Journal/AI Weekly Digest/8–15 May 2026

AI Weekly Digest8–15 May 2026

Five major launches in one week: Claude Platform on AWS (GA), Claude for Small Business, Microsoft 365 Word GA and Outlook beta, Advisor Tool public beta, and Opus 4.7 fast mode. Anthropic discloses $30B+ 2026 revenue run rate. Google I/O countdown begins.

Period
8–15 May 2026
Published
May 15, 2026
Covers
Anthropic · OpenAI · Gemini · Copilot

Dateline: May 15, 2026 | Next update: May 22, 2026

The most product-dense week of 2026 so far. Five major launches landed between May 8 and 15: Claude Platform on AWS (GA), Claude for Small Business, Microsoft 365 Word and Outlook add-ins (GA/beta), the Advisor Tool in public beta, Opus 4.7 fast mode, and 20+ legal MCP connectors. Anthropic also disclosed a 2026 revenue run rate above $30 billion — up from $9 billion in 2025 — and confirmed that the number of companies spending $1M+ annually has more than doubled in two months.


Claude / Anthropic

Claude Platform on AWS — generally available

Launch: May 11–12, 2026 | Availability: AWS customers globally | Regions: 18 at launch

Claude Platform on AWS is now generally available — the first time AWS customers can access the full native Claude API experience directly through their existing AWS account, credentials, and billing. No separate Anthropic contracts, no extra API keys. AWS is the first cloud provider to offer this. It is distinct from Claude on Amazon Bedrock: Bedrock keeps AWS as the data processor; Claude Platform on AWS is operated by Anthropic, with data processed outside the AWS boundary.

★ What's new

Full native Claude API through AWS — IAM authentication, AWS Marketplace billing, and CloudTrail audit logging. Includes everything on the native Claude API: Messages API, Files API, Message Batches API, Claude Managed Agents, Agent Skills, Advisor tool, code execution, web search, web fetch, prompt caching, citations, MCP connector, and Claude Console access. Models: Opus 4.7, Sonnet 4.6, Haiku 4.5. New features ship on the same day they go live on the native Claude API. Available in 18 regions across US, Canada, South America, Europe, and Asia Pacific.

Technical details

Auth: AWS IAM credentials | Billing: AWS Marketplace (retires existing AWS commitments) | Audit: AWS CloudTrail | Base URL: https://aws-external-anthropic.<REGION>.api.aws | Workspace ID: ANTHROPIC_WORKSPACE_ID env var | Note: existing Bedrock private offer holders must contact AE before switching — discounts cannot be applied retroactively

Best for: AWS-native engineering teams wanting the full Claude API feature set without a separate Anthropic vendor relationship

Claude for Small Business — new product launch

Launch: ~May 13, 2026 | Availability: Pro and above | Toggle install inside Claude Cowork

Anthropic launched Claude for Small Business, a toggle install inside Claude Cowork that connects Claude to the tools small businesses already use — QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365 — with 15 ready-to-run agent workflows and 15 skills built for common small-business tasks. It ships with payroll planning, month-end close, invoice management, lead triage, campaign attribution, contract signing, and more. Nothing sends, posts, or pays without owner approval.

★ What's new

15 agent workflows across finance, operations, sales, marketing, HR, and customer service. 15 skills targeting the repeatable tasks owners said slow them down most. Integrated tools: Intuit QuickBooks (payroll, cash-flow, close, tax prep), PayPal (settlements, invoicing, disputes), HubSpot (lead triage, customer pulse, campaign attribution), Canva (content creation and publishing), Docusign (contract send, status tracking, filing). Anthropic is also partnering with Workday Foundation and LISC to provide Claude credits and AI curriculum to an initial cohort of 15 solopreneurs.

