AI in the News — Weekly Brief

Anthropic has confidentially filed for an IPO that could value the company near $1 trillion, with a Nasdaq or NYSE listing possible as early as October 2026 — annualized revenue reportedly jumped from $9B at the end of 2025 to over $44B by May 2026. OpenAI has pushed its own IPO target to 2027, leaving Anthropic positioned to be the first trillion-dollar AI company to go public.

Why it matters: a trillion-dollar public debut turns AI valuation questions from a tech-press curiosity into a boardroom conversation — a clean opener for “why now” on readiness training.

In an all-stock deal announced June 16 and expected to close in Q3 2026, SpaceX agreed to acquire Anysphere, maker of the AI coding tool Cursor, which hit $2B in annual recurring revenue by February — reportedly the fastest-growing business software company on record. The deal follows SpaceX's earlier merger with xAI and cements its move into enterprise AI.

Why it matters: capital is consolidating fast around agentic coding tools — a useful data point for clients asking whether AI coding assistants are a fad or infrastructure.

Across 1,100 senior business and technology leaders in eight countries, just 23% say their workforce is fully AI-ready, a six-point decline from 2025. 79% agree AI's speed will outpace their organization's workforce, governance, and operating models, and while AI is deployed in 57% of enterprises, only 11% are hitting their top objectives with it.

Why it matters: readiness is getting worse, not better, despite heavier investment — the strongest “the gap is widening, not closing” data point of the week.

Cognizant announced an expanded strategic partnership with Anthropic on July 27, becoming one of a small number of Global Premier Partners in the Claude Partner Network. Its CEO framed the moment bluntly: “AI capability is rising faster than enterprises can absorb it, and that gap is the defining problem of this moment.”

Why it matters: a major systems integrator naming the absorption gap as “the defining problem” is third-party validation of the exact thesis behind readiness training.

More than 80% of Citigroup employees adopted AI tools in Q1 2026, generating 42 million interactions since launch — a 50% increase since Q4 2025. More than 10,000 engineers now use agentic AI tools daily, automated code reviews have topped 1.5 million, and AI processes over 4,400 documents in Markets each month, freeing roughly 1,700 hours of oversight capacity.

Why it matters: this is what mature, at-scale enterprise adoption actually looks like — a useful benchmark for clients still arguing about pilot programs.

Across 1,550 AI decision-makers in six countries, 73% say AI is used regularly or across most business processes, yet only 10% say it's core to how the business operates. 42% say AI is capable of meeting today's needs but their organization isn't set up to capture the value, and only 20% feel fully equipped to meet their own 12–24 month scaling expectations.

Why it matters: “adoption without transformation” is the precise gap readiness training addresses — this report puts numbers on that distinction.

Employees who received AI training are 8.7 times as likely to say AI has transformed how their work gets done and 7.4 times as likely to say it lets them do their best work. Yet 47% of employees who use AI say their organization hasn't offered any training on how to use it in their job — and training without manager support falls flat too.

Why it matters: a data-backed frame for pitching structured, manager-inclusive training over a self-serve tool rollout.

47% of leaders rank upskilling existing employees in AI as their single top workforce strategy for the next 12–18 months. At the same time, 52% of senior leaders say it has become harder to find people with the right skills to advance their AI strategy — a dual squeeze of build-versus-buy pressure on talent.

Why it matters: when leaders can't hire their way out of the skills gap, internal upskilling stops being a nice-to-have and becomes the only lever left.

Mollick's practical guide, rewritten because “using AI” has shifted from chatbot conversation to handing an agent a computer and a real task. He distinguishes AI-hosted agent modes from computer-hosted modes: the former return a finished result, the latter expose the work itself. His test case — prepping an MBA seminar from Gmail — had one system send an email on his behalf while the other drafted and waited for approval. He flags prompt injection as a real, unsolved risk and closes with blunt advice: pick one, pay the $20, hand it one real task.

Why it matters: “one agent sent the email, the other asked first” is a ready-made, non-technical illustration of why approval settings and prompt-injection awareness belong in every executive's first AI session.