CRUISE TERMINALS INTERNATIONALStrategic AI Masterclass
AI is your intern. Now put ten of them to work.
A practical, hands-on session built around what the nine of you actually told us — so you leave able to hand one real piece of your job to AI on Monday.
July 21, 2026 · Facilitated by Future Foundation · Rodrigo Faerman
Open warm. One line: "This isn't a generic AI talk — it was built around what the nine of you told us over WhatsApp." Set the promise: leave able to hand one real task to AI, with a prompt you can run Monday.
⋀ YOUR FACILITATOR
RODRIGOFAERMAN
Leader in Strategy & AI
I help companies turn AI from a buzzword into real results. I take what's working in the world's most advanced AI markets and apply it hands-on, inside real operations — not slideware.
in rodrigofaerman
◎ @rodrigofaerman
30–45 seconds. Establish credibility fast, then get out of the way — the session is about them, not you. The "turn AI into real results, not a buzzword" line sets up the ROI-reality slide next.
Who's guiding you
20+ years turning technology into results — across three continents
🇺🇸 USA🇮🇱 Israel🇧🇷 Brazil
Education
Business + technology
Business Administration, Boston University · MSc in Information Technology, Bocconi (Italy) · MBA in Leadership & Innovation.
Leadership
Ran it at scale
Former CIO of Zumba Fitness · ex-BankBoston, Telefónica, Accenture — tech, finance, BI & marketing in multinationals.
Focus today
AI, strategy & innovation
Advised hundreds of companies and leaders. Takes what's working in the world's most advanced AI markets and applies it hands-on — not slideware.
AI doesn't replace vision, execution, and courage. It multiplies whoever has all three.
Authority beat (real CV, from Rodrigo's keynote). ~45 sec. Boston University, Bocconi (Italy), MBA; ex-CIO of Zumba Fitness; BankBoston / Telefónica / Accenture; 20+ years, 3 continents. The "hands-on, not slideware" line pre-seeds the Act-2 reveal (this session itself was built, not talked about). Confident and brief, then move on.
A few of the names along the way
Where that experience comes from
Logos are language-neutral, so this stays as the original. 15–20 sec — let the names land, don't read them out. Reinforces the credibility from the previous slide, then pivot to the ROI reality.
Let's start honest
74–95% of enterprise AI projects deliver no ROI.
Not because the tools are weak — because "just add AI" isn't a strategy. The winners don't adopt more AI. They focus, prioritize what works, and put people in charge of it.
His signature hook. This reframes the whole session: you're not here to "use AI more," you're here to use it well — which is exactly the intern/ownership model coming up. Land the number, pause, then: "so today isn't about hype. It's about what works."
Today · ~90 minutes
Where we're going
1
How to think about AI — the intern, and the team of ten
20 min
2
Your challenges — and a reveal about today
15 min
3
Your tool today: Claude — and how to use it well
15 min
4
Live & practical — your challenges, step by step
30 min
5
Monday — one thing you'll actually do
10 min
Don't over-explain the agenda — 20 seconds. The point is to signal it's practical and ends with them doing something real.
Before we start
Quick show of hands
Who used AI at work this week?
Who used it for something that actually mattered — not just a test?
Who chatted with our WhatsApp bot before today?
Hold that last one — we'll come back to the bot. 👀
Diagnostic. Most hands will go up on Q1 — say so: "You're already past 'why AI' — today is about doing it well." Q3 seeds the Act-2 reveal. Keep it under 3 minutes.
Part One
How to think about AI
Two mental models. Get these right and everything else is easy.
This act is the heart of the session. If they remember one thing, it should come from here.
First, a reframe
Wherever you are with AI, you're somewhere on this line
0–1 yr
2–5 yrs
6–12 yrs
13–18 yrs
19–40 yrs
40+ yrs
From "just met it" to "grew up with it." None of these is wrong — the only goal today is to move one step to the right.
Rodrigo's maturity metaphor. Inclusive and disarming: the skeptic and the power user are both just at different ages with AI. Ties back to the opening poll — most of the room is already well along the line. Keep it light, ~30 sec.
Mental model #1
AI is a brilliant new intern
It has read almost everything ever written. But it doesn't know your business, your numbers, your standards, or your audience — yet.
Treating it like Google — one question, one answer, judged by the first reply — is why most people are underwhelmed. That's not how you get great work out of a smart junior.
The reframe. Most disappointment with AI comes from expecting an oracle instead of managing a junior. Land it slowly.
"ChatGPT as Google" vs. "a thinking partner" is just the difference between an intern you ignore and one you actually manage.
