AI Working Session
0%
Saved
Luxe Business Backend · Working Session

Operationalizing
Your Expertise With AI

Tools · Skills · Reusable Systems

You already know how to do the work. The problem is that the knowledge lives in your head, so every recap, report, proposal, and assessment gets rebuilt from scratch by the only person who can do it. This session is about moving that expertise into something that can run without you sitting at the keyboard — and doing it with the tools you already pay for.

Session Date
Format Live Training + Self-Paced Reference
Outcome Reusable AI Infrastructure
00

How To Use This Page

Read this first

This page does two jobs. During the session it's what we follow together. Afterward it's the manual — every prompt is here, every exercise can be redone alone, and everything you type into it is saved in your browser so you can come back mid-thought.

Teal Prompt Blocks
Ready to use as written. Hit Copy, paste into the tool named at the top of the section, swap the bracketed parts for your own information.
Gold Template Blocks
The same thing with the specifics stripped out. These are what you use the second and third time, so you're filling in blanks instead of starting over.
Your Turn Boxes
Short exercises. Each one ends with how you'll know it worked, so you can judge your own output without asking anyone.
Fields And Checkboxes
Type directly into the page. It saves as you go. Save Backup writes a file you can reopen later or move to another computer.
Quick Reference
My Workbook collects everything you've typed in one view. Prompt Library and Template Library pull every block into one scrollable list. Glossary defines the vocabulary.
One Thing
To Know
Your work saves to this browser on this device. It does not follow you to your phone. Use Save Backup to keep a copy, and Import Backup to bring it back or move it somewhere else.
01

What Tool Does What

The rule that ends the guessing
ClaudeChatGPTSavvySuite AI StudioFathomClickUp

"Which AI is better" is the wrong question. The right one is what job each tool holds, so you stop opening whichever tab is already up and hoping.

The ruleIf it produces a document, use Claude or ChatGPT. If it produces a record or an action, use SavvySuite.
ToolIts JobWhy That One
ClaudeVoice-sensitive, document-heavy, reusable workSkills. You teach it your standards once and it applies them every time without re-explaining.
ChatGPTFast generative volumeOutreach variants, invitations, brainstorm passes — where speed beats polish and voice matters less.
SavvySuite AI StudioAnything that touches the CRMBuilds apps and assessments that write back to the contact record instead of producing a document.
FathomCapture onlyOne recording, one transcript, one place everything else pulls from.
ClickUpWhere work and reusable assets liveThe index. If you can't find what you built, you rebuild it.
The Consolidation Question

If you're paying for an outside app builder while SavvySuite AI Studio does the same job inside the platform that already holds your contacts, that's two tools with one job. The app gets rebuilt where the data already lives, and one subscription goes away. Ask this question about every tool on your list.

1

Your Turn

3 minutes

Write a one-line job description for each tool you pay for. If a tool has no job, or two tools have the same one, you just found money.

Claude
ChatGPT
SavvySuite
Fathom
Other tool
You'll know it worked whenYou can say out loud which tool you'd open for a task you've never done before.
02

Stop Starting From Scratch

Skills, instructions, and reusable intelligence
Claude

There's a difference between asking AI to do something and teaching AI how you do something. The first one you repeat forever. The second one you build once.

Most people live in the first mode without knowing the second exists. They write a great prompt, get a great result, close the window, and write it again next week slightly worse.

The Vocabulary
Chat
One conversation. Disposable. Use it for anything you won't need again.
Project
A folder of related chats that share context and files. Use it per client or per initiative.
Custom Instructions
Standing preferences that apply everywhere — how you want to be spoken to, what you never want. Set once.
Knowledge File
Reference material you upload so AI can consult it. Your frameworks, your templates, your standards.
Skill
A documented process for producing a specific output to your standard. The important one. This is your expertise written down in a form that runs.
What Makes
A Good Skill
You've done it well at least once, you'll do it again, and you can name what separates a good version from a weak one. If you can't name the standard, the skill can't hold it.
Your Candidate List

Everything below is something you've already done well and will do again. Each one is a skill waiting to be written: your writing voice, session recaps, executive summaries, client reports, proposals, training follow-up, intake summaries, scope documents.