Technical details

Access: toggle inside Cowork | Plans: Pro and above | Workflows: 15 pre-built | Skills: 15 pre-built | Human-in-the-loop: approval required before any action sends, posts, or pays | Partners: Intuit, PayPal, HubSpot, Canva, Docusign, Google, Microsoft | Solopreneur program: Workday Foundation + LISC

Best for: Small business owners, solopreneurs, freelancers managing finance, sales, and operations

Microsoft 365 — Word GA, Outlook public beta

Launch: this week | Availability: all paid plans

The Microsoft 365 Claude add-in family is now complete for everyday office workers. Claude for Word launches in general availability, and Claude for Outlook enters public beta — joining the already-live Excel and PowerPoint add-ins. All four add-ins share full conversation context across apps, so Claude's actions in one application are informed by everything that happened in the others.

★ What's new

Claude for Word is now generally available — draft, edit, reformat, and summarise documents directly in Word. Claude for Outlook is now in public beta — inbox triage, draft replies, meeting prep summaries, and follow-up actions inside Outlook. All four M365 add-ins (Excel, PowerPoint, Word, Outlook) share full cross-app conversation context and support LLM gateway connections (Amazon Bedrock, Google Vertex AI, Microsoft Foundry).

Technical details

GA: Excel, PowerPoint, Word | Public beta: Outlook | Cross-app context: shared across all four add-ins | Gateway support: Amazon Bedrock, Google Vertex AI, Microsoft Foundry | Plans: Team and Enterprise for admin-managed deployment; individual paid plans for personal use

Best for: Office workers, legal and finance teams, knowledge workers in Microsoft 365 environments

Advisor Tool — now in public beta

Platform: Claude API | Beta header: anthropic-beta: advisor-tool-2026-03-01 | Availability: all API users

The Advisor tool moves from Claude Code-only to a full API public beta this week. It lets you pair a fast, cheaper executor model (Sonnet 4.6) with a higher-intelligence advisor model (Opus 4.6 or 4.7) that provides strategic guidance mid-generation. The result: long-horizon agentic workloads get close to advisor-solo quality, while the bulk of token generation happens at executor-model cost.

★ What's new

Advisor tool in public beta on the Claude API. Benchmark gains: Sonnet 4.6 + Opus advisor scores 50.5% on SWE-bench Multilingual vs 43.0% for Sonnet 4.6 solo — a 17% relative improvement. BrowseComp also improves substantially. The advisor runs alongside your existing tools in the same Messages API loop — web search, code execution, and advisor calls can all coexist in one agent.

Technical details

Beta header: anthropic-beta: advisor-tool-2026-03-01 | Tool name: advisor_20260301 | Recommended pairing: Sonnet 4.6 (executor) + Opus 4.6/4.7 (advisor) | SWE-bench Multilingual: 43.0% Sonnet solo → 50.5% Sonnet + Advisor | Works alongside web search, code execution, and other tools in the same loop

Best for: Developers running long-horizon agentic workloads who want near-Opus quality at below-Opus cost

Claude Opus 4.7 Fast Mode — research preview

Platform: Claude API | Beta header: fast-mode-2026-02-01 | Availability: waitlist

Fast mode, previously available only for Opus 4.6, now supports Opus 4.7. It delivers significantly faster output token generation at premium pricing for latency-sensitive workloads that still need the full Opus 4.7 capability level.

★ What's new

Opus 4.7 fast mode in research preview. Same pricing, rate limits, and access as Opus 4.6 fast mode. Join the waitlist to access. Useful for real-time agentic pipelines where Opus 4.7 quality is required but streaming latency matters.

Technical details

Model: claude-opus-4-7 | Parameter: speed: "fast" | Beta header: fast-mode-2026-02-01 | Pricing: premium (same tier as Opus 4.6 fast mode) | Access: waitlist only

Best for: Latency-sensitive enterprise pipelines requiring full Opus 4.7 intelligence

Legal MCP connectors — 20+ new tools and 12 plugins

Platform: Claude Cowork + Claude API | Availability: Pro and above

Legal professionals have become the most engaged Cowork users of any knowledge-work function, and Anthropic responded this week with the largest single expansion of legal tooling to date: 20+ new MCP connectors covering legal research, contract management, e-discovery, matter management, and legal aid — plus 12 practice-area plugins for deeper workflow support.

