ARGENTUM × SAVANT-AI
Argentum AI Leadership Institute · Powered by SAVANT-AI

Welcome. This home is private to the Founders Cohort.

Your key is in your welcome email.

The Argentum AI Leadership Sprint · Powered by SAVANT-AI

Welcome, Founders Cohort.

Nineteen leaders. Three mornings. Everything you need for the week lives on this page, and everything we promised you will be kept here, in writing.

AI opens up to firms of all sizes multiple opportunities that are safe and profitable: when recognized, when imagined, and when implemented with logic, systems thinking, and the willingness to explore, iterate, and share.

That sentence is the whole sprint. Monday we recognize. Wednesday we imagine. Friday we implement.

Start here

Two kinds of AI

You are not new to AI. Your industry runs some of the most measurable AI in American business, and many of you already use the newer kind personally. This week is about the difference, and about leading both.

The AI your company already runs

Narrow AI: it watches, predicts, and alerts.

  • Fall-detection cameras in memory care
  • Predictive health analytics and sensors
  • Sales lead scoring in your CRM
  • Staffing and scheduling optimizers

Built by vendors. Trained for one job. Your teams operate it; you govern it and approve its budget.

The AI you are here to lead

Generative AI: it reads, writes, reasons, and builds.

  • Works in plain English, no integration required
  • Reads your documents and spreadsheets: research and analysis are its native gifts
  • Broad, not narrow: one tool, a thousand jobs
  • Used directly by you, not only by specialists
  • Frontier models advance monthly; production models are the dependable workhorses

This is the kind you will build with, with your own hands, starting Monday morning.

The AI you know works for your buildings. The AI you are learning works with you.

Both matter. Leaders who understand the second become sharper buyers of the first, because they can finally ask vendors the questions that matter.

The end from the beginning

The Week

All sessions live on Microsoft Teams; your calendar invites carry the link. Come ready to work: this is a workshop, not a webinar. You are consuming something wonderful, a bite at a time.

Day One · Recognized

Get Oriented and Create

Monday, July 13 · 10 a.m.–1 p.m. CT · session slides
  • The state of AI in 2026: two kinds of AI
  • The R.O.C.K. framework, and your prompt library
  • Build your AI chief of staff: your first autonomous agent
  • Build a skill: your know-how, made reusable
  • Assemble your digital twin: your first custom assistant
  • Bring it to life: from Markdown to a designed page
  • Workspace agents, and the bigger picture
  • Demonstrate: your work, your voice
You leave with: two working builds of your own and a ten-minute evening assignment that pays off Wednesday.
In your world · your chief of staff reads the same census, move-in, and NOI numbers your regionals do, and hands you the one-glance morning brief.
Day Two · Imagined

Connect and Automate

Wednesday, July 15 · 10 a.m.–1 p.m. CT · session slides
  • Your twin learns to think like you
  • Durable memory: your knowledge, in files you own
  • Skills, uploads, and councils of experts
  • Simple automations: a prompt proven once, then scheduled
  • See your information: data made visible
  • Demonstrate: your work, your voice
You leave with: an assistant that carries your knowledge and your first scheduled automation.
In your world · a twin trained on your operating standard answers "how would we handle this?" the same way in every community at once.
Day Three · Implemented

Lead the Way

Friday, July 17 · 10 a.m.–12 p.m. CT · session slides
  • The week's work, in your cohort's own words
  • Your 90-day roadmap home: the first draft of your firm's AI operating system
  • Your Index results and personalized next steps
  • Leading the change: your team, your policy, your standard
You leave with: a roadmap you wrote yourself and a cohort that speaks your new language.
In your world · a transparent, logged, human-in-the-loop standard your board, your regulator, and your caregivers can all trust.
Who is assembling

The Room

19
senior leaders, coast to coast
12
organizations, from single-campus to the nation's largest operator
~1,250
communities operated by the people on your screen

Chief executives sit beside chief technology, people, and financial officers; presidents beside the specialists who run their systems. Two organizations sent full delegations. Whatever seat you hold, someone in this room holds the seat next to it, and by Friday you will share a language.

What your surveys are telling us

Updating as baselines arrive, through Monday morning.

The early returns already say something important: this cohort spans the entire map. Personal baselines have arrived from Level 2 through Level 5; organizational baselines from Tier 2 through Tier 4-5. Leaders early in the journey are sitting beside leaders at the frontier, and the course is built for exactly that: every build has a first rung and a depth rung, and your baseline is private to you. Wherever you start Monday, the measurement in October is against yourself, no one else.