I'm going to give you a piece of work I created and was happy with. I want you to reverse-engineer it into a reusable skill so I can produce this same quality of output again with different client information, without re-explaining myself every time.

Analyze what I give you and identify: the purpose this output serves and who reads it, what inputs are required to produce it, the structure and section order, the voice and tone rules that make it sound like me, the standards that separate a good version from a weak one, what should never appear in it, and the format and length.

Then write it as a skill: a short description of when to use it, the inputs it needs, step-by-step instructions for producing the output, the quality standards it must meet, and one example of what good looks like.

Ask me questions about anything you had to guess at before you finalize it.

Here is the work:
[PASTE YOUR WORK HERE]
Here is a skill I've been using, and here's a recent output from it that wasn't quite right.

Tell me which specific instruction in the skill allowed that output, then rewrite only that part. Don't rebuild the whole skill. Show me the before and after of what you changed and explain why it fixes the problem.

THE SKILL:
[PASTE SKILL]

THE OUTPUT THAT MISSED:
[PASTE OUTPUT]

WHAT WAS WRONG WITH IT:
[ONE OR TWO SENTENCES]
2

Your Turn — Build Your First Skill

5 minutes
  1. Find a recap, summary, or report you wrote and were happy with.
  2. Copy Prompt 2A above and paste it into Claude.
  3. Paste your work where it says to.
  4. Answer whatever it asks you.
  5. Save what comes back as a Skill in Claude.
You'll know it worked whenYou run the skill on a different client's material and the output comes back in your voice without reformatting.
03

Recording To Client Deliverable

The chain that kills manual recap work
ClaudeSavvySuiteFathom

One recording produces four things for three different readers. The mistake is writing each one separately from memory. The move is extracting once, then shaping that extraction for each audience.

TranscriptExtractionClient RecapExecutive SummaryFollow-Up Email
What Stays
Yours
Judgment calls, anything sensitive, and the final read before it sends. AI produces the draft and the structure. You decide what's true and what's appropriate to say.
Here is a transcript from a client session. Extract and organize the following, using only what's actually in the transcript — do not infer or fill gaps:

Decisions made, and who made them.
Insights or realizations the client reached.
Recommendations I gave.
Commitments — who agreed to do what, by when, including mine.
Open questions left unresolved.
Anything flagged as sensitive or confidential.

Note anything ambiguous rather than guessing at it.

TRANSCRIPT:
[PASTE]
Using the extraction above, write a session recap addressed directly to the client.

Open with what we covered in one or two sentences. Move to what they worked through and arrived at, in their language where possible. State their commitments and mine clearly, with dates. Close with what happens before we meet next.

Write to them, not about them. No corporate filler, no motivational language, no summary headers. Reflect their own words back where they said something that mattered. Keep it under 400 words.
Now write a version for the sponsor who is paying for this engagement, not the participant.

Report themes, patterns, and progress against the engagement's stated goals. Include nothing the participant shared in confidence — no personal disclosures, no specific admissions, nothing that identifies a struggle as belonging to that individual.

If the sponsor should be aware of something I need to raise with them directly rather than in writing, flag it to me separately at the end.

Keep it under 250 words.
Write the follow-up email that sends with this recap. Warm, brief, professional.

Confirm the next session date, name the one thing they're focused on before then, and nothing else. Under 100 words.
Build me a skill for writing client session recaps.

WHO READS IT: [CLIENT / SPONSOR / BOTH]
INPUTS IT NEEDS: [TRANSCRIPT, PRIOR RECAPS, ENGAGEMENT GOALS]
STRUCTURE: [YOUR SECTION ORDER]
VOICE RULES: [HOW YOU WRITE — SENTENCE LENGTH, WHAT YOU NEVER SAY]
LENGTH: [WORD COUNT]
NEVER INCLUDE: [CONFIDENTIAL CATEGORIES, JARGON, ANYTHING ELSE]
GOOD LOOKS LIKE: [PASTE YOUR BEST EXAMPLE]

Write it so I can hand it a transcript and get a recap that needs no reformatting.
3

Your Turn

2 minutes

Pull a transcript from a recent call. Run Prompt 3A only — just the extraction. Don't write the recap yet. Watch how fast the raw material appears.

