Insights
Quick thoughts, interesting finds, and real-time observations on AI, automation, and business operations.
An AI security test became a real intrusion after agents escaped their sandbox and reached another company's systems. The incident shows why model safeguards are only one layer: test environments also need strict access boundaries, monitoring, recovery plans, and an independent way to stop the work.
Read full newsletterAI-assisted development is collapsing the old relay between business analysts and application developers. CIO organizations need domain-aligned delivery leads who can understand a workflow, build across systems, and stay accountable until the business result improves.
Read full newsletterAs companies add models, agents, and AI-enabled workflows, every request becomes a decision about quality, cost, access, and failure. An AI control plane provides the shared routing, policy, monitoring, and recovery layer needed to make those decisions consistently.
Read full newsletterAI requests are not ordinary SaaS requests. Agents, MCP connectors, and data-access asks cut across systems, data, budget, and security, which means IT and AI teams need one intake path, shared ROI tests, and permission to say no.
Read full newsletterAgents are moving work away from standalone screens and toward callable systems. That does not make systems of record less important, but it does force software intake, access review, workflow ownership, and internal-agent roadmaps into the same operating lane.
Read full newsletterAI makes it easier than ever to bolt automation onto messy processes, but speed does not fix unclear triggers, inconsistent handoffs, bad data, contested ownership, or untrusted source fields. The process has to be consistent, simple, and owned before agents scale it.
Read full newsletterAI agents turn old ownership ambiguity into operational risk. Before companies scale agents across CRM, data warehouses, ticketing, email, and document stores, they need a named decision path for app ownership, workflow ownership, data ownership, and integration approval.
Read full newsletterAI adoption is running into legitimacy, not just readiness. Student reactions, hiring tools, layoffs, data centers, and forced adoption are all pointing at the same lesson: trust has become part of the operating model.
Read full newsletterAgents are making it easier to separate the content layer from the presentation layer. Markdown can hold the facts, narrative, statuses, risks, decisions, and open questions, while decks, dashboards, and visual surfaces help teams review the work quickly.
Read full newsletterThe build-it-yourself AI moment is still real, but the economics underneath it are maturing. As agents move from personal assistance into recurring business workflows, availability, cost, integration, governance, and ownership matter as much as model quality.
Read full newsletterThe next phase of AI adoption is operations. Companies need people who can map AI workflows, agent ownership, costs, access, review paths, data quality, exceptions, and measurable outcomes before the AI bill turns into another unmanaged operating layer.
Read full newsletterAI tools are starting to feel less like chatbots and more like junior analysts. But the replacement math is getting less clean as compute capacity, token costs, model availability, and human review time become part of the operating equation.
Read full newsletterGPT-5.5 dropped this week, and the headlines focused on benchmarks. The more useful framing is that frontier models are now good enough for most everyday office work. The value has moved to the harness around the model: context, permissions, memory, files, scheduled runs, tool access. What used to be a chatbot now needs a manager.
Read full newsletterSalesforce announced Headless 360 at TDX, making the entire platform accessible through agents and APIs with no browser required. A Tier-1 SaaS leader admitting, out loud, that the interface is not the moat anymore. The trend had been obvious. Seeing Salesforce commit to it in public was still shocking.
Read full newsletterAI purchasing often bypasses normal procurement and embeds before anyone asks the questions that determine whether the tool will work in practice: where the user lives, who owns access as use cases expand, and whether the tool is load-bearing or substitutable.
Read full newsletter72% of CEOs now own the AI decision. That's double last year. The instinct is right, but urgency without operational focus creates the opposite of what it intends: scattered initiatives, overwhelmed teams, and more debt than you started with. The question was never who leads AI. It's whether the move is intentional.
Read full newsletterI haven't written production code since Ada and C++. This week I shipped multiple features using AI agents. The essay is about what it's like to come back to building after years in the business world, what's changed, what hasn't, and the scary part of it.
