Enterprise Strategy

Amplitude tripled its pull requests with the same headcount. Chief Engineering Officer Wade Chambers says the leaders who win at AI are the ones willing to go first.

EY will pay $100 million in bonuses for judgment and adaptability, a response to the entry-level work its industry automated away.

The company that proved software could sell itself is now hiring engineers to go install it.

The largest consulting firms have converged on the same AI models and the same four workstreams. Their clients' positioning is converging with them.

A payments company just paid around $7.5 billion for a startup that decides which AI model answers your prompt. Most of the coverage assumes you already know what that means. This piece doesn't.

Wade Foster opened Zapier's books on token spend. The number nobody in the industry can produce is the one on the other side of the ledger.

Stripe is paying north of $7 billion for the router behind AI's most-cited usage data. That router's impartiality was never audited, published or contractually committed.

Joseph Davis, Chief Security Advisor for Microsoft's U.S. Health & Life Sciences business, explains how enterprises are wiring AI into code review, and why the ownership question outlasts the tooling one.

Companies with mature AI governance are more than three times as likely to have agents in production. The direction of causation is arguable. The drag hypothesis isn't.

Frontier model prices fell 99% in three years and open weights are six months behind. Who's left standing when the models are worth nothing, and who gets paid?

Sequoia ranked 20 years of Hacker News hype against the companies actually founded each year. The winners were almost never in the conversation.

Investors are benchmarking AI companies against the fastest growth figures software has ever produced. Very few people will say out loud what those figures are actually made of.

In its first earnings report as a public company, SpaceX posted the fastest AI revenue growth of any major infrastructure spender and the widest gap between what it earns and what it spends.

Mark Cuban says the next great job isn't a developer. Nine billion dollars in corporate hiring says he's right, and the Census data says he's early.

OpenAI's Presence ships the entire product page of a dozen funded startups. Token prices fell 99 percent in three years, so the labs went where the contracts are.

Uber has open-sourced ADR, the system it built after realizing its security tools could see what its AI agents did but not why.

The bank's Personal Banking chief named the three moments: the house purchase, the renewal, the account opening. What nobody at RBC is publishing is whether your decision came out any better.

Two-billion-dollar seed rounds are closing before products exist. Inside the collapse of tech's most overread announcement.

Jensen Huang wants Chinese models used and Anthropic's Mythos unleashed. Every gate he opposes, we've already watched close. The fight may be happening at the wrong layer anyway.

ClickUp's CFO is publishing her AI transformation playbook while she writes it. We checked it against what everyone else is finding.

Swapping frontier models for cheap ones changes what a call costs. It does nothing about how many calls you make, and that is where the budget went.

Writer's researchers held six models constant and swapped only the orchestration layer. Cost per task fell 41%, quality held, and the model choice barely moved the bill.

In one week, Jamie Dimon predicted $1 trillion in AI spending and told young Americans to pick up a welding torch. Both statements describe the same economy.

Murati's Thinking Machines gave away a 975B model for free, and the reason why reveals where the real moat in AI now lives: your data, not your model.

Kimi K3 is being celebrated as the moment open-source AI caught the frontier. But at 2.8 trillion parameters, openness has stopped meaning access and started meaning influence.

BNY's record quarter and ten blunt words from its CEO mark the end of token consumption as the metric that defined enterprise AI.

Databricks' new raise at a $188 billion valuation is a bet that the enterprise AI wars will be won not by the best model, but by the layer that routes, governs, and pays for all of them.

The AI services venture backed by Blackstone, Hellman & Friedman, and a deep bench of global investors now has a name, a brand, and a clear mandate: turn frontier models into working systems inside real companies.

Veterans of Google, Salesforce, and other tech giants are betting that the future of customer service AI will be built by startups solving the hardest problems: production reliability, control, data, and the role of humans.

Satya Nadella breaks ranks with the frontier labs, warning that every prompt and correction quietly teaches AI providers how your business runs.

Starbucks spends $400 million a year on software, and it just told Microsoft and IBM it might not need them.

For most of the past two years, the story of AI adoption inside big tech companies has been an engineering story. Uber has now given one of the clearest public looks at the next chapter.

The cloud giant is the latest and largest name to embrace a deployment model pioneered by Palantir. The race has shifted from who has the best model to who can get it running inside your company.

James Shira oversees technology and security for one of the world's largest professional-services firms. His argument to other leaders: the hardest part of the AI transformation is the weight it places on people.

Good enterprise software was assumed to be expensive, slow, and painful to replicate, which is exactly why acquirers paid steep multiples for it. Bain & Company just put a price on that assumption.

A new post on how the company runs its own analytics doubles as a quiet argument: the thing knowledge platforms charge a premium for is becoming the floor.

Two months ago it was Meta logging keystrokes to train AI. Now a Canadian bank is tracking its financial-crime staff to lift productivity.

Meta is reining in the very behavior it spent the past year encouraging.

Two months ago, Anthropic's most powerful model was too dangerous to hand out. Today you can put it on a company credit card.

OpenAI's CEO says cost concerns went from a non-topic in January to one of the loudest complaints he hears, as firms blow through annual AI budgets in a single quarter

"At the time, my team hated me," Mike Blandina said of January cuts he framed as a "provocative" way to prove AI could replace lost headcount.

Google, OpenAI, Anthropic, and Meta are racing to embed engineers inside customer operations. The reason has less to do with the job market than with a number the labs rarely volunteer.

The most dangerous number in Bain & Co.'s new survey is not 40%. It is 44%.

DeepSeek's V4-Pro now costs $0.87 per million output tokens. Permanently.

Anthropic now leads OpenAI in U.S. business AI for the first time, and four patterns from a 1957 Walt Disney sketch explain what changed.

The race in AI has moved one layer up, from model capability to architecture, and the five companies shaping enterprise AI are running four fundamentally different bets.