Echo the sponsor's own framing (unattributed). Tie "step away from the Excel and act like an owner" to this: managing an AI intern is the same muscle — context, ask for insight, push back when it stays shallow.
AI is a tool.
A powerful one — but a tool. The result depends entirely on the person holding it. That's good news: the advantage isn't the AI. It's how well you brief it and how sharply you review it.
Rodrigo's "IA é ferramenta." Reinforces the intern model and defuses both fear ("it replaces me") and hype ("it does everything"). The craftsman, not the chisel, makes the work.
Mental model #2
Don't hire one intern who knows everything.
Hire ten.
The counter-intuitive leap. People want one perfect AI. The power is many, each briefed for one job, in parallel.
Mental model #2
Specialized & in parallel
🧾 The recon intern
Matches your ledgers, surfaces only the exceptions.
✍️ The commentary intern
Turns actuals-vs-budget into board-ready narrative.
📊 The deck intern
Drafts the MBR slides from your numbers.
They don't get tired, don't wait in line, don't need each other to finish first.
So "which AI is best?" is the wrong question. The right one: what's the job, and how do I brief the intern for it?
Each demo in Act 4 is one of these interns. This slide plants that structure.
And it's already happening here
One of your colleagues already runs their own custom bots — trained on their own feedback — with AI doing hands-on work in Excel and PowerPoint.
This isn't theoretical for CTI. It's already in the building.
Guillermo. Keep anonymous unless he's happy to be named — confirm with him beforehand. If he'll say one sentence about his bots here, that peer proof beats anything you can claim.
The part that stays human
Ten interns. One orchestrator — you.
This isn't AI replacing you. It's you conducting it — setting the direction, judging the work, making the call, owning the result. The interns draft; the orchestrator decides. That job matters more now, not less.
Say this plainly and with confidence — it's the antidote to "so the machine does my job." AI does the drafting; the judgment, the standards, the accountability stay human. You are not being automated away; you're being promoted to orchestrator. Land it before Part 2.
Part Two
Your challenges
We asked all nine of you. Here's what we heard.
Shift from concept to mirror. This is where they see themselves in it.
Data. Not opinion.
This session wasn't designed from a template or a hunch. It was built from your nine real answers — what you actually told us eats your day. So let's look at the data.
Rodrigo's "Dados. Não opinião." Bridges his method to our discovery: everything you're about to see came from them, not from us guessing. Sets up the mirror and the reveal.
The warm-up
We asked. You answered.
Nine short WhatsApp conversations. One question at a time. Your role, your week, your frictions, and the one thing you'd hand off.
Two clear groups emerged — and they need very different help.
What we captured
Where your time actually goes
Your honest relationship with AI
Your biggest "there's got to be a better way"
Your dream hand-off
Answers were promised private — keep quotes unattributed on screen. You may attribute live only if someone's clearly happy to own it. Never show scores or rankings.
Two tracks
You split into two groups
Track A
Tell the story
Turning numbers into an executive-ready narrative — close commentary, MBR decks, board messaging, forecasting.
Same close cycle. Different pain. We'll go deep on one live example from each.
This two-track split is the core finding. It's why today isn't one-note. Point out both halves are stages of the same month-close.
Track A · Tell the story
In your words
"They stay at the surface level… and I'm still catching spelling and grammar mistakes, which with all the tools we have is kind of inexcusable."
"Every single word matters and requires thoughtful consideration — it takes up a considerable amount of time."
"I'd like a bot that holds the historical context of my work. Like Iron Man's Jarvis."
Unattributed. Let them recognize themselves. The "Jarvis" line is a great bridge to Projects/persistent memory in Act 3.
Track B · Do the data
In your words
"The biggest time-takers are reconciliations."
"Invoice inputting and vendor setup definitely take up a lot of my time."
"Go through my emails and clean up redundant / duplicate invoices before I start my day."
These are the four whose day is data, not narrative. If you'd only seen the FP&A leaders, you'd have built the wrong session.
Quick round
How was it — talking to the bot?
Weird? Useful? Surprisingly natural? Let's hear a couple of you.
Take 2–3 answers, 2 minutes. This is the setup for the reveal on the next slide — let them praise/critique the "bot" before you tell them what it was.
Here's the thing
That bot was AI. The analysis that found these two tracks and profiled each of you was AI. These slides were AI.
This entire experience was designed and built with the same tools you're about to learn — in a fraction of the usual time. And a human decided, at every step, what actually mattered.
Emotional peak. Three beats, slow: the bot → the analysis → the slides. The concept from Act 1 becomes undeniable because they lived it. Pause before moving on.