Haven't taken the surveys yet? The links are below, and if the weekend runs out, it will simply be your first activity in class.

And come ready to answer one question out loud Monday: what do you want from this week? Say what you want. It shapes what we teach.

Set up your tools and self

How to Get the Most Out of This Course

  • Say what you want
  • Write it down on paper
  • Imagine
  • Record your thoughts: your voice, your ramble
  • Wear many hats
  • Explain yourself to a recording
  • Have AI structure your thinking
  • Get your hands on the tools
  • Get your ideas in the tools

Take the two surveys 10 min total

The Argentum AI Leadership Index survey (9 questions) sets your personal baseline; your results arrive immediately by email.

The Argentum AI Maturity Index survey (14 questions) establishes your organization's baseline.

Today, tomorrow, or Monday morning: all fine. If the weekend gets away from you, taking the survey will simply be your first activity in class. Your results are private to you, and progress can only be measured from a baseline.

Bring your own laptop Monday, with a paid ChatGPT or Microsoft 365 Copilot account

A paid subscription keeps your work private: required for Monday's hands-on workshop. For Microsoft users this means the paid Microsoft 365 Copilot license (the add-on your IT team assigns), not the free Copilot Chat that comes with Office. Not sure which you have? Ask IT today: "Do I have a Microsoft 365 Copilot license?"

Optional: a personal sandbox recommended

A personal laptop or personal email address on hand. Some leading-edge tools we will reference live outside company policies and firewalls: not for use at work, but worth exploring personally.

Arrive already thinking: the R.O.C.K. framework

You will use it Monday within the first hour. Four elements of clear, reliable communication with AI: Role (who should the AI be), Objective (what outcome you want), Context (the situation: what it needs to know), and Know-how.

Know-how is the element only you can supply. Context is context for the situation; Know-how is context for the person. It is your earned knowledge, brought into the prompt: a CFO brings a CFO's know-how, a CMO brings theirs, a CHRO brings twenty-eight years across every company they ever served. The judgment, the standards, the rules of thumb, what good looks like to you. It is why your prompt produces work an intern's prompt cannot: the tool supplies the labor; you supply the mastery. The Pathway below has the working template.

If you don't get to all of this, come anyway. You'll be caught up. The only thing you cannot skip is the laptop.

Your questions, answered before you ask

Is the hype real?
The one AI number your industry already trusts is peer-reviewed: 41% fewer falls, 69% fewer ER visits. The rest you will judge by building, not by believing. That is the whole design of the week.
Will this touch resident data?
No. Everything built this week lives on the leadership surface: strategy, drafting, planning, your own knowledge. No resident names, no health records, no PHI, ever.
Is this about cutting staff?
No. The standard we teach is augmentation: AI that returns hours to your people, transparent and human-in-the-loop. The industry already has its cautionary tale of the other kind; we build its opposite.
Am I too far behind? Is this too technical for me?
Different starting points are expected by design. If you can describe what you want in plain English, you can do everything in this course. Nothing here requires a technical background.
What does our existing AI investment mean here?
It means you already govern AI. Fall detection, predictive analytics, lead scoring: that experience makes you a sharper leader of the kind you are about to use yourself.
The working surface · copy, paste, build

The Pathway

This section follows the course in order. During workshop portions, copy the block for the current activity, paste it into your platform, and make it yours by replacing the bracketed parts. Same outcome on every platform; three dialects. And while the industry waits for its systems to talk to each other, notice: nothing you build today needs any integration at all.

1
Day One · Activity 1 of 10

Your first R.O.C.K. prompt

Pick one real task from this week. Use the framework to hand it to AI clearly. Slides: the R.O.C.K. framework.

Role: You are [the seat you need at your table, e.g., an experienced senior living operations analyst who writes plainly].

Objective: [One outcome, named. e.g., Research X and brief me in one page. / Analyze this spreadsheet and tell me the three things that matter. / Draft the memo for Monday's leadership meeting.]

Context: I am the [your role] of a senior living organization with [size/scope]. What matters here: [2-4 facts the AI needs]. Do not use any resident or health information.

Know-how: Here is what my experience tells me matters, which you cannot know on your own: [2-3 hard-won rules of thumb or standards from your years in this seat, e.g., "occupancy alone misleads; read move-outs against acuity mix"]. What good looks like to me: [your standard for done].