You'll know it worked whenEvery commitment made on that call is on the list, including the ones you'd forgotten.
04

Build An Assessment That Feeds Your CRM

Scored, branded, and connected
SavvySuite AI StudioClaude

AI Studio builds apps from a description. You write what you want in plain language and it produces a working web app — which means an assessment is no longer a development project, it's a paragraph.

The reason it belongs inside SavvySuite rather than a document: results become contact records. A score lands on the contact, a tag fires, and follow-up runs on its own.

What Has To Be True Before You Automate
Scored vs. Open
Keep them separate. Scale questions produce numbers. Open-ended questions produce insight. Mixed together, neither works.
Categories
Group questions so each category produces its own subscore. A single total tells someone they have a problem. Subscores tell them where.
Interpretation Bands
Score ranges tied to what that range actually means. You write these. This is your professional judgment and it is the part that has value.
The Hard Line
AI applies your interpretation. It never invents what a score means. That's your IP, and it's what makes the report sellable later.
Why this matters more than it looksScore forty leaders inside one organization and you no longer have forty individual results. You have a bench diagnostic — and that is a different product at a different price.
The Working Example

The build prompt below produces a Performance Management Health Check: twelve questions, three categories, scored one to five, with lead capture and a CRM connection. Paste it into a new AI Studio project exactly as written and it builds.

Build me a web-based assessment called the Performance Management Health Check. It should look professional and work well on mobile.

STRUCTURE
A welcome screen explaining the assessment takes about four minutes and evaluates how well their organization's performance management system is actually working. Then 12 questions, one per screen with a progress indicator. Then a lead capture screen collecting first name, last name, work email, company name, and job title. Then a results screen.

QUESTIONS — each answered on a 1 to 5 scale where 1 is Strongly Disagree and 5 is Strongly Agree.

Category 1 — Clarity of Expectations
1. Every employee can describe what success in their role looks like without checking a document.
2. Performance expectations are set in writing before the period being evaluated, not during it.
3. Managers across our organization define "good performance" in roughly the same way.
4. When someone's role changes, their performance expectations are formally updated.

Category 2 — Feedback and Coaching Cadence
5. Employees receive performance feedback on a predictable schedule, not only when something goes wrong.
6. Managers are trained and confident in delivering difficult feedback.
7. Feedback conversations are documented consistently across managers.
8. Employees would say they know where they stand right now.

Category 3 — Accountability and Documentation
9. Performance concerns are documented at the time they occur, not reconstructed later.
10. Two employees with similar performance issues would be handled consistently by different managers.
11. Our documentation would hold up if a termination decision were challenged.
12. Performance decisions — promotions, raises, discipline — can be traced back to documented evidence.

SCORING
Each category scores out of 20. Total out of 60. Show all three category scores and the total on the results screen.

CATEGORY BANDS
4-9 Fragile — this area is running on individual managers' instincts rather than a system. Risk is high and inconsistency is likely already visible.
10-15 Developing — the foundation exists but isn't applied consistently. Results depend on who the manager is.
16-20 Established — this area functions as a system. Focus shifts from building to maintaining.

OVERALL BANDS
12-27 Significant exposure.
28-43 Functional but inconsistent.
44-60 Strong foundation.

RESULTS SCREEN
Show the total, the three category scores visually, the interpretation for each category, and name the lowest-scoring category as the priority area. End with a button to book a consultation.

CRM CONNECTION
On submission, POST the results as JSON to a webhook URL I will provide. Include first name, last name, email, company, job title, the three category scores, the total score, and the overall band name.

DESIGN
Clean and professional. Deep teal and white, generous spacing, large readable type.
Build me a web-based assessment called [NAME]. Professional, mobile-friendly.

Welcome screen explaining it takes about [X] minutes and evaluates [WHAT IT MEASURES]. Then [N] questions, one per screen with a progress indicator. Then lead capture for first name, last name, work email, company, job title. Then results.