Ten interns built your masterclass
Not one genius AI — a team of specialists
Intern 1
The interviewer
Ran 9 WhatsApp conversations, one question at a time.
→
Intern 2
The analyst
Read all nine, found the two tracks, profiled each of you.
→
Intern 3
The designer
Shaped this session and built these slides.
By the end of today, you'll put a few of these to work on your challenges.
This is the concrete payoff of "10 interns." Tie each intern back to the mental model. Then pivot: "so how do you actually do this? Let's pick a tool."
Part Three
Your tool today: Claude
The concepts work with any AI. Today we go deep on one.
No tool war. Pick one, learn it deeply, the skills transfer to any of them.
Why this one, today
Why Claude fits this room
Strong at reasoning & writing — exactly what board commentary and MBR narrative need.
Great with documents & data — hand it the variance table, the invoice, the reconciliation.
Projects with persistent memory — the intern that remembers your business (the "Jarvis" idea).
Hands-on with Excel & PowerPoint — already in use inside CTI.
You can use any AI. We picked one so we can go deep, not shallow.
Keep it 2–3 minutes. The message is "pick one and master it," not "Claude is the only good one."
How to use it properly
How to brief your intern: 5 parts
Role
Who
"You are a senior FP&A analyst at CTI…"
Context
The situation
The business, the audience, the numbers.
Task
The one ask
The single clear thing you want done.
Format
Shape
A table, 3 bullets, a board paragraph.
Constraints
Guardrails
Tone, length, "flag anything uncertain."
A vague brief gets vague work. This is just good delegation — written down.
This is the craft. Everything in Act 4 uses this shape. Say it out loud: "onboard, then delegate — same as any intern."
The difference is night and day
Weak brief → strong brief
❌ Weak
"Summarize this month's results."
Gets you: generic bullet points restating the obvious.
✅ Strong
"You are a senior FP&A analyst at CTI. Explain each material variance below — the likely driver, the forward-looking implication, and any risk. Board tone, 3 short paragraphs, flag anything you're unsure of."
Gets you: an executive-ready first draft you can sharpen.
If you want a live moment here, run both prompts in Claude on the same numbers and let the room see the gap. Optional — depends on clock.
The intern that remembers
Onboard once — with Projects
Load your close checklist, chart of accounts, last quarter's commentary once.
Every new chat starts already knowing your business.
This is the "Jarvis" one of you asked for.
⚠️ The opposite of this
Deleting every conversation erases everything the intern just learned about you. Persistent memory is the whole point — don't throw it away.
Directly (kindly, unnamed) answers the "deletes every chat" habit the sponsor flagged. Persistent context is what turns a generic tool into your analyst.
One more habit
Give it your data — then own the output
Feed it the real thing — paste the variance table, upload the invoice. AI is far better with your numbers than with your description of them.
The first answer is a draft from a junior. Push back: "sharper," "more forward-looking," "you missed the FX impact."
You sign off. The owner reviews; the intern drafts. That never changes.
Reinforces the intern metaphor one last time before the demos. Ownership is the through-line.
Before you paste real numbers
Real data? Treat it like any outside tool.
Don't paste what you couldn't send outside without sign-off — live financials, PII, anything confidential — until CTI IT approves the setup. (That's why today's data is fake, on purpose.)
Training: on business / enterprise plans, your inputs are not used to train the model; on personal plans, check the data-training setting. Confirm CTI's exact plan with IT.
It's a drafting partner, not a source of truth — verify every total and figure yourself before anything leaves your desk.
When in doubt: use sample data, or bring it to a 1:1.
This WILL come up in the room — probably right here. Have the answer ready and give it proactively; don't let it come up cold. If you don't know CTI's exact Claude plan/account setup, say so honestly and route it to IT and the 1:1s. The "not a source of truth — verify totals" line builds trust with a finance audience rather than undercutting it.
The concept
The real skill? A process.
Introduce PROCESSES as its own idea. A great result isn't luck or a flash of insight — it's a repeatable method, run the same way every time. This is the frame for the whole demo block.
Why processes + AI
Great work is a process, not a flash of insight
A process is a repeatable method — the same steps, the same way, every time.
It finds what a glance misses, because it never skips the boring parts.
AI's real superpower isn't creativity — it's running your process tirelessly: step 500 as carefully as step 1.
Your month-close, your reconciliation, your reporting are all processes. Perfect work to hand an AI.
The bridge from concept to practice. AI doesn't get bored, doesn't skip a line, doesn't lose focus at 6pm. That's exactly why it's so good at finance processes.
A 90-second exercise
How many squares do you see?
Commit to a number. Shout it out. Then watch what a process finds that your eye skipped.