Working instructions: ask me up to three clarifying questions before you begin, then produce [format: a one-page memo / a table / an email draft]. Plain language. Keep it under [length].
ChatGPT: paste into a new chat; save the good version as a Project instruction. Copilot: paste into Copilot Chat (work tab uses your Graph). Claude: paste into a new chat; save as a Project instruction.
2
Day One · Activity 2 of 10

Save it: your prompt library

A good prompt is an asset. Keep every one that works in a single document you own. Today one document; Wednesday we turn the best of them into files, and files are where skills come from.

# My Prompt Library
Owner: [your name] · Started: July 13, 2026

---
## [Prompt name, e.g., Monday Briefing Draft]
Date: · Platform: · What it does:
[Paste the full prompt here, exactly as it worked.]
---
Paste into a Word doc, Google Doc, or OneNote page named "My Prompt Library." One document, one owner: you. Every prompt that works goes in, named and dated.
3
Day One · Activity 3 of 10

Schedule it: your AI chief of staff, your first agent

An assistant helps you while you are there. An agent works while you are not. A scheduled assistant is your first agent. Slides: from assistant to agent.

Role: You are my AI chief of staff.

Objective: Produce my morning briefing.

Context: [Paste or attach your R.O.C.K. self-description from Activity 1.] My standing priorities this quarter: [list 2-3].

Know-how: This is my standard for a briefing, earned the hard way. Every briefing has exactly four parts, in this order:
1. My top three priorities today, one line each, based on what I have told you.
2. One question I should be asking that I have not asked.
3. One risk I am not seeing.
4. One thing to stop doing.
Keep the whole briefing under 200 words. No preamble.

Run it once now so I can correct it. Then I will schedule it.
ChatGPT: prove it, then save as a scheduled task (daily, your time). Copilot: prove it, then schedule the prompt (requires the paid Microsoft 365 Copilot license). Claude: prove it in a chat, then save it as a scheduled task in Cowork; or run it as your first click of the morning.
4
Day One · Activity 4 of 10

Build a skill: your know-how, made reusable

A skill is a piece of your expertise, written down once, used forever. It is the seed of every playbook and workflow to come. Slides: skills.

# Skill: [Name it, e.g., Email Triage, My Way]
Owner: [your name] · Version: 1

## When to use this skill
[The situation that calls for it.]

## The know-how (what I know that the tool does not)
- [rule of thumb 1, e.g., anything from a family member outranks anything internal]
- [rule of thumb 2]
- [what good looks like when this is done right]

## The steps
1. [step]
2. [step]

## Never
- [the line this skill must not cross, e.g., never auto-reply to a family concern]
Save it as a Markdown (.md) file in your library. ChatGPT: add it to a Project or custom GPT instructions. Copilot: keep it in OneDrive/Notebook where Copilot can reach it (work grounding needs the paid license). Claude: add it to a Project's knowledge. Then the callback: give your scheduled chief of staff this skill. And notice what you just made: a file that travels. Email it to a colleague and your know-how works in their platform too, as-is or as a template for their own version. That is one leader's intelligence becoming a team's consistent result.
5
Day One · Activity 5 of 10

Assemble: your digital twin

Files plus a system instruction, gathered in one place, equals an assistant that is unmistakably yours. This is your digital twin, version one. Slides: assembling the twin.

You are building my digital twin: an assistant that thinks alongside me in my own patterns.

First, interview me one question at a time. Wait for my answer before the next question:
1. Who am I? (role, organization, what I am responsible for)
2. How do I think? (how I approach problems, what I trust, what I doubt)
3. What do I think? (my current priorities and positions)
4. What do I do, and how do I get it done?
5. What do I want to do next?
6. What is missing, and how could I create it?

Then write a one-page profile of me in Markdown titled "[My Name] - Twin Profile v1". I will save that file and attach it, along with my skills, wherever we work together. It is durable memory I own.
The assembly: ChatGPT: a custom GPT or Project with your profile + skills attached, and your R.O.C.K. prompt as the instruction. Copilot: a custom agent with your files in reach. Claude: a Project with your files as knowledge and your prompt as Project instructions. The system instruction is simply the R.O.C.K. prompt that never leaves.
Day One · Bring It to Life

From a file to a page, before your eyes

You just created a Markdown file: your Twin Profile. Now watch durable memory become something visual, designed, and shareable. Slides: bring it to life.

Take the Markdown file below and turn it into a single, clean HTML page I can open in my browser: something I would be proud to show a colleague.

What I envision: [one page, my name at the top, calm and readable, big ideas first].