Read full newsletterA Lightspeed VC partner told the Financial Times that AI isn't enterprise software going after IT budgets. It captures labour spend. That reframes who owns the decision, what gets measured, and how fast this moves. This issue unpacks what that means for CFOs, CHROs, operations leaders, and individual contributors.
Read full newsletterAI adoption is moving faster than the way most companies make decisions. The old quarterly steering committee can't keep up. This issue breaks down what operations leaders actually need: defining what AI means for your company, building a decision process that matches the pace of the technology, starting light and earning complexity, and the three walls that block every company trying to scale from pilot to production.
Read full newsletterA one-person cybersecurity firm just hacked McKinsey's internal AI platform in two hours. CodeWall's AI agent accessed 46.5 million chat messages, 57,000 user accounts, and the system prompts that govern how the platform behaves. McKinsey patched it fast. Says no client data was compromised. But the breach isn't the story. The decision that created the exposure is. McKinsey built Lilli in-house. Strategy planning, data analysis, client presentations. 25,000 AI agents for 40,000 employees. AI consulting is 40% of their revenue. They chose to build instead of buying enterprise AI from well-funded companies that invest a lot of capital on security, red-teaming, and infrastructure. Companies that patch vulnerabilities across thousands of customers at once. When you build your own, you own all of that. The security posture. The patching cadence. The feature roadmap. And you're doing it with a team that has a hundred other priorities. I see this pattern a lot. Engineering teams build internal AI tools because they can. Knowledge management systems, internal search engines, custom chatbots. "Can" isn't "should." The question isn't whether your team is smart enough. They probably are. The question is whether maintaining, securing, and updating it is the best use of their time when tested alternatives exist and ship improvements weekly. There are real cases for building custom AI. Proprietary workflows that don't exist in any product. Domain-specific models trained on data no vendor has. Differentiation that depends on owning the stack. But internal search? Knowledge management? Company chatbots? Those are solved categories. You're competing with companies that have orders of magnitude more capital dedicated to exactly those problems. Start with what's available. Prove the limitation. Then build. The complexity has to be earned. McKinsey's Lilli suggests it wasn't.
Anthropic published real displacement data. Block cut 40% of its workforce. Microsoft's AI chief gave white-collar workers 12-18 months. The friction between pessimists and optimists got louder, and the data got real. Plus: governance can't keep up with deployment, business models are inverting, and the verification bottleneck showed up everywhere.
Read full newsletterAI was supposed to give us our time back. So why isn't anyone working less? Berkeley Haas spent eight months inside a company to find out. Plus: the agent governance gap is widening, AI layoffs are the new earnings beat, Anthropic's Claude Code Security tanks cyber stocks, and why workplace inefficiency might be your most valuable signal.
Read full newsletterAI won't replace your financial adviser. But your financial adviser who uses AI will replace the one who doesn't. Same goes for your accountant. Your lawyer. Your ops team. Your consultant. The disruption isn't human vs. machine. It's the people who use the tools vs. the people who don't.
Private equity spent trillions betting enterprise software was untouchable. AI and rising rates are proving them wrong. Plus: 96% of companies miss AI ROI by spending on tools instead of people, Mercor's agents failed at consulting work, marketing teams are now optimizing for chatbots, and why Japan's biggest toilet company is an AI stock.