★ What's new

20+ legal MCP connectors spanning: legal research platforms, contract lifecycle management, e-discovery tools, matter and case management systems, and legal aid services. 12 new practice-area plugins for more opinionated, repeatable legal workflows. Connectors are open and customisable for law firms building their own integrations.

Technical details

Connector categories: legal research, contract management, e-discovery, matter management, legal aid | Plugins: 12 practice-area specific | Customisation: open architecture for firm-built connectors | Available on: Cowork and Claude API

Best for: Law firms, in-house legal teams, legal aid organisations, compliance teams

Claude Code — agent view, session tools, and fixes

Platform: terminal / VS Code / web / mobile | Availability: all plans

Claude Code shipped several point releases this week alongside the broader platform announcements, adding new session management tools, improving auto mode transparency, and fixing a range of stability issues.

★ What's new

Agent view now in research preview (run claude agents to opt in) — a dashboard for managing background and parallel agent sessions. New: 'Summarize up to here' in the Rewind menu to compress earlier context while keeping recent turns intact. ANTHROPIC_WORKSPACE_ID env var added for workload identity federation scoped to a specific workspace. claude agents --cwd <path> scopes the session list to a directory. Auto mode permission dialogs now explain when a permissions.ask rule caused the prompt. Background agents launched via /bg now preserve the current permission mode. CLAUDE_CODE_PLUGIN_PREFER_HTTPS added for GitHub plugin installs over HTTPS instead of SSH.

Technical details

Agent view: research preview, opt-in via claude agents | ANTHROPIC_WORKSPACE_ID: workload identity federation scope | CLAUDE_CODE_PLUGIN_PREFER_HTTPS: HTTPS GitHub plugin clone | Fixed: spurious stream idle timeout 5 min post-response | Fixed: /model picker default row not reflecting ANTHROPIC_DEFAULT_OPUS_MODEL overrides | Fixed: background jobs on custom base URL not getting auto-named | Fixed: /model in one session silently changing autocompact threshold in other concurrent sessions | Fixed: crash loop when piping >10MB input to claude -p via stdin

Best for: Developers managing parallel agent sessions, enterprise teams using workload identity federation

Business metrics — $30B+ revenue run rate disclosed

Disclosed: May 2026 | Source: Anthropic briefings and press coverage

Anthropic disclosed its 2026 annual revenue run rate has climbed above $30 billion, up from $9 billion in 2025. The number of companies spending more than $1 million per year with Anthropic has more than doubled in two months — from 500 to over 1,000. CEO Dario Amodei separately warned at Anthropic's Financial Services briefing that SaaS companies that do not adapt to the AI shift risk losing significant market value or going bankrupt.

★ What's new

2026 revenue run rate: above $30 billion (vs $9B in 2025). $1M+ annual spend customers: more than 1,000 (up from 500 in approximately two months). These figures were disclosed informally at industry events and have since been reported by major press outlets.

Technical details

Figures: informal disclosure, not audited | Context: Anthropic reportedly in pre-IPO discussions per CoinDesk and others | SaaS market impact: Salesforce, ServiceNow, Intuit, Docusign, Box all down year-to-date amid AI substitution concerns

Best for: Investors, enterprise buyers evaluating Anthropic's trajectory, competitive intelligence

Plans and Pricing

No changes to API or consumer plan pricing this week. Claude Platform on AWS is available via AWS Marketplace on a consumption basis; contact your Anthropic or AWS account executive for private offer pricing.

Technical details

Opus 4.7: $5/$25 per MTok | Sonnet 4.6: $3/$15 per MTok | Haiku 4.5: low-cost tier | Claude Platform on AWS: consumption via AWS Marketplace | Claude for Small Business: included in Pro and above | Opus 4.7 Fast Mode: premium pricing, waitlist

Best for: No action needed on pricing — unless setting up Claude Platform on AWS (contact AE)


ChatGPT / OpenAI

Dateline: May 15, 2026 | Next update: May 22, 2026

Over the past week, OpenAI has continued refining system reliability and agent behaviour, while quietly strengthening enterprise-readiness and workflow integration. This is another consolidation week — fewer flashy updates, more quiet improvements that make the system dependable enough to integrate into real workflows.

