Claude isn't a chatbot anymore. If your team is still treating it like one, you're already behind the ones who aren't.
In March, I walked into a National Sales Meeting. Room of eighty reps. Strong team. Good numbers. The VP introduced me, I thanked him, I opened my laptop, and the first question came from the third row before I'd said a word.
"Mike, which prompts should we copy?"
That question was the whole problem.
Not the answer. The question.
Because while that team was asking me which prompts to copy, a competitor three blocks over was already running a Digital Crew out of Claude Projects, with a HubSpot connector reading live pipeline, six custom Skills drafting follow-ups in their brand voice, and an Excel add-in that was quietly rebuilding the quarterly forecast every Monday at 7 AM. Same AI tool. Same price tag. Completely different gravity.
That's the GenAI Divide. And it's widening faster than most revenue leaders realize.
If you're in sales, marketing, agency work, or running your own shop, there are six moves your team probably hasn't made yet. They are not hard. They are not technical. They do not require a new tool. And if you make them in the next thirty days, you will find yourself on the right side of that divide.
Let's walk through them.

First, the reframe you need
Claude in April 2026 is not a chat window.
Claude is a stack.
It's three current models, four product surfaces, a growing library of Skills and connectors, and a desktop app that writes actual files to your hard drive. The chat you're using is one-sixth of the product. If you're still using Claude the way you used ChatGPT in 2024, you're operating on roughly the same fraction of capability that your smart speaker uses when you only ask it for the weather.
Here's the harder truth underneath that.
Prompt engineering is a secondary skill. Workflow Architecture is the one that matters.
That's a bigger claim than it sounds like, so let me say it plainly. If you have not redesigned the workflow, a better prompt is not going to save you. If your sales reps are still writing follow-ups by hand after their AI tool drafts one, you didn't redesign the workflow. If your marketing manager still copies the brand voice into every new chat, you didn't redesign the workflow. If your leadership still treats AI as a tool that lives in a separate browser tab, you didn't redesign the workflow.
The teams winning in 2026 are not better at prompting. They're better at staffing. They set Claude up the way you'd set up a new hire - with a desk, a folder of reference documents, a few instructions pinned to the wall, and the right hands on the right tools.
Here are the six moves that get you there.
Move 1: Pick your models. Stop mixing them up.
As of this month, Claude ships three production models you should know the names of.
Opus 4.7 is the flagship, released April 16, 2026. It's the one you want for strategy, long-horizon coding, complex market analysis, and anything where the output has to be genuinely good on the first pass. It's also the expensive one.
Sonnet 4.6 is the workhorse, released February 2026. It's the default for most daily work, including Cowork and most marketing and analysis tasks. If you're picking one model for your team, this is it.
Haiku 4.5 is the fast, cheap one, released October 2025. It's for volume work - classification, extraction, high-volume customer-facing traffic. Not for your keynote deck.
Every team I've worked with this spring had at least one person burning Opus tokens on tasks Haiku could have handled in a tenth of the time, and another person trying to get Haiku to do strategic work it was never designed for. That's Admin Drag in its purest form. Spend one meeting picking your default, your heavy-lift, and your volume model, and you'll see the cost curve and the quality curve both move in the right direction the same week.
Move 2: Build 3-5 Projects. Not one giant one.
A Project in Claude is a self-contained workspace with three pillars: a knowledge base you upload, custom instructions for how Claude should behave inside it, and an isolated chat history you can come back to.
Think of a Project the way you'd think of a client folder on your laptop. It has the brief, the brand guide, the examples, the meeting notes. Everything Claude needs to do work about that thing without you pasting the context in every time.
The single biggest mistake I see teams make here is building one megaproject called Marketing and dumping everything into it.
Don't do that.
The proven pattern - the one I've now watched work at marketing agencies, sales teams, and solo operators alike - is 3 to 5 focused Projects per client or per function. Here's the shape I'd copy first:
A Content Marketing Project with your brand voice, buyer personas, top-performing past posts, style guide, and banned phrases.
An Email Campaigns Project with your copywriting frameworks, segment definitions, welcome sequences, compliance rules, and past performance.
A Cold Outreach Project with your ICP, value prop, objection library, and first-line examples that have actually landed.
A Client or Workshop Project for each major engagement - transcripts, briefs, decks, and notes.