Questions on a 1 to 5 scale, Strongly Disagree to Strongly Agree, grouped as:

Category 1 — [NAME]
[FOUR STATEMENTS]

Category 2 — [NAME]
[FOUR STATEMENTS]

Category 3 — [NAME]
[FOUR STATEMENTS]

SCORING: each category out of [X], total out of [Y].

CATEGORY BANDS
[LOW RANGE] [WHAT IT MEANS]
[MID RANGE] [WHAT IT MEANS]
[HIGH RANGE] [WHAT IT MEANS]

OVERALL BANDS: [RANGES AND MEANINGS]

RESULTS SCREEN: total, category scores visually, interpretation per category, lowest category named as the priority. Button to book a consultation.

CRM CONNECTION: POST results as JSON to a webhook I'll provide — contact details, category scores, total, band name.

DESIGN: [COLORS], one question per screen, progress indicator.
I've built an assessment with a category called [CATEGORY NAME] measuring [WHAT IT MEASURES]. I need interpretation language for three score bands: low, middle, and high.

For each band, write two to three sentences that name what's actually happening operationally at that level, name the specific risk or cost it creates, and point toward what changes it — without prescribing the solution, because that's the conversation I want to have with them.

Write to the reader directly. No jargon, no motivational language, no softening. These are senior leaders who will respect being told the truth.

Here are the four statements in this category:
[PASTE THE FOUR STATEMENTS]
Connecting It Back To SavvySuite
  1. In SavvySuite, create custom fields on the contact record for each category score, the total, and the band name.
  2. Create an inbound webhook in a new workflow. Copy the URL it gives you.
  3. Paste that URL into your AI Studio app where the build prompt says to.
  4. Submit a test. Map each incoming value to the matching custom field.
  5. Add a tag based on the band, and let that tag trigger the follow-up you want.
4

Your Turn

5 minutes
  1. Open a new project in AI Studio.
  2. Paste Prompt 4A exactly as written and let it build.
  3. Take the assessment yourself and look at the results screen.
  4. Then rewrite one category's interpretation language in your own voice.
You'll know it worked whenA test submission lands on a contact record with the right score and the right tag.
05

A Recurring Event System

Built once, rerun with three swapped fields
ChatGPTSavvySuiteClaude

Anything you intend to run repeatedly — a roundtable, a workshop, a quarterly session — should be built as a system with swappable inputs, not a campaign you rebuild each time. The difference between those two is about six hours per event.

ChatGPT Handles

  • Ideal participant profile
  • Outreach sequence
  • Invitation variants

SavvySuite Handles

  • Registration form
  • Confirmations and reminders
  • Tagging and list building

Claude Handles

  • Pre-event briefing
  • Post-event summary
  • Follow-up and insight recap
The Part
People Skip
Registration questions aren't admin. They're the input that makes the briefing, the follow-up, and every personalized message possible without you writing anything twice.
I run invitation-only executive roundtables for senior leaders. The next one is on [TOPIC].

Define the ideal participant: title and seniority, company size and type, what they're currently dealing with that makes this topic urgent, what they'd want from a room like this, and what would make them decline.

Then give me five screening signals I can look for on LinkedIn that suggest someone is a fit, and three that suggest they aren't.
Write a three-message LinkedIn sequence inviting someone matching this profile to an executive roundtable on [TOPIC], held [DATE].

Message one: connection request, under 200 characters, references something specific about them, no pitch.
Message two: sent after they connect. The invitation — what the room is, who's in it, why them. Under 150 words.
Message three: one follow-up a week later. Brief, no pressure, gives them an easy no.

Peer to peer, not marketing. No exclamation points, no hype, no "I'd love to pick your brain."
Here's my base invitation:
[PASTE MESSAGE TWO FROM 5B]

I'm sending it to these people:
[NAME, TITLE, COMPANY, AND ONE SPECIFIC DETAIL FOR EACH]

Rewrite it once per person. Keep the structure and the offer identical. Change only the opening line so it references their specific situation. Don't flatter, and don't overreach on what you know about them.
Here are the registration responses for my roundtable on [TOPIC]:
[PASTE RESPONSES]

Build me a briefing:
Who's in the room, by name, title, and company.
The themes that appear across multiple responses.
The two or three tensions where people clearly disagree — that's where the good conversation is.
Anyone whose situation is unusual enough that I should be careful how I frame things.
Five questions to open with, ordered so the room warms up before it gets honest.

