FINANCE, DATA AND AI · FULLY REMOTE
Your board wants AI in the finance function. Your data isn't ready.
Most mid-market companies run finance on spreadsheets and disconnected systems, then wonder why every AI pilot stalls. I build the data foundation first, then the FP&A that runs on it. Ex-AWS EC2.
monthly-close · source status
From 12 days to 4| Source | System | Status | Days |
|---|---|---|---|
| Accounting | ERP | Connected | 1 |
| Billing | In-house system | Connected | 1 |
| Operations | Spreadsheets | Manual | 6 |
| Banks | Statements | Manual | 4 |
Monthly close once all four sources are connected
4 days
5 years at AWS leading financial planning and pricing for EC2
USD 35B cost base
Monthly close cut 30% with Python automation
IF THIS SOUNDS FAMILIAR
The problem isn't your team. It's the foundation under it.
01
The close takes weeks
The numbers live in spreadsheets, inboxes, and one person's head. Every month-end is archaeology.
02
Your systems don't talk to each other
Accounting says one thing, billing says another, ops has its own version. Nobody sees a single source of truth.
03
AI pilots keep stalling
The tools work. Your data doesn't. Every AI initiative dies on the same rock: nobody built the foundation first.
04
Reports describe the past
You learn what happened last month, three weeks late. You never see what's coming. That's reporting, not finance.
THE APPROACH
Foundation first. Then the finance that runs on it.
Most FP&A advice assumes clean data. Mid-market reality is the opposite. I start where the numbers actually live, build the pipes, then the models. In that order, because the other order doesn't work.
01
Map where the data lives
Accounting, billing, ops, spreadsheets. We inventory every source, what's reliable, what's broken, and what the business actually needs to see.
02
Build the foundation
Your finance data connected into a clean cloud stack. One automated reporting pack: monthly close and board financials that build themselves.
03
Run the cadence
Monthly close, forecasts built from real drivers, budget vs. actuals, board pack. AI tooling where the data can now support it. Direct access, no hand-offs.
TWO WAYS TO ENGAGE
Two ways to work together.
The core work: build the finance data foundation, then operate it. I recommend exactly one per conversation, the one that matches your problem.
Finance Data Foundation Sprint
USD 5,000–10,000
4 to 6 weeks · fixed price
Full inventory of where your finance data lives. Core migration into a clean cloud stack. One automated reporting pack that replaces the spreadsheets you build by hand every month. An AI-readiness roadmap: what your data supports today, what needs fixing first. Fixed price after one scoping call.
Ideal if
- Your close runs on spreadsheets and takes weeks.
- AI is on the board agenda but every pilot stalls on data quality.
- You want a working foundation before committing to ongoing finance support.
Fractional FP&A
USD 8,000–12,000/mo
Ongoing · maximum 2 clients
Ongoing finance leadership that runs on the foundation. Monthly close support, board deck financials, forecasting built from usage and pricing drivers, budget vs. actuals, scenario models, fundraise prep. For clients with material cloud or AI spend: cost unit economics and spend governance, the AWS EC2 specialty.
Ideal if
- You need a senior finance operator, not a generalist.
- You can't justify a full-time VP of Finance yet.
- Your reporting exists but nobody turns it into decisions.
Maximum 2 fractional FP&A clients at a time. Direct access, no associates.
ALONGSIDE · AI ADOPTION
What if the problem is the time, not the numbers?
The same data foundation that makes FP&A work is what makes AI work. If you'd rather start there, two smaller and faster ways in.
AI Assessment
USD 990
One-off · one 45-min call
A recorded 45-minute interview and a written report, in your hands within 5 business days. You leave with an effort-versus-impact matrix, 3 to 7 tools with cost and setup time, and one page of financial impact. If we don't find at least 5 recoverable hours a week, you don't pay.
Ideal if
- You know time is being lost but not how much or where.
- You've tried scattered tools and nothing stuck.
- You want a plan before committing money to implementation.
AI Mentoring
USD 1,200–3,000/mo
2 sessions a month, or 1 hour weekly
Ongoing support so your team actually adopts AI, rather than for three weeks. In the light version we review what you built and what comes next, twice a month. In the intensive version I'm in the room building with you, one hour a week. Price follows intensity, not positioning.
Ideal if
- You've done the assessment and want to hold the pace.
- You have someone internal who can execute, or you need someone who will.
- The licenses are already paid for and nobody uses them.
The assessment fee credits toward implementation if we continue.
ONE AI INTEGRATION A WEEK
Every Tuesday I publish an integration I actually use.
The job it kills, what it costs today in hours, the steps, the prompt to copy and paste, and what it actually saved. Including where it broke. Published in Spanish, in under three minutes of reading.
Get it on Tuesdays
One a week. No filler. Unsubscribe in one click.
BACKGROUND
Where the methodology was built.
$35B
Cost base managed at AWS EC2
30%
Monthly close reduction through automation
15
Years across finance, operations and technology
AWS EC2
Amazon Web Services
Led financial planning and pricing for EC2 capacity, the infrastructure behind the AI buildout and one of the largest cost lines in tech. Built the pricing frameworks behind EC2's largest cost-savings programs and Python automation that cut monthly close by 30%.
Current Fractional Work
Consumer electronics distribution
Leading the cloud migration and finance data build for a distributor operating across Argentina, Mexico, and the US. Databases, automated reporting, and FP&A rebuilt from the inside, alongside the monthly close and forecasting.
Automation stack
Health and sports performance
Fractional CFO and COO for a profitable operation running entirely through WhatsApp and Instagram, with no CRM and no structured data. Built the cash model, the P&L tracker, the client database and weekly KPI reporting, plus the automation stack design on n8n, the WhatsApp Business API and the Claude API.
ABOUT
Finance from inside the machine.
I ran financial planning for EC2 capacity at AWS, the infrastructure behind the AI buildout. My whole career is turning messy cost data into operating models leadership can plan against: at AWS that meant one of the largest infrastructure cost bases in tech, today it means building that same capability inside mid-market companies. Foundation first, then the finance.
15 years across tech, finance, and operations. Goldman Sachs, PE-backed tech companies, Amazon.
2 clients at a time. Direct access. No associates. Fully remote, any time zone.
LET'S TALK
If your finance function runs on spreadsheets and one person's memory.
First call is 30 minutes. No pitch deck. A direct conversation about where your finance data lives and what a foundation would take. If there's a fit, we scope the sprint. If not, you leave knowing exactly where your data problem is. That's not nothing.
or email agustin.lastra@silverandes.com