AI Bulletin · week of May 11–15, 2026

AI Bulletin · week of May 11–15, 2026

Recap of the week of May 11–15, 2026 in AI. Six news items and a framework on how AI is reshaping the software P&L.


Recap of the week of May 11–15, 2026 in AI. Six news items and a framework on how AI is reshaping the software P&L.


1. Cisco announces 4,000 layoffs framed as a reorganisation toward AI and cybersecurity

Cisco has announced a reduction of around 4,000 employees, roughly 5% of its workforce, despite reporting better-than-expected quarterly results. The company says the move will let it redirect spending toward AI and cybersecurity.

It is the second case in two weeks, after Cloudflare’s announcement of 1,100 layoffs framed as an “agentic AI-first” restructuring, in which a publicly listed technology company has tied a workforce reduction to a redistribution of spending toward AI right after posting record quarterly results. A separate analysis published by CIO argues that there is no correlation between AI-attributed layoffs and the measurable ROI of AI or automation projects.

Sources: TechCrunch — “Cisco cuts nearly 4,000 jobs to spend more on AI, reports record quarterly revenue”, 14 May 2026; CIO — “AI-driven layoffs aren’t making business sense”.


The bulletin's mascot taking notes in front of a board of corporate logos while two human hands cross figures off a roster sheet

Cisco announces 4,000 layoffs as part of a reorganisation toward AI and cybersecurity.

2. SAP unveils “Autonomous Enterprise” and blocks external AI agents from its data

At Sapphire 2026, SAP unveiled a framework it calls “Autonomous Enterprise”: specialised AI agents that execute end-to-end processes in finance, supply chain and HR, supported by a new “Knowledge Graph” intended to give those agents business context. The company has announced a €100 million partner fund and integrations with NVIDIA and Microsoft.

In parallel, its 2026 API policy prohibits external AI agents from accessing SAP data directly, forcing reliance on its Joule assistant. According to reports, only 3% of SAP customers use Joule in production. ServiceNow and Salesforce have instead adopted headless architectures with open data access for external agents: this week ServiceNow introduced Action Fabric, an architecture that allows third-party agents to execute workflows directly on the platform through MCP servers, with audit and permission controls.

Sources: SAP News — “SAP unveils Autonomous Enterprise”, 13 May 2026; report via TLDR IT, 15 May 2026, on SAP’s 2026 API policy; Techzine — “The UI is dead, long live the agent: ServiceNow goes headless and opens its platform”.


3. Anthropic overtakes OpenAI in business customers, according to Ramp data

According to Ramp data cited this week, in April Anthropic surpassed OpenAI for the first time in number of businesses using its models. Anthropic’s business adoption has quadrupled over twelve months, while OpenAI’s grew by 0.3% over the same period.

A separate analysis published by SaaStr on the same dataset adds another cut: OpenAI’s share is still the largest (56%), but it has dropped 8 points over the year; Claude has doubled adoption (+128%) and Gemini has gone from 27% to 40%. The coding assistant segment is one of the main drivers of the shift.

In parallel, Anthropic launched Claude for Small Business, a package with prebuilt connectors to QuickBooks, PayPal, HubSpot, Google Workspace and Microsoft 365, and announced that its paid plans will begin applying metered pricing to Claude usage via API and agents, with a dedicated monthly credit included starting 15 June.

Sources: TechCrunch — “Anthropic now has more business customers than OpenAI, according to Ramp data”; SaaStr — “Who’s winning enterprise AI now: Claude up 128%, Gemini up 48%, OpenAI down 8%, Grok still a rounding error”; Anthropic — “Claude for Small Business”; InfoWorld — “Anthropic puts Claude agents on a meter across its subscriptions”.


The bulletin's mascot reading a long telex strip filled with adoption bars going up and down, with three generic logos in the background

Anthropic overtakes OpenAI in business customers using its models, according to Ramp data.

4. Cerebras debuts on the market with the year’s biggest IPO: $5.55 billion

AI chipmaker Cerebras has closed its US IPO with $5.55 billion raised and a market valuation of around $40 billion. The offering, led by Morgan Stanley, Citigroup, Barclays and UBS, drew orders for more than 20 times the number of shares available.

