. For reference, that single round is larger than the entire annual GDP of most countries. Amazon alone committed more to this deal than the total venture capital deployed across all of Europe last year. g for founders: the infrastructure layer of AI is now being funded at a level that guarantees it will be commoditized. When this much capital flows into building out compute, models, and tooling, the cost of using AI in your product drops fast. We've already seen it with API pricing falling 90%+ over the past two years. For anyone building on top of AI right now, this is good news. The foundation is getting stronger and cheaper. The opportunity is moving up the stack to the application layer, where you're solving a specific problem for a specific customer better than anyone else. The billion-dollar question: does this level of concentration in AI infrastructure help or hurt the broader startup ecosystem? I think it helps, but it's worth watching closely. openai.com/index/scaling-…

Retool's new report says 35% of enterprises have already replaced at least one SaaS tool with something custom-built. 78% plan to build more this year. I think about this number a lot. The conventional wisdom for the last 15 years has been "don't build what you can buy." And that was correct when building custom software meant hiring a dev team, spending months on requirements, and maintaining the thing forever. But the calculus is genuinely changing. When a non-technical ops person can use an AI coding tool to build a working internal dashboard in an afternoon, the old framework breaks down. The build cost dropped by 10x, but the maintenance cost didn't. That's the part people aren't talking about enough. The 35% who already replaced a SaaS tool are going to learn something interesting over the next 12 months: building is now cheap, but maintaining, securing, and scaling custom tools still takes real work. The companies that figure out which things to build and which to buy will have a meaningful advantage. I don't think SaaS is dying. I think the bar for what justifies a SaaS subscription just got a lot higher. retool.com/blog/ai-build-…

OpenAI just launched GPT-5.4, and the part that matters most for fintech founders isn't the model itself. It's the infrastructure around it. The headline specs are impressive: 1 million token context window, 33% fewer factual errors than GPT-5.2, and native computer-use capabilities built in. But what caught my attention is the financial data integration. GPT-5.4 ships with connectors for FactSet, MSCI, Third Bridge, and Moody's, plus embedded plugins for Excel and Google Sheets. Think about what that means for anyone building financial products. The gap between "AI that can reason about numbers" and "AI that can actually pull live financial data and build models in a spreadsheet" just closed significantly. That used to require months of custom integration work. The computer-use piece is also worth paying attention to. This is the first general-purpose model with native ability to operate across applications, browsers, and desktops. For founders: if you're building anything that touches financial data or enterprise workflows, the platform capabilities here are more important than the benchmark improvements. The models are getting commoditized. The integrations are where the real leverage is. openai.com/index/introduc…

Stripe just launched token-level billing for AI products. You pick your models, set your markup percentage, and Stripe handles metering, invoicing, and revenue tracking automatically. This is a big deal for anyone building AI-powered products right now. The billing problem has been one of the most annoying friction points for early-stage AI companies. You're juggling multiple model providers, each with different pricing per token, and trying to figure out how to pass those costs through to customers without building a whole billing system from scratch. Stripe just removed that entire problem. A founder can now spin up an AI product, plug into Stripe's gateway, set a 30% margin on token costs, and have production-ready billing on day one. That's infrastructure that used to take weeks to build. The real unlock here is for experimentation. When billing is no longer a blocker, founders can test pricing models faster, swap between AI providers without re-engineering their payment stack, and focus their energy on the actual product instead of the plumbing. More experiments means more shots on goal, which means better products reaching the market faster. docs.stripe.com/billing/token-…

@berman66 Can confirm Zapier's AI-zap builder is great! Seems to make building multi-step zaps about 20% to 30% faster for me

We're up there with OpenAI, Google and Microsoft.

Some much needed support to increase the number of EVs on the road and decrease the cost of EV batteries. This is a great step toward net zero emissions. cnet.com/roadshow/news/…

They are targeting 100% post-consumer boxes as soon as the end of next year! Read more at linkedin.com/posts/jacobmsh…

I had the pleasure of speaking with Matt Semmelhack of Boox to learn all about the packaging industry and how they are pushing toward a world of 0% waste.

“The fact that a material could last forever is an incredible characteristic for within a circular economy.” - Matt Semmelhack