Flag anyone who indicated something they'd rather not discuss publicly.
Here's the transcript from my roundtable on [TOPIC]:
[PASTE]

Produce four things:

1. A summary of themes and the strongest insights, attributed by role rather than by name.
2. A follow-up email to attendees with the summary and one specific thing worth acting on this week.
3. An insight recap I can publish publicly — no names, no companies, no identifying details.
4. A short private note to me flagging which attendees showed signals worth a direct follow-up, and why.
What's the biggest challenge you're facing in [TOPIC AREA] right now?

What would make this session worth your time?

How many people do you have responsibility for?

What have you already tried that didn't work?

Is there anything you'd rather not discuss in a group setting?
5

Your Turn

3 minutes

Run Prompt 5A for an event you're actually planning. Read the profile it gives you and notice whether it matches who you've been inviting.

You'll know it worked whenRunning it again next quarter means changing the topic, the date, and the profile — nothing else.
06

Sales Calls And Proposals

Prepared before, drafted after
ClaudeSavvySuite

Two halves. Before the call, whatever you already know about the prospect becomes a prep brief. After the call, the transcript becomes scope and proposal language that drops into your existing templates.

The outcome worth naming: you stop writing proposals from a blank document, and you stop walking into calls trying to invent good questions in real time.

Here's what I know about a prospect before our call:
[PASTE APPLICATION, INTAKE FORM, LINKEDIN PROFILE, EMAIL THREAD — WHATEVER YOU HAVE]

Build me a prep brief:

What they said their problem is, and what it's more likely to actually be.
Which of my services this points toward, and which would be wrong for them.
Eight diagnostic questions in the order I should ask them, moving from situation to impact to cost.
What they'll probably push back on, and what's behind each objection.
What I should be listening for that would change my recommendation.
Two things I should not assume.

Be direct about where the information is thin and I'll be guessing.
Here's the transcript from a discovery call:
[PASTE]

Produce:

A one-paragraph statement of their situation in their own language — close enough that they'd read it and feel understood.
The engagement I should recommend, and why that one.
Scope: what's included, stated specifically.
Exclusions: what's explicitly not included, so it doesn't get assumed in later.
Deliverables with a realistic timeline.
The three biggest risks to this engagement succeeding.
Proposal language for the situation and approach sections.

If anything in the call suggests they aren't a fit, say so before you write any of it.
Build me a skill for drafting proposals.

MY SERVICES AND WHAT EACH INCLUDES: [LIST]
MY STANDARD INCLUSIONS: [WHAT EVERY ENGAGEMENT CARRIES]
MY STANDARD EXCLUSIONS: [WHAT NEVER GETS ASSUMED IN]
PRICING STRUCTURE: [HOW YOU PRICE, WHAT CHANGES IT]
PROPOSAL SECTIONS IN ORDER: [YOUR STRUCTURE]
VOICE: [HOW YOU WRITE TO PROSPECTS]
NEVER INCLUDE: [DISCOUNTING LANGUAGE, OVERPROMISING, ANYTHING ELSE]

Input is a call transcript. Output is a draft proposal I only have to review, not write.
6

Your Turn

4 minutes

Run Prompt 6A against a real prospect you're speaking with soon. Use whatever you already have — even just their LinkedIn and one email.

You'll know it worked whenYou walk into the call with questions you didn't have to invent on the spot.
07

The Reuse Rule

So you stop rebuilding what you already built
ClaudeClickUp

Here's the pattern that eats the most time and gets noticed the least. You make something good. You don't write down how. Weeks later you need it again, can't find it, and rebuild a slightly worse version from memory.

The ruleIf you expect to need it again, the work isn't finished until the process for recreating it is captured.
Document exactly how we created this, including the inputs, the decision logic, the standards applied, the formatting, the output requirements, and the quality criteria. Turn that process into reusable instructions so I can reproduce this outcome with new client information.