[Paste your Twin Profile v1 here, or any Markdown file you own.]
Claude: renders it live as an Artifact. ChatGPT: produces the file; open it in your browser. Copilot: ask for the HTML file, save, open. Wednesday we go deeper: real work data, made visible.
Day One · Stretch

Workspace agents

The newest tools in ChatGPT and Microsoft 365 Copilot: agents your whole team can call by name. We will demonstrate; building them is a stretch goal today and a theme Wednesday. Slides: workspace agents · Slides: the bigger picture.

The bigger picture, in minutes: multimodal (image and voice) · translation (one document, every language your workforce and families speak) · synthetic data (practice data created from nothing: specific, useful, and safe to practice on, as long as it is generated fresh and never converted from real records) · Claude inside Excel · connectors, the bridge to Wednesday.

Day One · Evening · 10 minutes

Show your twin how you think

The evening assignment. Two open questions, answered in your own rambling words: record yourself or type freely. Bring the result Wednesday, when your twin learns to think like you. Details arrive at the end of Monday's session.

6
Day Two · Activity 6 of 10

Your twin learns to think like you

Monday your twin met you. Today it learns how you think: your evening answers become part of its durable memory. Slides: the twin upgrade.

Here is how I actually think, in my own words: [paste or attach your evening answers].

Read it. Then tell me three things you now understand about how I work that you did not know before. Update my Twin Profile to version 2 with what you learned. I will save the new file.
7
Day Two · Activity 7 of 10

Durable memory: your knowledge, in files you own

Durable knowledge lives in Markdown files you keep. Temporary knowledge is fetched in the moment. Files used to be the hard drive; context is the RAM. Whole teams can be stored in files. Slides: durable memory.

Three things one file can be: an instruction (a skill your tools follow), a record (analysis you want to keep: a spreadsheet's findings, a document's summary, preserved and shareable), and a template (your structure, handed to a colleague to make their own). It is the first and easiest way, on any platform, to carry a durable instruction from one workspace into another. How your team manages and versions the files is your organization's choice; that the intelligence travels is the point.

# [Topic] - Knowledge File v1
Owner: [your name] · Updated: [date]

## What this is
[One paragraph: what knowledge this file holds and who should use it.]

## The facts that matter
- [fact]
- [fact]

## How we do it here
[Your organization's way: standards, steps, preferences.]

## What good looks like
[An example, pasted in. Show the tools what you want to imitate.]

## Open questions
- [what is missing]
Save as a .md file (any text editor). ChatGPT: upload to a Project or custom GPT. Copilot: keep in OneDrive/Notebook where Copilot can reach it (work-data grounding needs the paid license). Claude: upload to Project knowledge.
8
Day Two · Activity 8 of 10

A council of experts

One prompt, many hats: the seats you wish were at your table.

Convene my council of experts. Five seats:
1. A chief financial officer who prices everything in labor hours and occupancy.
2. An operations veteran of 25 years.
3. A compliance and risk mind.
4. A marketing leader who thinks in families and trust.
5. [Your choice: the seat you most wish you had.]

When I bring a question, each seat answers in turn, three sentences maximum, in their own voice. Then the council states where it agrees, where it disagrees, and the one action it recommends.

My first question: [ask something real].
9
Day Two · Activity 9 of 10

See your information

From Markdown to HTML: data made visible. Tell the tools what you envision; show them what you want to imitate.

Take the Markdown file below and turn it into a single, clean HTML page I can open in my browser and share with my team.

What I envision: [describe: a one-page dashboard / a briefing the eye can scan in 30 seconds / big numbers first, detail below].

Style: calm, readable, no clutter. [Optional: paste or attach an example whose style you want to imitate.]

[Paste your Markdown here.]
Claude: renders it live as an Artifact. ChatGPT: produces the file; open it in your browser. Copilot: ask for the HTML file, save, open.
10
Day Three · Activity 10 of 10

Your 90-day roadmap home

This is the first draft of your firm's AI operating system: what you build, what you teach, what you govern, how you measure. You decide the rest when you return home. It starts here. Slides: the roadmap scaffold.

# My 90-Day AI Roadmap
Name: [you] · Written: July 17, 2026

## What I build next (weeks 1-2)
[The one assistant or agent you extend first, and for what.]

## What I show my team (weeks 2-4)
[Who sees it first. What they build with you. Champion to team: from one to many. Teach one person one build within seven days.]

## The policy and governance questions I raise
[What our AI policy must answer. What data may never enter a tool.]

## How I measure
[Retake the Leadership Index survey in October. What number moves, and why.]

The common language

By Friday these are yours. Say them plainly and your whole organization can follow.

