If you came here from the carousel, this is the part the carousel could not hold. The five workspaces are the easy part to name. The value is in how you build them, and almost nobody publishes that. So here it is, start to finish.
First, the five words that make this work
Before the workspaces, the vocabulary. Get these five terms straight and the rest is just execution.
Project. A saved workspace inside Claude with its own instructions and its own knowledge base. You set it up once. Every chat you start inside it already knows your context. This is the brain.
Project instructions. The standing rules for that workspace. Your role, the purpose, the tone, and the rules that always apply. Claude loads them in every conversation inside the project.
Knowledge. The documents you upload once. Claude references them automatically. On a paid plan, Claude searches them even when they grow past what fits in a single conversation.
Connectors. The links between a workspace and your live tools, like Google Drive, your email, and your calendar. A connector lets Claude pull from the source instead of waiting for you to paste.
Skills. Reusable instruction sets Claude loads on its own when a task matches. You write the rule once, for example your SOP format or your brand voice, and it applies everywhere without re-pasting. Skills need code execution turned on in settings.
Cowork. The desktop agent. Regular chat answers you. Cowork does the work, reading, editing, and creating files in folders you grant it, inside an isolated environment. This is where a workspace stops advising and starts producing. It runs on the desktop app on paid plans.
Before you build
— A paid Claude plan if you want document search at scale and Cowork. Projects and Skills work on a free account too, with a five-project limit and code execution turned on.
— Your real documents, gathered per workspace. The quality of the output is set by what you load.
— Code execution enabled in settings if you plan to use Skills.
One honest note. None of this is a single upload that returns a finished system. You put your real context in once. The workspace assembles, structures, and scales from there. Treat the setup the way you would onboard a person.
1. The Numbers Co-Pilot Open Claude, click New Project in the left sidebar, name it "[Practice] Finance."
PROJECT INSTRUCTIONS
You are the finance partner for [Practice Name], a [single or multi] location aesthetic practice. You know our service menu and pricing, our payroll and commission model, provider and room utilization, and our monthly revenue and margin targets, all in project knowledge. When I bring you a decision, model the scenarios with real numbers, show the downside I am most likely to miss, and ask the questions a careful CFO would ask before I commit. Always state your assumptions and flag anything you are estimating. Never present a single option as the only answer. Do not give legal or tax advice. Tell me when to involve our accountant.
Load into knowledge: the last twelve months of P&L, your current price list, your payroll and commission structure, membership terms, and a provider utilization export.
Connectors: Google Drive, pointed at your finance folder, so it reads the live spreadsheets instead of waiting for a paste.
STARTER PROMPTS
Here is an offer I am considering: [paste]. Model gross margin at 20, 40, and 60 percent redemption. Show me where this loses money.
We are considering a fifth injector at [pay structure]. At what monthly revenue does that hire break even, and what am I not accounting for.
Compare raising tox per unit by [x] against raising membership price by [y]. Which protects margin with the least patient fallout.
Cowork: point it at your finance folder and have it build a one-page monthly scorecard from the raw exports, then set it to refresh on a schedule so the number is waiting for you each month.
2. The SOP Builder
New Project, named "[Practice] Operations."
PROJECT INSTRUCTIONS
You build standard operating procedures for [Practice Name]. I will give you a recording transcript, a rough description, or a messy existing document. Return a clean SOP in this structure: purpose, who owns it, when it runs, the step by step, and what good looks like. Write at a level a brand new hire can follow. Use our terminology from project knowledge. When a step is unclear, ask me one question rather than guessing.
Load into knowledge: any existing SOPs, your org chart, role descriptions, and your preferred SOP template.
Skill to build: an SOP format Skill so every SOP comes out in your exact structure without re-pasting the template. Use the skill-creator to generate it, or upload a folder with your template inside.
Connectors: Google Drive, so finished SOPs save straight to the right folder.
STARTER PROMPTS
Here is a transcript of [staff member] explaining how they open the clinic: [paste]. Turn it into an SOP.
Here is our current closing checklist: [paste]. Rewrite it as a full SOP and flag the three steps most likely to be skipped on a busy day.
Cowork: hand it a folder of recorded walkthroughs or transcripts and have it produce a full set of SOPs, named and filed consistently, in one pass. This is the step that gets a quarter of undocumented procedures written in an afternoon.
3. The Compliance Brain
New Project, named "[Practice] Compliance." Treat this one with more care than the rest.
PROJECT INSTRUCTIONS
You hold the compliance standards for [Practice Name]. Project knowledge contains our protocols, consent templates, scope-of-practice rules by role and state, and our documentation standard. When I ask, check the material I give you against our written standard and tell me exactly where it falls short. Cite the standard you are checking against. You support documentation and consistency. You do not give legal advice and you do not make the clinical decision. Flag clearly when a provider, medical director, or attorney needs to weigh in.
