Maor × Angel
discovery call

29 May 2026 Bali · 15:00–16:00
01 what I heard

Let me play back what I think I heard.

01
The symptom: workflow going in circles. Versions, context, alignment.
02
The team: lawyers, accountants, valuator already cover the substance.
03
The gap: core analysis is there. Technical structure around it is what's missing.
04
The stakes: high. Submission quality matters. Money matters.
05
The wish: consistent high-quality output, without redoing things every time.
if we worked together · the goal would be

A workflow you run yourself, where Claude handles the repetitive heavy lifting — and you + your team focus on the judgment calls.

02 working with AI

Three basic principles, for any AI work.

01
Be specific

Vague prompts get vague answers. "Improve this" ≠ "find 3 unsupported claims in section 2."

02
Verify, don't trust

AI sounds confident even when wrong. Treat every output as a first draft. Always verify numbers, citations, names.

03
Iterate, don't restart

First answer is rarely the best. "Try again, but tighter." Refinement > starting over.

honest about quality
AI errors don't go away.
Good systems catch them earlier.
Your lawyers' and valuator's final review still matters — that's not changing.
03 where information lives

Imagine your physical desk. Three areas, three jobs.

The bookshelf · always there, rarely opened
Reference material Claude can refer to anytime, doesn't need to re-read.
The desk surface · today's work, clear after
The 3–5 documents in your hand right now. Done → close the chat → next task starts fresh.
The notebook · what today taught me
End-of-session summary, small enough to become tomorrow's bookshelf.
takeaway Today's notebook = tomorrow's bookshelf. The system gets sharper every week, not heavier.
04 how I'm set up · briefly

My own architecture — 4 levels, builds top-down.

Same skeleton as yours will be — just bigger. My stack: Claude Code + Notion as my visual index. For non-technical users, Notion alone handles most of what I do.

my OS
Cowork OS
Claude Code + Notion · loads top-down
0
level 0 · always loaded
ROOT
CLAUDE.md MEMORY.md memory/ Notion 3-DB
1
level 1 · domain groupings
Workstations
career-hq/ business-hq/ content-hq/
2
level 2 · per-project
Projects · 18+
color = parent workstation
jobs-pipeline career-tracker aus-job freesoul-website content-analytics contacts-pipeline jiwa-surf-content personal-finance + 10 more
3
level 3 · automation
Skills & Rituals
/log-decision /ship-it /project-checkup Friday cron
takeaway Four layers, each doing one job. The system works because nothing tries to do two jobs at once.
same architecture · different proportions

The skeleton stays.
Most of the weight shifts.

Yours isn't a portfolio with 18 things. It's one deep mission with a finite end. Same four layers. But most of mine simplifies dramatically — and the depth lands where it matters for you.

06 how yours could look

Your setup — built for one deep mission.

Same four-level shape as mine, but everything simplifies. The exact files, prompts, and configurations are what we'd design together if there's a fit.

your setup
The Valuation Case
single mission · finite end
0
level 0 · always loaded
ROOT
your baseline + running state
1
level 1 · domain
One workstation
your case material, where it already lives
2
level 2 · the depth
A few focused projects
each one does one job, output feeds next
3
level 3 · daily practice
A small toolkit
5–6 saved prompts + a few weekly rituals
07 what to take away

Three things — even if we don't work together.

01 · today's quick win

Build a document index — let AI find before it reads.

Massive doc loads thin Claude's attention fast. Build a small index (Notion DB or even a spreadsheet) with topic · summary · category · file. AI scans the index first, then reads only the doc that matters. Speed + accuracy win, no architecture needed.

02 · the structural fix

Separate concerns — three places, focused units.

Reference shelf · today's desk · notebook. One job per unit. This alone catches most of the friction.

03 · setting expectations

AI accelerates, humans verify.

The system catches errors earlier — but the final judgment stays with you and your team. Set this expectation early.

08 what's next

What happens after today.

Today is just a discovery call — no commitment from either side. Here's the process I'd suggest after this.

Step 01
Today's call

You walk me through your current workflow. I share how I'd approach it. We figure out together if there's a real fit.

60 min · on me
Step 02
I send a proposal

If I can genuinely help, you get a written proposal — scope, options, pricing, timeframe. Take your time deciding.

early next week
Step 03
If it's a fit, we start small

We'd begin with a foundation phase — understanding your actual setup, mapping your evidence, putting the structure in place. We figure out timing based on your data volume and deadline.

small foundation
Step 04
If not, I say so

If after today I'm not the right person, I'll tell you honestly and point you toward someone better placed.

no pressure
FREE SOUL Angel Chen · FREE SOUL 如魚得水