Then tell me what you had to guess at, so I can make those decisions explicit instead of leaving them to chance next time.
Where It Lives

A reuse rule without a filing system just builds a faster junk drawer. One home, one naming pattern, one index. ClickUp holds the index — what it is, what it does, where it lives, when you last improved it.

AssetWhat It ProducesWhere It LivesLast Updated
Session Recap SkillClient recap from a transcriptClaude — SkillsDate
Proposal SkillDraft proposal from a callClaude — SkillsDate
Assessment Build PromptScored assessment appClickUp — DocDate
Here's a task I'm about to start:
[DESCRIBE IT]

Here's the index of skills and assets I've already built:
[PASTE YOUR INDEX]

Tell me whether something on this list already covers it, partially covers it, or doesn't. If something partially covers it, tell me what to add rather than what to rebuild.
7

Your Turn

4 minutes
  1. Pick something you built earlier in this session, or anything you've made recently that worked.
  2. Run Prompt 7A on it.
  3. Decide where it lives and what it's called.
  4. Add the row to your index.
You'll know it worked whenSomeone asks for that deliverable a month from now and you open a file instead of a blank page.
08

Your Practice Plan

One build a week for four weeks

Nothing on this page matters if it stays on this page. One thing per week, in this order, because each one makes the next easier.

Practice Progress0 of 12
Weeks 1–2
0/6
Weeks 3–4
0/6

Week One · Your Recap Skill

Start here
Target date
Checklist

Week Two · Your Assessment

Builds on week one
Target date
Checklist

Week Three · Your Proposal Skill

Highest time return
Target date
Checklist

Week Four · Your Index

Makes the rest permanent
Target date
Checklist
Your Top Three
Priority One
Priority Two
Priority Three
09

Reference

Glossary, platform notes, and what to keep out of AI
Glossary
Chat
One conversation with AI. Disposable. When you close it, the context is gone.
Project
A container for related chats that share files and context. One per client or per initiative.
Custom Instructions
Standing preferences applied to everything — tone, format, what you never want.
Knowledge File
Reference material uploaded so AI can consult it: frameworks, templates, past work.
Skill
A documented process for producing a specific output to your standard. Your expertise, written down so it runs.
Context
Everything AI can currently see. It doesn't know what you didn't give it in this conversation.
Prompt
The instruction you give. A good one names the audience, the output, the constraints, and the standard.
Transcript
The text version of a recording. The raw material for every deliverable in Section 03.
Webhook
A URL that receives data automatically when something happens. How your assessment talks to your CRM.
Custom Field
A slot on a contact record for information specific to your business — a score, a category, a band.
Tag
A label on a contact that can trigger automation. How a score turns into follow-up.
Vibe Coding
Describing an app in plain language and having AI build it. What AI Studio does.
Where The Buttons Are
Claude — Projects
Left sidebar. Create one per client or initiative, add knowledge files to it.
Claude — Skills
Settings, then Capabilities. Add a skill and it becomes available in your conversations.
ChatGPT — Projects
Left sidebar. Same idea: grouped chats with shared files.
ChatGPT — Custom GPTs
Explore GPTs, then Create. Best for something you'll run often with the same instructions.
SavvySuite — AI Studio
Describe the app you want. It builds it. Connect it to a workflow with an inbound webhook.
Keeping Client Information Clean

You handle information other people don't. Before anything goes into an AI tool, decide what belongs there.

Strip First
Names of third parties, health information, compensation figures, anything disclosed in confidence, anything about a protected characteristic, active legal matters.
Replace With Roles
"The VP of Operations" instead of a name. The analysis works the same and the transcript stops being identifiable.
Tell Clients Plainly
That sessions are recorded, that AI assists with drafting summaries, and that you review everything before it's shared. Nobody objects to the version they were told about in advance.
Never Automate
Language between two named parties in a dispute, anything involving a complaint about a protected characteristic, and judgment calls about a specific person's future. That's the work they pay you for.
Review Before Send
Nothing reaches a client or a sponsor without you reading it. The draft is AI's. The decision is yours.

My Workbook