Load into knowledge: your protocols, consent templates, state scope-of-practice references, your charting standard, and your medical director agreement.
Handle with care: keep protected health information out of it. Use de-identified examples only, and connect a dedicated, access-controlled folder rather than opening your whole drive. The sensitivity is exactly why this workspace is worth building, and exactly why you set it up deliberately.
STARTER PROMPTS
Here is a de-identified chart note: [paste]. Does it meet our documentation standard. List what is missing.
Draft a consent addendum for [treatment] that matches the structure of the consents already in knowledge.
Build an audit checklist for [treatment category] from our protocols.
Cowork: have it compare a batch of de-identified notes against the standard and return an exceptions list, so the chart audit is documented in advance rather than reconstructed after a complaint lands.
4. The Marketing Brain
New Project, named "[Practice] Marketing."
PROJECT INSTRUCTIONS
You write marketing for [Practice Name] in our voice, which lives in project knowledge along with our offers, our patient segments, and our best past campaigns. When I give you an offer and an audience, draft the email and the text message ready to send. Match our voice, not a generic template that sounds like every other med spa. Keep every claim compliant: no guarantees and no results promises we cannot stand behind. I approve every message before it goes out.
Load into knowledge: your brand voice notes, ten to twenty of your best past emails and texts, your current offer list, your segment definitions, and a do-not-say list for claims.
Skill to build: a brand voice Skill. Because Skills load automatically across projects and the Claude add-ins for Word, Excel, and PowerPoint, you write your voice once and every workspace inherits it.
Connectors: your content calendar in Drive, and your calendar for scheduling sends.
STARTER PROMPTS
Offer: [paste]. Audience: patients who have not booked in six months. Draft a three-email sequence and two texts in our voice.
Rewrite this promo to sound like us and to stay inside our claims rules: [paste].
Cowork: give it the quarter's promo calendar and have it draft the full set of emails and texts as files, each labeled by send date, ready for your review. The campaign that used to sit in drafts for a week goes out the same afternoon.
5. The Onboarding Brain
New Project, named "[Practice] Onboarding." This one draws on the SOPs you built in workspace two.
PROJECT INSTRUCTIONS
You build onboarding for [Practice Name]. Project knowledge contains our SOPs, role descriptions, and protocols. When I name a role, produce a structured first-thirty-days plan: what they learn each week, which SOPs to read, who they shadow, and how we confirm they are ready. Write it so a manager can hand it over without editing.
Load into knowledge: your SOPs, role descriptions, and any training checklists you already use.
Connectors: the Drive folder where your SOPs live, so onboarding always reflects the current version.
STARTER PROMPTS
Build a thirty-day onboarding plan for a new front desk coordinator from our SOPs.
Turn our injector protocols into a week-one competency checklist a manager can sign off on.
Cowork: have it generate a complete onboarding packet per role as formatted documents, pulled straight from the SOP folder. New hires ramp from how the practice actually runs, not from what a busy coworker remembered to mention.
If you build only one thing, build this
Build a single practice-context Skill that holds how you sound and your non-negotiables: your voice, your claims rules, your terminology. Skills load automatically across your projects, the Claude apps, and the Office add-ins, so you write it once and every workspace inherits it. Start there. It makes all five workspaces sound like you on day one.
A note on agents and scheduling
Cowork is where these workspaces become operational. It runs multi-step tasks, it can run them on a schedule, and through plugins it can bundle Skills, connectors, and specialized sub-agents into a single install, so the workspace shows up as a specialist rather than a blank assistant. You do not need any of that to start. You need one project, loaded properly. The agent layer is what you grow into once a workspace has earned its place.
The build order
You do not build all five in a weekend. That is the fastest way to end up with five half-loaded workspaces and no momentum. Start with the one tied to whatever is costing you the most right now. If margin decisions are being made on instinct, start with the Numbers Co-Pilot. If the practice stalls every time one person is out, start with the SOP Builder. If documentation is inconsistent and an audit would expose it, start with the Compliance Brain. Load it with your real context, use it for two weeks, and let it earn its place before you add the next. By the time you reach the Onboarding Brain, you already have the SOPs it needs, because you built them in step two.
What stays human
None of this replaces clinical judgment, and none of it replaces your voice. The workspace handles consistent execution against the parameters you define. The provider still makes the call at the chair. You still approve the campaign. The system removes the repetition, not the responsibility.
Start with one, not five
The practices pulling ahead are not the ones with the most tools. They are the ones who treated AI as infrastructure and built it in the right order. Pick one workspace this week. Write the instructions. Load the documents. Run three real tasks through it. If it feels right, it will become one of the most useful things in your operation.
For the rest of the Real Stack series and the full AI-Ready Practice build order, subscribe. Everything here is independent analysis, built from inside aesthetic operations.