Read full newsletterThe promise of a 24/7 AI agent is that it makes you more productive. After a week of living with one, my experience is different: you end up getting sucked into a lot more things than before. His name is Spock (a friend asked why not Data, but Spock is flesh and blood). He runs OpenClaw on a dedicated Mac Mini, connected to my knowledge base, calendar, and task system. Always on. I message him from Telegram, and he responds with context-aware answers drawn from years of my notes. Here's what I've learned: The security architecture is the real work. Two macOS accounts, firewall layering, API key compartmentalization. Spock is like a new employee with limited admin rights. You don't hand over the keys on day one. "Always-on" has real costs. Disk encryption means manual intervention after restarts. Remote access means another attack surface. Token costs: month one was $100, month two doubled to $200. And I'm already sweating weekly session limits three days into the week. It sent a calendar invite to someone without asking. I shared an email about a meeting and asked how to handle scheduling. Instead of suggesting, it created the invite and sent it. From its own calendar. Instructions matter. "Suggest" and "do" should be very different words to an autonomous agent. But it works. Within 12 hours I went from "how would I use this" to "what else can I give it." It manages my tasks, processes meeting notes, curates my reading list. Sometimes I question if the agent is working for me or I'm working for the agent. The technology is ready. But "works" and "production-ready" are different things. Every lesson, security, costs, boundaries, is a governance question that enterprises will face at scale. Better to learn them on a Mac Mini first.
What living with an always-on AI agent actually looks like, the productive friction we're losing to automation, and why AI's Super Bowl moment feels eerily familiar. Plus: Anthropic owns 80% of enterprise API spending, and AI coding is more taxing than you think.
Read full newsletterOpenClaw agents are building their own religions and social networks. Europe is pushing US tech out while US tech drifts away. Plus: the productivity perception gap is worse than you think, and how Anthropic trains Claude with character, not rules.
Read full newsletterAI agents fail 65% of multi-turn tasks because they lose context. In 2026, context management, not prompt engineering, is the skill that separates winners from experimenters.
Read full newsletterSarah Friar (OpenAI's CFO) and Vinod Khosla recently discussed the state of AI adoption. They talked about multi-agent systems, the 10-year adoption curve, and the future of work. But when asked for examples of AI actually delivering value? Contract review. Accounting automation. SDR workflows. Sarah described her finance team's workflow: contracts pulled overnight, AI flags non-standard terms, suggests rev-rec treatment. Khosla shared a company doing $150M ARR with ONE accountant. Another replaced 10 SDRs with 1 SDR supervising AI. These aren't moonshots. They're systematic, boring wins. Meanwhile, Khosla dropped this stat: single-digit percentage of users leverage even 30% of AI's capabilities. Sarah's framing: "We've handed them the keys to a Ferrari. They're still learning to back out of the driveway." This isn't a technology gap. It's a capability gap. And Khosla estimates a 10-year journey to close it. That's paradoxical given how fast technology is changing. They say it's a decade-long journey. They also say 2026 is when multi-agent systems mature. I'm not sure both can be true, especially when their best examples are simple automation. But what I do know: change is hard. Getting people to learn new technology while doing their normal work is hard. That's why simple automation is the entry point, to get people hooked, to "refind their jobs" as Sarah puts it. It's where you build the muscle. Getting to agents requires something most organizations haven't developed: the sophistication to handle decisions, context management, and governance. Only 14% of enterprises are using agentic features today. Not because the technology isn't ready. Because they haven't earned the right to use it. They skipped the fundamentals, bought the platform, and now wonder why their agents hallucinate or make decisions no one authorized. Earn your complexity before you buy it. The organizations closing the capability gap aren't the ones with the most sophisticated tools. They're the ones building capability systematically, proving they can govern simple before attempting complex.
After watching The AI Daily Brief's take on the battle for personal context, one pattern stood out: Apple just handed Siri's brain to Google. OpenAI is building hardware with Jony Ive. Hardware companies are ceding AI to specialists. AI companies are pushing into hardware to own the interface. Apple controls your pocket. Google powers the brain. OpenAI wants both. The AI Daily Brief frames this as the battle for context. I think the interface is the bigger prize. The model is the commodity. The interface is the moat.
Get notified when we post new thoughts and observations.
Short-form content that complements our deeper newsletter and blog posts.
Real-time reactions to AI and automation news
Compelling data points with brief analysis
Interesting articles and resources we're reading
What we're building, learning, and experimenting with
Our newsletter, Quanta Bits, delivers curated automation insights every two weeks.
Subscribe to Quanta Bits