Get them to commit out loud — most say 16. The gap between their guess and the real answer is the whole point. Then click through the frames.
Click through — the animation counts squares of every size, plus the ones hiding between the lines. The final frame lands on 35.
Here's the point
You saw 16. A process found 35.
A method more than doubled what intuition found — and that's exactly what AI does for your close. It runs every line, every size, every time. Your reconciliation has a "35" hiding in it too.
The bridge to the live demos. The squares = the exceptions in a recon, the real story under a flat variance. AI is the tireless process that finds them. Now go live.
Part Four · Focus. Prioritize what works.
Live & practical
We won't cover everything — we'll go deep on one real example from each track. Every step. Copy the prompt and use it Monday.
Rodrigo's "Focar. Priorizar aquilo que funciona." — the reason we pick 1–2 demos, not all. Live in Claude; deck holds the challenge, the prompt, and what to notice. Use realistic but fake data — never live CTI financials.
Look at the non-obvious.
The value isn't in the number everyone already sees — it's in what's hiding underneath. That's the job we're about to hand the AI.
Observewhat the data actually says
Perceivethe story under the surface
Transformit into a decision
Rodrigo's "Olhar para o não óbvio — Observe, Perceba, Transforme." Perfect setup for A1: the month-close total looks flat, but the real story is underneath. This is literally what the demo proves.
Track A · Demo 1
Month-close commentary
The challenge: actuals-vs-budget lands, and the write-up comes back late, surface-level, sometimes with errors. We want a forward-looking, board-ready draft — fast.
The intern: your commentary specialist. Onboarded on CTI finance, briefed for the MBR.
This serves the FP&A five — the strongest shared theme. Set it up as a pain everyone in Track A recognizes.
Track A · Demo 1
The prompt
paste your actuals-vs-budget, then:
You are a senior FP&A analyst at CTI, a cruise-terminal developer. Below is this month's actuals-vs-budget for [project]. Write the executive commentary for the MBR: 3–4 short paragraphs for a board audience. For each material variance, give the likely driver, the forward-looking implication, and flag any risk. Be specific, no filler, no jargon. Flag anything you're unsure of rather than guessing. Then give me 3 headline bullets to open the review.
What to notice: it explains why, not just what; it flags uncertainty instead of bluffing; the tone is board-ready. Then push once — "sharpen the risk section" — to show iteration = ownership.
Run it live. Have a known-good result from your dry run in your back pocket in case the room's numbers are messy.
Track A · Demo 1
Make it yours
Save it as a Project with your close checklist + last month's commentary.
Next close, it starts already knowing your business — you edit, not write.
Add your house style once ("we never open with revenue") and it sticks.
The takeaway is the Project, not the one-off. That's what compounds month over month.
Track B · Demo 1
Reconciliation exceptions
The challenge: reconciliations eat the close. You scan hundreds of lines to find the few that don't tie.
The intern: a meticulous staff accountant who only brings you the problems.
Serves Sami, Yessi, Roy, Crow. The reframe: review exceptions, not everything.
Track B · Demo 1
The prompt
paste both sources, then:
You are a meticulous staff accountant. Here are two sets of records: [GL entries] and [bank / subledger]. Match them and return ONLY the exceptions — unmatched items, amount mismatches, and likely duplicates. For each, show both sides, the difference, and your best hypothesis for the cause. Put it in a table sorted by dollar impact. Do not guess a match you're unsure of — list it as "needs review."
What to notice: you review 6 exceptions, not 600 lines; it proposes a cause; it's honest about low-confidence matches.
Plant one clear exception and one duplicate in your demo data so the output is obviously useful.
Track B · Demo 1
Make it yours
Next close, paste your own two sources — same prompt.
Add your materiality threshold as a standing instruction ("ignore < $50 differences").
You go from scanning noise to clearing a short, ranked list.
Emphasize how little changes month to month — the prompt is reusable, only the data swaps.
Optional · Track A
Audience-tailored rewrite
One message, three audiences — "every word matters," and it eats time.
Here's my core message: [paste]. Rewrite it three ways — (1) board: concise, strategic; (2) operations: direct, action-oriented; (3) lender: precise, risk-aware. Keep the facts identical; change only tone, emphasis, and length.
What to notice: one draft → three send-ready versions in seconds; facts locked, only framing shifts.
Run only if you're under ~70 min. Serves Paty and Lory. Otherwise skip and bank the time for Q&A.
Track B · Demo 2
Document intake — W-9 filing
The challenge: new-vendor setup and W-9 filing is pure manual typing — one form at a time, field by field, with the risk of a mistyped tax ID slipping into the vendor master.