Prompt · what you say to AI; the R.O.C.K. framework makes it reliable
Know-how · the K in R.O.C.K.: your earned expertise brought into the prompt. Context is for the situation; Know-how is for the person
Playbook · a written, repeatable way of doing something
Workflow · steps chained together toward an outcome
System · workflows plus memory plus schedule, running as one
Assistant · AI that works when you ask (R.O.C.K. and skills)
Agent · AI that works when you are not asking (schedules, loops, orchestration)
LLM · large language model, the engine under every tool this week
NLP · natural language processing; why plain English is now an interface
Multimodal · reads and produces text, images, audio, video
Research and analysis · the native gifts of a language model: it reads, so it can research; it reads at scale, so it can analyze — documents, spreadsheets, even images
AGI · artificial general intelligence; the horizon, not this week
Broad vs. narrow · a model for everything vs. a tool for one thing
Frontier vs. production · the newest capability vs. the dependable workhorse
Commercial vs. open source · you rent the model vs. you can hold it
Cloud vs. on-prem · their computers or yours; privacy and control follow
Open weights & fine-tuning · models can learn your language: more accurate, more nuanced, more human
Durable vs. temporary memory · files you own vs. context fetched in the moment
Markdown (.md) · plain text with structure; the file format of durable memory. Emailable, platform-neutral, yours: instruction, record, or template
Everything promised, kept in writing

The Library

This section grows across the week and persists after it. What was promised is what appears here.

Recordings

Every session is recorded and shared with the group shortly after it ends.

Day One · posted Monday afternoon  ·  Day Two · posted Wednesday afternoon  ·  Day Three · posted Friday afternoon

Build cards

Step-by-step cards for each build, in your platform's dialect: Claude, ChatGPT, or Microsoft 365 Copilot. Posted as sessions progress. Until then, the Pathway above is your working surface.

Playbooks and reference materials

Templates and practical playbooks you can hand to your team, added across the week: starting with "AI in Senior Living: what already works, and the guardrails", plus who we learn from and the books we recommend.

Your Index results and next steps

Your survey results arrived by email the moment you finished. On Friday you receive your personalized next steps, so the baseline becomes a direction.

Practice files

Something real to analyze, so you never stall hunting for material. Entirely synthetic, created from nothing: specific, useful, and belonging to no one. That is a lesson in itself. Generate practice data fresh; never paste real resident records and ask the tool to disguise them. Practice data is a training tool, not a substitute for your own privacy obligations.

Sunrise Meadows portfolio (spreadsheet) · 12 fictional communities, six months of census, labor, and food cost. Try: "Which three communities need my attention, and why?"

A family letter and an incident note (document) · Try: "What matters most here, what is missing, and draft my reply."

Also fair game: any public page, like Argentum's public workforce research. "Read this and brief me" works on the whole readable web.

Sample skills, written as files

A skill is a set of working instructions your AI keeps and reuses, and it is nothing more than a Markdown file. These two are real ones we use. Open them, read them in plain English, then install: paste into ChatGPT and ask it to run the review, or hand the file to Claude or Copilot the same way.

Weekly inbox and calendar review (skill) · Your week, read back to you: unresolved threads, meeting follow-ups, and next week's priorities.

Monthly operating review (skill) · The same idea at altitude: themes, workstreams, relationship momentum, and what to continue, stop, or delegate.

These read Gmail and Google Calendar. Microsoft shop? Change those words to Outlook and Teams calendar, and the same file works. One edited line, and the skill travels. That is the point of Markdown.

Session takeaways

After each session: the takeaways and the words you just learned, in writing, posted beside the recording. Read it in two minutes; hand it to your team as-is.

With your permission, we may quote your wins in what we publish. We will always ask first, by name.

Day One · Monday afternoon  ·  Day Two · Wednesday afternoon  ·  Day Three · Friday afternoon

The specialist stack (for your personal sandbox)

Tools we reference beyond the big three, worth exploring on a personal account: Perplexity (research with sources) · Gamma (decks from a prompt) · NotebookLM (your documents, conversational) · HeyGen (video avatars) · ElevenLabs (voice) · Suno (music). More will be added as the cohort asks.

After Friday

This home stays open through October 16. The templates, practice files, and playbooks on it are yours to copy and keep. This field moves quickly, and we would rather retire the page than let it age on you. In October we will invite you to take the survey again, so you can see how far you have come.