And a Competitive Intelligence Project with competitor pages, pricing sheets, analyst reports, and the Reddit and G2 mentions you don't want to re-Google every month.
Notice what that structure does. Each Project is narrow enough to be retrievable. Each one gets sharper every month as you feed it what you learned. And when you open a chat inside one of them, Claude already knows who you are, who you're talking to, and what good looks like.
Move 3: Install 5 to 10 Skills. This is the one most people miss.
If Projects are the rooms, Skills are the specialists who walk into the room when you call them.
Anthropic launched Agent Skills in October 2025 and published the open agentskills.io standard in December 2025. A Skill is, at its simplest, a folder containing a SKILL.md markdown file with metadata, instructions, and optional reference documents. Simon Willison called it conceptually extremely simple. He's right. It's a folder with a README that tells Claude how to do one thing well.
Here's why it matters. Claude uses progressive disclosure. You can have fifty Skills installed, and Claude will only read the ones that fit the task in front of it. That means every Skill you build is a specialized apprentice, sitting quietly in a folder, activating only when you need them.
The Skills I'd build first for a sales and marketing team:
A brand-voice Skill with tone rules, approved examples, and the phrases your team has banned. A linkedin-post Skill with the hook structure your audience responds to. A cold-outreach Skill with your personalization rules and ICP references. A keynote-outline Skill for the content leader who speaks. A workshop-prep Skill with your exercise templates. A monthly-reporting Skill with your KPI rollups for client retainers. An account-scoring Skill with your intent logic.
Seven Skills. One weekend of focused work. And you have stopped pasting the same instructions into every chat for the rest of the year.
This is the move most teams have not made yet. It's also the move with the highest leverage. Professional services firms who have built even three good Skills report the same thing - the quality of output stops bouncing around, and new hires can produce work at senior quality on day one.
Move 4: Wire the stack to your workflow with connectors.
Claude's MCP layer - Model Context Protocol, if you want to know what the acronym means - is the set of bridges that let Claude reach outside itself and read or write in the tools your team already lives in.
There are ten first-party connectors Anthropic maintains directly: Google Drive, Gmail, Calendar, Microsoft 365, Outlook, SharePoint, OneDrive, and Teams. Beyond that, the Claude Directory now has more than fifty official Claude-approved connectors, including HubSpot, Slack, Notion, Asana, Atlassian, Airtable, Stripe, Zapier, DocuSign, WordPress.com, ActiveCampaign, Ahrefs, Box, and monday.com.
You don't need all fifty. You need the three that cover where your revenue data actually lives.
For most revenue teams I work with, that's: their CRM (HubSpot became the first CRM with a native Claude connector, and it's already the one sales leaders ask me about by name), their team chat (usually Slack), and their shared drive (usually Google Drive or Microsoft 365). Wire those three in, and Claude stops being a standalone thing on a separate browser tab. It becomes something that knows where your pipeline lives.
One important honesty note before you connect anything. Every connector is a trust boundary. Anthropic disclosed a vulnerability in March 2026 called Claudy Day - an invisible prompt injection attempt via URL parameters in the Chrome extension. They patched it. But the research on prompt injection is still public, and roughly 11.2% of attacks still succeed against current mitigations. Treat connectors and MCP servers the way you treat browser extensions. Install sparingly. Audit the source. Pin the version. Don't connect a server your security team hasn't looked at.
Move 5: Ship Artifacts as deliverables.
Artifacts are the part of Claude that turns an answer into a tool.
They started as a side panel for code snippets and charts. They've quietly grown into something much bigger. As of June 2025, Artifacts can call Claude's API using the end user's own subscription, which means you can build a personalized flashcard generator, a tutor, a coaching app, and the person using it pays for their own tokens. As of October 2025, Artifacts can persist state, so you can build leaderboards, habit trackers, or mini-CRMs that actually remember things between sessions. As of October 2025, Artifacts can also call MCP servers.
Here's the move most Claude users are not making. They use Claude. They do not publish with it.
The Project + Artifact pattern is where the real differentiation is. You build a Project with your brand voice and your pricing sheet. You use it to generate a reusable Artifact - an ROI calculator branded to your firm, an interactive landing page your sales team shares during demos, a proposal builder that pulls from your services doc. You publish that Artifact, and suddenly your clients are interacting with a tool that has your thinking baked into it.
