Confidential · Working Document
14 May 2026
A Midas × ECAL workshop proposal

AI Upskilling
for ECAL.

A tailored three-week program for the Hong Kong team.

From Midas Analytics
×
For
Eurizon Capital Asia Limited
Eurizon Capital Asia Limited
Hong Kong · SFC Type 1 · 4 · 9
Global AI literacy distribution — 84% have never used AI
Feb 2026
01 The bar today

AI literacy is on the floor.

Globally, 84% of people have never used AI. Pay-for-AI users are rounding-error. Even in HK financial services, sustained workflow use is far rarer than headcount surveys imply.

~6.8B (84%) have never used AI
~1.3B (16%) use a free chatbot occasionally
~15–25M (0.3%) pay USD 20/mo for AI
~2–5M (0.04%) use AI in actual coding / workflow
What this means for ECAL

Three weeks puts the entire team into the top 0.04% — building, not just chatting.

AI theoretical capability vs observed usage across job categories
Anthropic Economic Index · 2025 · radar chart
02 The opportunity at scale

ECAL's work sits in the bullseye.

  • Blue area: what AI could already do across job categories today.
  • Red area: what people are actually using it for.
Business & Finance · Research
~95% vs < 15%

ECAL Investment + Research sit here.

Sales · Business Development
~90% vs < 18%

Pitches, RFPs, client comms, prospecting.

Legal · Compliance · Office & Admin
90%+ vs < 20%

Compliance, Risk, Operations.

What this means for ECAL

Every one of ECAL's seven seats in scope sits in a high-capability / low-usage cell. The runway is years long — and the firms that compound first compound fastest.

03 Why this conversation now

Three forces. One window.

01

The shift is real.

3 in 4 HK financial firms have a GenAI use case live or in pilot.

02

Regulators caught up.

SFC GenAI circular · PCPD guidelines · EU AI Act. Doing nothing is no longer neutral.

03

The gap is real.

80% of HK financial institutions flag GenAI skills as a top gap.

Sources: HKIMR (2024) · SFC GenAI Circular (Nov 2024) · PCPD AI Guidelines (Mar 2025).

04 What we heard about ECAL

Where ECAL sits today.

Parent AUM€583 B
Founded in HK2015
SFC licencesType 1 · 4 · 9
In scope7 seats across 6 functions

HK subsidiary of Eurizon Capital SGR (Intesa Sanpaolo Group), focused on Asian equity markets.

Cross-country and regional sector coverage; on-the-ground China expertise via Putonghua-speaking analysts and the Penghua JV.

SFC-licensed under Type 1, Type 4, and Type 9 — AI use touches dealing, advising, and asset management activities, the three areas where the SFC circular has the most teeth.

This workshop scopes seven seats across six functions — the senior layer ECAL has prioritised for upskilling. Small enough that AI force-multiplication has outsized impact, varied enough that one workshop format won't fit all.

Milan Hong Kong Shanghai Beijing
Information flow that's central to ECAL's mandate.
05 The opportunity

How AI re-invents your day-to-day.

One slide, one promise per function. AI is not generic productivity — it speaks to what each of you actually does.

Investment & Research

From hours of reading to minutes of judgement.

Chat-with-documents on 10-Ks, A-share filings, SFC handbooks; first-pass summaries on Asian equity news; structured extraction of risks from Chinese-language footnotes. The leg-work compresses; your judgement layer is untouched.

Before: 3 hrs / filing  ·  After: 20 min
Sales · Business Development

RFPs answered in hours, not days.

RFP parsing + draft responses against your prior wins. Pitch-deck variants for different LP archetypes. Pre-meeting briefings on prospects. The same talent, but with three more pitches in the funnel each month.

Before: 2 days / RFP  ·  After: 3 hrs
Compliance

Handbook lookups answered live.

SFC handbook + circular Q&A on tap. Regulatory news monitor running between sessions. AI-policy templates ready for the parent group's review. You become the team's regulatory search engine — instantly.

Before: 45 min / question  ·  After: 2 min
Risk

Scenario narratives, on demand.

Stress-scenario narratives that translate quant outputs into Risk-Committee-ready prose. AI brings the structure; you bring the judgement. The Milan side reads cleaner papers, faster.

Before: 1 day / scenario  ·  After: 2 hrs
Operations

Reconciliation, surfaced before it hurts.

