Real workflows. Real results.

Q1 was about Anthropic shipping features at breakneck speed. Fourteen releases in March alone. Computer Use, Dispatch, Projects, Office add-ins, 1M context, scheduled tasks, connectors, a marketplace. The pace was extraordinary. It was also, for most operators, impossible to keep up with.

Q2 is different. The question is no longer "what did they ship this week?" It is "how deeply do I embed this into my operations?"

The financial picture tells you why this is infrastructure, not an experiment. Anthropic's annualised revenue crossed $19 billion in March, up from $9 billion at the end of 2025. Secondary market trades now value the company at roughly $600 billion. Goldman Sachs and JPMorgan are lining up for an October Nasdaq listing that bankers believe could raise more than $60 billion. Eight of the Fortune 10 are customers. The platform changes in the past fortnight reinforce the direction: Computer Use went cross-platform with Windows support on 3 April. Third-party harnesses like OpenClaw lost subscription access on 4 April, pushing users toward first-party tooling. Copilot Cowork goes GA on 1 May, putting Claude inside Microsoft 365 for enterprise. Anthropic acquired biotech startup Coefficient Bio for $400 million, signalling that vertical specialisation is coming.

The feature set is now comprehensive. Screen control across Mac and Windows. Phone-to-desktop delegation. Persistent workspaces. 60+ connectors. Scheduled tasks. 1M context at standard pricing. Excel and PowerPoint add-ins with shared context. The building blocks are all there.

The honest caveat: reliability has not kept pace. Four service incidents hit in the first week of April alone (1, 3, 4, and 6 April), following nine in the previous fortnight. The platform is powerful but not yet dependable enough for truly mission-critical, zero-downtime workflows. Build with that reality in mind. Embed Claude into operations where a two-hour outage is inconvenient, not catastrophic. The reliability will catch up. The operators who embedded early will be the ones who benefit most when it does.

Skills have been available since launch. They work on every plan, including Free. And most operators have never built one. That is a mistake, because Skills are the single most effective way to make Claude consistently follow your processes instead of improvising every time.

What a Skill actually is. A folder containing a markdown file (SKILL.md) that describes how you want Claude to behave for a specific type of task. Your report format. Your onboarding checklist. Your email tone. Your variance thresholds. You write it once, upload it, and Claude loads it automatically whenever you give it a relevant request. No re-explaining. No "remember how we format this."

Skills are not prompts you paste every session. They are persistent instructions that Claude discovers and applies on its own. The difference is significant: a saved prompt requires you to remember to use it. A Skill activates when Claude recognises the context.

What you can build without writing code. The SKILL.md file is plain markdown with a small YAML header. No programming required. Here is what a weekly report Skill looks like:

Save that as SKILL.md inside a folder called weekly-ops-report/, zip it, and upload it in Claude Desktop under Settings, then Capabilities. Done.

Five Skills every operator should build. Based on what we see working:

Each of these takes under ten minutes to write. The return is every future session where Claude already knows how you work.

Skills vs Projects vs Memory. These three overlap, and operators ask which to use. The short version: Memory stores facts about you and your business. Projects store context for a specific workspace. Skills store processes and formats. Use Memory for "our financial year ends 30 June." Use a Project for "this is my month-end close workspace." Use a Skill for "this is how we format a weekly report." They layer together; they do not compete.

The honest take. Skills are genuinely powerful and genuinely underused. The main barrier is not technical difficulty; it is that most operators do not realise they can create them. If you have ever pasted the same formatting instructions into Claude more than twice, that should be a Skill. The ten minutes you spend writing it pays back on every subsequent session.

The Skill ecosystem is still thin outside of Anthropic's built-in document creation Skills. Community-shared Skills are emerging but there is no central marketplace yet. For now, the value is in building your own. Our Building Custom Skills guide walks through the full process with five copy-paste templates. The Skills platform page covers the technical details.

Most operators check competitors when they feel like it, usually the week before a strategy meeting or when someone on the team mentions a new feature a rival shipped. The result is inconsistent: sometimes thorough, sometimes a quick glance at a pricing page.

Here is how to set up a system that runs every fortnight and produces a structured brief you can actually use.

Create a Cowork Project called "Competitive Intelligence." Link a local folder where you will store outputs (e.g., /Documents/competitive-intel/). Write custom instructions:

Open the project and start a session. If you have existing competitor materials (previous analyses, screenshots of their pricing pages, product comparison spreadsheets), drop them into the linked folder first. Then:

Claude browses each competitor's site, pulls the relevant information, and structures the output per your instructions. The first run takes the longest because there is no baseline to compare against. After that, each brief builds on the previous one.

The brief gives you the overview. Follow-up prompts take you deeper:

That last prompt is where the Cowork Project pays off. Because prior briefs are in the linked folder, Claude can identify trends across multiple fortnights, not just snapshot a single moment.

Schedule a fortnightly reminder (calendar event, not a Claude scheduled task, since competitive analysis benefits from your input and steering). Every second Monday, open the project and run the brief. Over a quarter, you build a searchable archive of competitive movements that no one on your team has to maintain manually.

For teams: share the output document (the brief itself, not the Cowork Project) via your usual channels. The brief is a standard markdown file that works in Slack, Notion, email, or wherever your team reads.

Full guide: Competitive Analysis workflow | Setting Up Persistent Workspaces

Computer Use on Windows. As of 3 April, Claude's screen control works on Windows in both Cowork and Claude Code. Previously macOS only (Apple Silicon). Still a research preview, still roughly 50% success on complex tasks, but cross-platform support removes the biggest adoption barrier. Start with simple, well-defined tasks. WinBuzzer | Our Screen Automation guide

OpenClaw subscription lockout. On 4 April, Anthropic blocked Claude Pro and Max subscribers from using their flat-rate plans with third-party agent frameworks, starting with OpenClaw. Over 135,000 instances were running on Claude subscriptions. Users now face pay-as-you-go pricing for agentic workloads. One-time credit (equal to one month's plan cost) is available until 17 April. The message is clear: first-party tooling is the supported path. TechCrunch

Sonnet 4.5 and Sonnet 4 losing 1M context on 30 April. The 1M token context window beta for these older models is being retired. After 30 April, requests exceeding the standard 200K window will return an error. Migrate to Sonnet 4.6 or Opus 4.6, which support 1M context at standard pricing with no beta header required. Our Models page

IPO trajectory. Secondary market valuation has reached roughly $600 billion, up from the $380 billion Series G in February. October Nasdaq listing with Goldman Sachs and JPMorgan as lead banks. For operators: continued aggressive platform investment and a shipping pace that is unlikely to slow. TechCrunch

One thing to do this week: Set up a competitive intelligence project in Cowork.

You already know which five competitors matter most to your business. Create the project, write the custom instructions, run the first brief. Fifteen minutes. The output is a structured report you can share with your team today, and the project remembers the baseline for next fortnight's comparison.

This is what embedding looks like in practice. Not a one-off prompt. A system that runs on a cadence and gets sharper every time.

The experimentation phase is over. Q2 is about making Claude part of how your business actually operates. Start with the workflow that keeps falling through the cracks. Competitive intelligence is a good one.

Claude for Operators is an independent publication. Not affiliated with Anthropic.

Just getting started? Visit our Operator's Toolkit for free guides on every workflow.

Got a workflow worth sharing? Reply to this email.