The intern: a tireless data-entry clerk that reads each form, files the clean ones itself, and only hands you the ones it can't be sure of.
Serves Crow and Sami. Promoted from "optional" — this is the most tangible "wow" for Track B, and the clearest example of the loop: onboard once, delegate the repetition, review only exceptions. Use fake documents only.
Track B · Demo 2
One W-9 is a trick. The loop is the point.
Extracting one document is a party trick. The real win is the same tiny job, run over and over — document in, clean row out, and anything unreadable gets flagged, never guessed.
1 · Intake
A W-9 lands
PDF or a phone photo, into the chat.
→
2 · Read
AI reads it
Pulls every field from your map.
→
3 · Check
Sanity rules
TIN 9 digits? Signed? Nothing missing?
→
4 · Decide
Clear or flag
Clean → write. Unsure → raise a hand.
→
5 · Write
Row into the sheet
Then back to step 1 for the next.
The clean ~90% file themselves. You review only the flagged exceptions — not everything. That gap is the whole win.
The upgrade from the one-off. Say it out loud: "you're not building a robot, you're onboarding an intern and handing it the same job repeatedly." Ties straight back to the squares — the process runs every line, tirelessly, and surfaces the exceptions your eye would skip.
Track B · Demo 2
Onboard once: the field map is your columns
You think it through a single time. The fields you want are the same thing three times over — your spreadsheet columns, your instructions to the AI, and the shape of every row it returns. Paste this once into a Claude Project and it's remembered forever. No code.
Claude Project instructions — the W-9 field map
You are my W-9 intake assistant. For each W-9 I upload, extract these fields into a single table row, in this order:
Name · Business/DBA · Tax classification · Address · City · State · ZIP · TIN · Signed? · Date
• If a field is unreadable or missing, write NEEDS REVIEW — never guess.
• TIN must be 9 digits; if it isn't, flag it.
• Output the row tab-separated so I can paste straight into Google Sheets.
• After the row, list every field you flagged and why.
Monday, no code: Onboard (make the Project) → Delegate (upload the W-9) → Review & paste the row. Drag in the next document; the loop continues.
The prompt is the whole build — everything else is drag-and-drop. Two upgrades if asked: connect Google Drive/Sheets so it reads from a folder and you skip the paste (still no code); or wire this same prompt to the API for a hands-off folder-watcher (a Future Foundation build, not a workshop task).
Track B · Demo 2
What comes back — a real run
Upload the Gulfstream W-9 (a fake vendor). One row, ready to paste — and the obscured EIN comes back flagged, not invented.
the row Claude returns · tab-separated
Gulfstream Facilities Management LLC │ Gulfstream FM │ LLC · S-corp │ 1450 NW 78th Ave, Suite 210 │ Doral │ FL │ 33126 │ NEEDS REVIEW │ Yes │ 2026-05-30
⚑ Flagged — TIN: the EIN reads 47-38••291; two digits are obscured on the scan, so a 9-digit number can't be confirmed. It writes NEEDS REVIEW instead of guessing. That is what makes it safe to trust with volume.
Your known-good result — keep it in your back pocket in case the live upload is messy. The flag is the money moment: it stops a wrong TIN from silently entering the vendor master. Contrast with a junior who'd type something plausible and move on. Dry-run reference: demo-data/dry-run-reference-outputs.md.
Part Five
Monday
The session only counts if something changes this week.
Shift from "that was cool" to "here's what you do." End on their agency.
Your one action
Pick one challenge you saw today. Open Claude. Run the template.
That's it. One intern, one job. You don't have to boil the ocean — you have to start it.
Make it small and unavoidable. One challenge, one prompt, this week. Offer to share the templates + deck afterward.
The whole thing in one line
AI is an intern. Hire ten. You're the owner.
The spine, one more time. If they walk out repeating this, you won.
"Set up your first Project" — a one-page starter so your intern remembers.
Your personal challenge, from the warm-up — the obvious place to start.
Want a follow-up on your specific workflow? That's what the 1:1s are for.
Tease phase-2: the persistent monthly-commentary assistant (the sponsor's dream hand-off) is a real build, not just a prompt.
CRUISE TERMINALS INTERNATIONALStrategic AI Masterclass
Thank you. Now — what would you hand your intern first?
Rodrigo Faerman · Leader in Strategy & AI · LinkedIn / IG @rodrigofaerman · Facilitated by Future Foundation
Open Q&A on their terms — the closing question invites them to name a real task, which is the perfect segue to discussion and 1:1 follow-ups. Point them to your socials for follow-up.
