Market · Weekly Roundup
In the week to 11 July 2026, OpenAI and Anthropic spent their energy on the connective tissue around the models rather than the models themselves. The new pages cluster around three moves - a sweeping build-out of plugins and connectors, a push to put agentic work on every surface through Cowork, and a steady stack of enterprise proof and governance - and together they point at one direction of travel: the assistants are being wired into the tools businesses already run.
TL;DR
- Integrations move to the centre: OpenAI published a directory of more than seventy named business plugins - from Salesforce and Slack to Snowflake, Databricks, Figma and Stripe - while Claude widened its own marketplace and connectors.
- Cowork goes everywhere: new pages take Cowork to web and mobile and show it running real reporting and campaign work, alongside deeper Claude Code and agent SDK surfaces.
- Enterprise proof stacks up: fresh customer stories from Deutsche Telekom, LG CNS, Thomson Reuters and others, plus new admin controls over usage and spend.
- The models still ship: GPT-5.6 arrived and became a preferred model in Microsoft 365 Copilot, and GPT Live opened up - but the centre of gravity sat in the platform around them.
The flagship models kept coming, yet the story of the week to 11 July 2026 was the scaffolding. The clearest signal is the run of pages devoted to integrations, agentic workspaces and enterprise proof - the machinery that turns a capable model into something a company can plug in, deploy and defend to its board. Below is where the activity clustered across OpenAI and Anthropic. For the longer arc, our roundup from the week before and the one before that track the same build-out gathering pace.
The three things the labs built this week
Net-new pages published this week, by theme
Bars are relative to the largest theme. Counts reflect new pages observed across the leading AI platforms in the week to 11 July 2026.
1. Integrations move to the centre
The largest push of the week, by a wide margin, was integrations. OpenAI published a full business plugins directory - more than seventy named connectors in a single sweep. The list reads like a census of the modern work stack: Salesforce, Slack, HubSpot and Intercom for go-to-market; Snowflake, Databricks, BigQuery and Hex for data; Figma, Canva and Adobe for design; GitHub, GitLab, Linear, Replit, Vercel and Netlify for engineering; Notion, Asana, Monday, ClickUp and Atlassian for planning; Stripe, Ramp and Gusto for finance and operations. Sitting above them came a new plugins gateway, a Dropbox partner page and an operations solutions surface.
Anthropic pushed from the same direction. Claude added new marketplace entries for GitLab, Harvey, Lovable, Replit and Snowflake, a Lattice connector and a fresh plugin surface - the plumbing that lets the assistant reach the systems a business already runs on. You can see how the picture builds for each platform on our OpenAI intel page and Claude intel page.
Why it matters: every new connector is a new place the model can act, and a new reason to stay. The platform with the deepest, most trusted set of integrations stops looking like a chatbot and starts looking like a control layer over the whole toolset. It is the same instinct that took Claude into Microsoft 365 Copilot - meet the enterprise on the surfaces it already trusts. If you want to understand how those pages get surfaced inside AI answers, our guide on how LLMs decide which brands to cite is the place to start.
2. Cowork goes to web and mobile
The second cluster was about how the work actually gets done. Anthropic took Cowork - its shared space for delegating and running tasks - onto web and mobile, and backed it with pages showing the tool in real use: how people are using Cowork, and how its own marketing operations team leans on it to automate reporting and build campaigns. Around it sat a deeper developer story: new Claude Code pages on choosing a model and effort level and on dynamic workflows, a guide to building agents with the Claude agent SDK, and best practice for working in large codebases.
OpenAI moved on the same ground from the other end, with a new ChatGPT for work surface and an academy strand on using ChatGPT for everyday tasks. The framing across both labs has shifted from “ask a question, get an answer” to “hand over a task, get it run.” It is the arc we traced around the Claude Sonnet 5 launch and the GPT-5.5 Instant Mini rollout - capability moving steadily toward autonomy.
Why it matters: an assistant that holds a workspace, a schedule and a task list is far stickier than one that answers on demand. Once Cowork is on the phone as well as the desktop, the assistant is present wherever the work is - and that is where retention, and lock-in, is won.
3. Enterprise proof and governance
The third cluster was evidence and control. Both labs added a run of customer stories - Deutsche Telekom and Australian Payments Plus on the OpenAI side; LG CNS, Thomson Reuters, Advantage Solutions and Pacific Community Ventures on the Claude side - the named, logo-bearing proof that enterprise buyers ask for before they commit. Alongside the wins came governance: a new Claude page on giving admins more visibility and control over usage and spend, and material on bringing Claude Code and Cowork into government.
OpenAI rounded out the same theme with pages on government and national-security partnerships, an AI-first hire for small business, and modelling the AI jobs transition - the policy and workforce conversation that increasingly travels with any enterprise sale. It is the pattern we flagged in earlier weeks, from wire up, skill up and prove it onward: the proof layer is now a permanent part of the build.