That's not AI content. That's an asset. You can give it away as a lead magnet, embed it on a sales page, or hand it to a prospect mid-call.
Most Claude users have never built one. The ones who have are quietly turning that single move into retention, referrals, and a differentiator that is very hard to copy.
Move 6: Know when to leave the chat.
The sixth move is not about the chat window at all.
There are three places where the real output of your work is happening right now, and none of them are inside claude.ai.
Cowork is the desktop app Anthropic shipped on January 12, 2026 as a research preview, then expanded to all paid plans within weeks. Cowork is goal-driven. You tell it what you want. It reads and writes actual files in folders you authorize. It calls connectors. It returns a finished deliverable - an .xlsx, a .pptx, a .docx, or an organized folder sitting on your drive. For non-developers, it's the feature of 2026. For developers, it's Claude Code with friendlier guardrails. Governance caveat worth knowing: Cowork is not yet covered by Anthropic's audit logs or Compliance API, so if you're in regulated work, wait or segment.
The Excel add-in launched GA this spring. Paste a CSV into Claude inside Excel, and it produces an entire multi-tab workbook with real formulas, cross-sheet dependencies, and cell-level citations. When you update an assumption, it preserves the formula dependencies. This is not a text-to-numbers illusion. This is a real Excel model.
The PowerPoint and Word add-ins launched in February and April 2026 respectively. Both are template-aware - they respect your master slides, layouts, fonts, and color schemes before generating. If your firm has a house deck template, they'll use it.
Use these three when the deliverable is a file. Use the chat when the deliverable is a decision. Knowing which one you need for the task in front of you is the single biggest time win available in 2026.
The numbers behind the reframe
Before we get to your move, look at what this has already produced in teams that committed to it.
Klarna publicly reported a 152% revenue increase per employee after embedding AI across their workflow. One hundred and fifty-two percent. Not a productivity bump. A capacity shift. Microsoft's internal sales organization reported a 9.4% revenue increase per seller after Copilot adoption - and that's inside a company that was already one of the best-instrumented sales orgs on the planet. When I ran a workshop with the Nevco Sports sales team, they left the session with their first working agents built and a plan to eliminate the administrative drag that was quietly burning their best reps out before noon.
None of those numbers are coming from better prompts.
All of them are coming from redesigned workflows.
Here's the baseline most revenue leaders I work with never quite want to hear: sales and marketing teams spend up to 70% of their week on non-revenue activities. Call logging. CRM updates. Research. Reporting. Handoffs. That's the Admin Drag. The whole point of the six moves above is to pull that number down - not by pushing humans harder, but by staffing a Digital Crew that handles the robot work while the humans focus on strategy and selling.
I run my own crew through an AI Chief of Staff I call CLU. CLU is not a product. CLU is a configuration - Projects, Skills, connectors, the same six moves we just walked through. CLU handles my podcast prep, my research, my executive admin, and a lot of the writing scaffolding I used to do myself. That's not theory. That's the architecture behind everything you just read.
Your move this week
Here's what I'd suggest.
Take twenty minutes, look at the six moves above, and pick the one you haven't made yet where the cost of not making it is highest. For most sales leaders reading this, that's the connector - getting Claude wired into your CRM. For most marketers, it's the Projects - ending the brand-voice copy-paste cycle. For most agency owners, it's the Artifact - building one branded tool your clients can actually use.
Just pick one. Start there. Build it this week. See what happens.
And when you've picked it, two next steps depending on where you are.
If you run a revenue team and you want a pro to walk your organization through this, my keynotes and workshops do exactly that. I've spent the last year delivering the Revenue Architecture playbook inside National Sales Meetings, leadership offsites, and full-day hands-on labs where teams build their first working agents in the room. If that sounds like the catalyst your team needs, start here and we'll find the shape that fits.
If you want to quantify your Admin Drag before anything else, grab the Friction Audit. It's the same diagnostic I run at the start of every engagement. You answer a short set of questions about how your team actually spends their week, and you get a real number back - the true dollar cost of the manual friction your organization is carrying. Most leaders are shocked by it the first time they see it. Download the Friction Audit here.
The team three blocks over isn't waiting. You don't have to either.