GPT-5.3 Standard — default model

Release: late 2025 | Pricing: included | Availability: all users

★ What's new

Improved consistency in instruction-following across extended interactions.

Technical details

Context ~128k | Output ~4k–8k | Improved prompt adherence tuning

Best for: General use

GPT-5.3 Pro — high-reasoning model

Release: late 2025 | Pricing: Pro | Availability: Pro/Enterprise

★ What's new

Reduced logical breakdowns in multi-step analytical outputs.

Technical details

Context ~200k (est.) | Improved reasoning continuity

Best for: Deep analysis

GPT-5.3 Mini — fallback model

Release: late 2025 | Pricing: low-cost | Availability: all

★ What's new

Further reduction in visible quality gaps during fallback.

Technical details

Context ~64k | Refined model-switching logic

Best for: Quick tasks

Agent Mode

★ What's new

Improved ability to execute structured, repeatable workflows with fewer corrections.

Technical details

Better task decomposition | Improved execution reliability

Best for: Task delegation

Deep Research

★ What's new

Improved structuring of long-form research outputs.

Technical details

Enhanced formatting + synthesis pipeline

Best for: Research

Memory & Projects

★ What's new

Improved prioritisation of task-critical information over general memory.

Technical details

Refined weighting toward active-task relevance

Best for: Ongoing workflows

Advanced Voice Mode

★ What's new

More natural conversational rhythm and pacing.

Technical details

Improved response timing + conversational smoothing

Best for: Voice interaction

ChatGPT for Clinicians

★ What's new

Improved usability in drafting structured clinical notes and summaries.

Technical details

Further domain tuning | Reinforced safety constraints

Best for: Clinical support (non-diagnostic assistance)

What this means

This week reinforces a now very clear trajectory: OpenAI is prioritising reliability, workflow execution, and real-world usability over headline innovation.

Three things stand out. First, instruction-following and consistency improvements are becoming central — a shift from "what the model can do" to whether it can do it reliably every time, a necessary step for enterprise adoption. Second, Agent Mode is steadily evolving into a practical tool for structured work, not just experimental automation. Third, the continued iteration on ChatGPT for clinicians confirms that OpenAI is serious about embedding its models in high-trust, high-risk environments.

Overall, this is another consolidation week — fewer flashy updates, more quiet improvements that make the system dependable enough to integrate into real workflows.

Plans and Pricing

No changes this week.

Technical details

Plans unchanged | API pricing stable


Gemini

Date: May 15, 2026 | Next update: May 22, 2026 (Post-I/O Special)

This week serves as the final preparation window for Google I/O 2026. While core model updates have focused on developer accessibility and refining multimodal capabilities, the product ecosystem is shifting toward "Personal Intelligence" and deeper integration into daily workflows.

Gemini 3.1 Flash-Lite — migration deadline May 25

GA: last week | Deprecation of preview: May 11, 2026 | Final shutdown: May 25, 2026

Following its GA launch last week, the legacy gemini-3.1-flash-lite-preview model was officially deprecated on May 11. Developers must switch to the production version before the final shutdown on May 25.

★ What's new

Preview model deprecated May 11. Final shutdown: May 25, 2026. Migrate to the production gemini-3.1-flash-lite before this date.

Best for: Any team still on the preview version — action required before May 25

Enhanced Gemini 2.5 Flash

★ What's new

Updated version now fully rolled out in the Gemini app. Superior image understanding — particularly for diagrams and handwritten notes — and improved output formatting using headers and tables.

Best for: Document analysis, structured output, diagram-heavy workflows

Multimodal File Search — API update

API update: May 5, 2026

★ What's new

Native multimodal search now supported. Use gemini-embedding-2 to search through images within documents. Grounding metadata now provides specific page numbers and visual citations.