Why it matters: integrations open the door, but proof and controls are what get an assistant through procurement. Admin visibility over usage and spend is not a footnote - it is the feature that lets a CIO say yes. Expect the customer-and-controls page count to keep climbing.
New models keep shipping underneath
None of this means the frontier stood still. OpenAI shipped GPT-5.6 and, in the same breath, made it a preferred model inside Microsoft 365 Copilot - a reminder that distribution now matters as much as the benchmark. It also opened GPT Live, with an API route close behind, and published fresh work on coding evaluations and a bio bug-bounty programme. The models advance, but the energy sits in the platform, the integrations and the proof around them - the same shift we noted when GPT-4.5 was retired and when Perplexity levelled up deep research.
The developer story goes multilingual
One quieter detail is worth pulling out. Anthropic published its developer and agent content - the field guide to Claude Fable, dynamic workflows in Claude Code, building agents with the SDK, self-service data analytics and COBOL modernisation - across German, French, Italian, Japanese and Korean. Translating a technical library into five markets is not an experiment; it is a go-to-market. It says the labs expect their developer and agent audience to be global, and it means the AI answers your brand appears in will increasingly be multilingual too. Our note on Google’s llms.txt guidance covers one way to keep pace.
What it means for your AI visibility
If the labs are building outward into integrations, agentic workspaces and enterprise proof, the brands that want to appear inside those answers should build in the same direction. Here is a short checklist to work through this month:
- Be a connectable source: structure your data, pricing and documentation so an assistant - or one of its many new connectors - can reach and cite them cleanly.
- Map to the work stack: if your product sits next to Salesforce, Slack, Notion or Snowflake, make that relationship explicit on your public pages so the model can place you.
- Write for tasks, not just queries: as Cowork-style workflows spread, being the source an agent uses to complete a task matters as much as ranking for a question.
- Publish your proof: named customer stories, security and governance pages are what get an assistant - and a brand - through procurement. Make yours easy to find and cite.
- Localise where the labs localise: if a developer library ships in five languages, the answers will too; check how you are represented beyond English.
- Track the direction weekly: read the shifts against your own category using our intel dashboards and decide where to invest.
Not sure where you stand today? Start with is your brand ready for AI search, then browse the full archive on our news page.
What to watch next week
- Connector velocity: whether OpenAI keeps expanding its business plugins directory and how fast Claude matches it, connector for connector.
- Cowork depth: how quickly Cowork on web and mobile picks up named integrations, scheduled tasks and templated workflows.
- Proof cadence: which industries supply the next round of customer stories, and whether admin and governance pages keep multiplying.
- Model distribution: where GPT-5.6 and GPT Live show up next, and whether being the default inside another vendor tool becomes the real launch surface.
The takeaway for anyone competing for visibility inside these assistants: the platforms are racing to become the place work gets done - wired into every tool, present on every device, and backed by the proof a buyer needs - not just the place questions get answered. Read which integrations, agent roles and customers each lab is prioritising and you will navigate the next few months far better. Start from the reconnAI home page or talk to our team to put it to work.
Market Intelligence
Want to know where the AI platforms are heading next?
We track what the leading AI assistants are building and how brands appear inside their answers - so you can plan around the shifts before they reach your category.
Talk to our team →Frequently asked questions
What was the main direction of travel for the AI sites this week?
In the week to 11 July 2026, the new pages across OpenAI and Anthropic leaned into building outward - a sweeping directory of plugins and connectors, Cowork and agentic work spreading to web and mobile, and a steady stack of enterprise customer stories and governance controls. New models shipped underneath, but the centre of gravity sat in the platform around them.
How big was OpenAI’s plugins push?
OpenAI published a business plugins directory with more than seventy named integrations in one sweep, spanning go-to-market, data, design, engineering, planning and finance tools - from Salesforce and Slack to Snowflake, Databricks, Figma and Stripe. Anthropic widened its own marketplace and connectors alongside it.
Why does Cowork on web and mobile matter?
Putting Cowork on every surface turns the assistant from something you visit into something that is always present, holding a workspace, a schedule and a task list. That makes it far stickier - and it shifts the job of a brand from ranking for a query to being the source an agent uses to complete a task.
How should brands respond?
Make your data cleanly connectable and citable, map your product to the tools the labs are integrating, publish named customer and governance proof, write for tasks as well as queries, and localise where the labs localise. Our guide to how LLMs decide which brands to cite walks through the detail.
About reconnAI
reconnAI tracks how the major AI models represent topics and sources across ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview - across multiple regions. We monitor how those models answer and how they change over time, so you can stay ahead of shifts in the AI landscape. Talk to our team to put this intelligence to work.
Based on source data gathered by reconnAI.