It is the largest IPO of 2026 to date, according to published data.

Source: report via TLDR AI, 14 May 2026 — “AI chipmaker Cerebras raises $5.55 billion in year’s biggest IPO”.


Bloomberg reported that OpenAI has hired outside counsel and is exploring legal action against Apple over the performance of the two-year-old agreement to integrate ChatGPT into iOS, iPadOS and macOS. OpenAI argues the integration has not produced the expected subscriber growth and that Apple “has not made an honest effort” within the partnership. Apple is set to open its platform to other AI providers this autumn.

In parallel, it has been reported that Microsoft signed an amendment to its OpenAI agreement at the end of April: it removed exclusivity over OpenAI’s models (which can now be sold on any cloud), dropped the AGI clause that would have altered Microsoft’s IP rights, and retained its licence through 2032 with a 27% stake valued at around $135 billion. According to the report, Microsoft is in talks to acquire Inception, a startup that builds diffusion-based language models.

Sources: TechCrunch — “OpenAI is reportedly preparing legal action against Apple, it wouldn’t be the first partner to feel burned”, 14 May 2026; The Next Web — “Microsoft is quietly shopping for an OpenAI replacement”.


The bulletin's mascot seen from behind looking at two towers of contract folders on a desk, a law-firm stamp on top of one of the piles

OpenAI explores legal action against Apple and Microsoft amends the terms of its OpenAI agreement.

6. Anthropic signs a seven-year, $1.8 billion compute deal with Akamai

Anthropic has closed a seven-year, $1.8 billion compute services deal with Akamai. The company has been expanding capacity following repeated user complaints about Claude usage limits.

The deal stacks on top of those announced in recent weeks with SpaceX (access to more than 220,000 NVIDIA GPUs at the Colossus 1 datacenter in Memphis), Amazon, Google, Broadcom, Microsoft, NVIDIA, Fluidstack and CoreWeave.

Source: Sherwood News — “Akamai climbs to highest level since 2000”, 11 May 2026.


7. Framework of the week: AI-attributed ROI, the software P&L and organisational readability

Four pieces that circulated this week converge on the same theme: why reported AI adoption is not yet translating into comparable economic shifts.

  • CIO publishes an analysis arguing that AI-attributed layoffs show no correlation with the ROI of AI projects. Short-term headcount savings do not demonstrate durable economic value from AI initiatives.
  • A CIO Dive report notes that almost every enterprise is investing in AI, yet only 5% say their data is ready to support it. A separate CIO Dive report cites that 81% of engineering leaders say the time AI saves on writing code is being reinvested in reviewing and debugging what it generates, creating “invisible work” that traditional productivity metrics miss.
  • An analysis published by GPTomics argues that the typical SaaS gross margin (70%) could collapse to 17% in AI-native software, because every model call is personalised (caching and multi-tenancy contribute little), and reasoning models burn 10–100x more tokens than the models they replace.
  • A piece distributed via TLDR Founders rounds out the reading from the organisational side. The author estimates that only around 1,000 companies with more than $5M ARR have actually rebuilt themselves so that agents operate inside the core workflow rather than at its edges, because pricing rules, refund policies and sales processes still live in spreadsheets and internal notes that an agent cannot read.

Sources: CIO — “AI-driven layoffs aren’t making business sense”; CIO Dive — “Nearly every enterprise is investing in AI, but only 5% say their data is ready” and “Engineering roles shift toward managing AI”; GPTomics — “How AI changes software P&L”; analysis via TLDR Founders, 13 May 2026, on becoming an AI-native company.


Closing

The six items above cover a fresh layoff at a public company with public AI justification, a dual move by SAP (internal opening to its own agents and external closure of its API), a crossover in enterprise adoption with Anthropic overtaking OpenAI according to Ramp data, the year’s biggest IPO (Cerebras), tensions in OpenAI’s alliances with Apple and Microsoft, and an additional compute deal between Anthropic and Akamai. None of the items include public productivity data measured in specific companies, and the framework gathers four independent readings on why declared AI investment is not yet translating into comparable economic metrics.