Technical details

Model: gemini-embedding-2 | Grounding: page-level citations for images | Use case: searching visual content within PDFs and documents

Best for: Developers building document search with visual content

"Memories" rebrand and switching tools

★ What's new

"Past chats" officially rebranded as Memories. New suite of switching tools in Settings allows users to import chat history and personal context from other AI providers by uploading a ZIP file or using a specific import prompt.

Best for: Users migrating from other AI platforms

Gemini in Google Chat — expanded language support

Update: May 14, 2026

★ What's new

Expanded language support for message refining — users can now polish drafts in French, German, Italian, Japanese, Korean, Portuguese, and Spanish.

Best for: Global teams using Google Chat

Guided Learning Mode — now open to all

★ What's new

Now available to all users via the "Learn" (mobile) or "Guided Learning" (desktop) chips. Moves beyond direct answers to provide interactive, step-by-step concept breakdowns.

Best for: Learning, training, and education workflows

AlphaEvolve — Gemini-powered coding agent

★ What's new

Google DeepMind highlighted AlphaEvolve, a Gemini-powered coding agent designed to scale scientific impact through automated research and code generation.

Best for: Research teams, scientific computing

Google I/O 2026 — keynote May 19

Google I/O 2026 begins May 19. Expect a major focus on "Agentic AI" — assistants capable of navigating computers and performing complex tasks with minimal supervision. Android 17 Beta 4 was showcased May 12, featuring "App Bubbles" for enhanced multitasking, setting the stage for deep Gemini-Android integration.

Action required: A breaking change for the Interactions API is scheduled for May 26, where the request/response schema will shift from outputs to steps.


Microsoft Copilot

Dateline: May 15, 2026 | Next update: May 22, 2026

The biggest story of the week is employee experience and productivity: Microsoft rolled out Copilot-powered survey insights in Viva Glint, added contextual nudges in Edge, and improved enterprise readiness tools. These updates focus on helping organisations interpret data faster and streamline workflows.

Edge for Business — contextual nudges for page summaries

Announced: May 9, 2026 | Effective: immediately | Applies to: Microsoft Edge for Business

Microsoft Edge now surfaces contextual nudges that prompt users to summarize webpages directly with Copilot Chat, reducing time spent scanning long documents.

★ What's new

Contextual nudges in Edge for Business. Summarize any open webpage via Copilot Chat. Summarization + follow-up Q&A supported.

Technical details

Roadmap ID: 515167 | Copilot Chat integration | Nudges attached to Copilot entry point

Best for: Knowledge workers, researchers, teams consuming large volumes of web content

Viva Glint — multilingual AI survey insights

Announced: May 10, 2026 | Availability: General

Copilot in Viva Glint now generates AI-powered summaries of employee survey results directly in Team and Executive Summary reports, with multilingual support for global organisations.

★ What's new

Multilingual support added — summaries now available in multiple languages. AI-generated summaries include score changes, benchmarks, and response confidence.

Technical details

Roadmap ID: 558111 | AI-generated summaries | Benchmarks + response confidence | Multilingual expansion

Best for: HR leaders, global enterprises, managers interpreting survey data

Viva Insights — expanded report sharing

★ What's new

Expanded sharing permissions for Copilot-generated reports across teams, making organisational insights more accessible.

Technical details

Integration with Microsoft 365 compliance | Role-based access | Secure sharing across departments

Best for: Enterprise leaders, managers, compliance-focused organisations

Copilot Dashboard — readiness guidance improvements

★ What's new

Improved readiness guidance in Copilot Dashboard. Admins receive tailored recommendations on licensing, compliance, and rollout sequencing.

Technical details

Dashboard enhancements | Licensing + compliance checks | Rollout sequencing recommendations

Best for: IT admins, enterprise deployment teams

Plans and Pricing

No pricing changes this week. Updates focus on feature expansion rather than cost.

Technical details

Edge contextual nudges: included in Microsoft Edge for Business | Viva Glint multilingual summaries: included in Viva suite | Dashboard guidance: enterprise only

Best for: All subscribers — no action needed


Filed under: AI Weekly Digest
First published: May 15, 2026

← Previous issue4–8 May 2026All issuesNext issue →15–22 May 2026