Calendar and to-do AI agents. Reconciliation flaggers running overnight. Process-flow audits driven by 5 Whys — done with AI as the patient interviewer. Less firefighting, more shaping.

Before: weekly fire-drills  ·  After: caught at 8 a.m.
Marketing

Factsheet commentary, drafted weekly.

Monthly factsheet commentary, LinkedIn variants, content idea generators tuned to ECAL's investment narrative. Your voice, multiplied across channels.

Before: 1 channel / week  ·  After: 4 channels
06 Our proposed approach

A three-week working program, on site.

Three in-person sessions of three hours each, at ECAL's office. Designed around your team and your business. Each session builds on the last.

1
Week 1

Your AI superpowers.

Start building your AI superpowers. Prompt frameworks, the four superpowers, tailored to asset-management workflows.

2
Week 2

AI Assistant.

Everyone walks away with their own AI Assistant — a persistent agent that runs between sessions and works the way you work.

3
Week 3

Build your own tools.

Use AI to build small, specific tools that make your day-to-day easier — tailored to what you actually do.

All content is tailored to ECAL's mandate and your team's functions. This document is a starting point — every element is open for discussion.

07 Program at a glance

Three weeks. Three sessions. Seven people. One AI-native team.

Week 1 Week 2 Week 3
Theme Your AI superpowers. AI Assistant. Build your own tools.
Tools
Manus + Claude Gemini ChatGPT
Hostinger VPS + Claude API Gemini API Moonshot AI
Claude Code + GitHub
Format 3-hour workshop, on site 3-hour workshop, on site 3-hour workshop, on site
Core concept Individual fluency with consumer AI A persistent agent working for you Custom tools, built with AI
Hands-on Function-specific drills Each participant ships their own agent Each function prototypes one tool
Take-away Personal prompt library Working AI assistant on Telegram One working tool per function
Individual
Persistent
Programmatic

▲ A VPN is strongly recommended for all participants — some required services are restricted in mainland HK and we want zero friction during the workshops.

08
Workshop 1 · Week 1

AI superpowers for asset management.

The four "superpowers" framework, applied to asset management. Each function leans into the superpower most relevant to their day.

1 · Writing Wizard

Stakeholder Communications

30 + 20 min

Drafting IC memos, factsheet commentary, RFP responses, regulatory filings, client letters. Exercise: turn a sloppy bullet briefing on a portfolio change into a polished IC-ready memo using C.R.A.F.T.

Best for Research BD Marketing
2 · Research Rocket

Investment & Regulatory Research

30 + 20 min

Chat-with-documents for parsing 10-Ks, A-share annual reports, regulatory handbooks. Live web research for Asian market intelligence. Exercise: extract the key risks buried in the footnotes of a real Chinese-listed-company filing in under 10 minutes.

Best for Research Compliance Risk
3 · Creative Muse

Marketing & Business Development

30 + 20 min

LinkedIn content, pitch-deck refreshes, client newsletters, brainstorming campaign angles for institutional audiences. Exercise: develop three pitch-hook variations for a hypothetical new China A-share fund targeting European institutional allocators.

Best for Marketing BD
4 · Strategic Thinker

Risk, Compliance, and Operations

30 + 20 min

Business frameworks (SWOT, 5 Whys, scenario planning) applied to operational and risk questions. Exercise: build a risk matrix for "ECAL's potential adoption of AI in regulated activities" against the firm's governance principles.

Best for Risk Compliance Operations
Opening
SP 1
Break
SP 2
Break
SP 3
SP 4
Closing
3-hour session flow
09
Workshop 2 · Week 2

AI Assistant. Working for you, between sessions.

A hands-on build session. Everyone walks away with a personal AI agent that runs between sessions. We provide the vehicle; you decide what it does.

The main concept · in plain language

An AI agent is just an AI you can talk to from a single place (Telegram, email, your phone) that remembers context, runs while you sleep, and does small jobs — like reading the news, summarising filings, drafting a reply, or flagging something to your attention. No code on day one. You decide what jobs it does.

Use cases — pick yours
Compliance

Regulatory news watcher.

Scans HK FSI news daily; pings you when something material lands.

Research

Morning market brief.

A 60-second summary on your covered names, in your inbox at 7 a.m.

BD

Meeting-prep agent.

Tomorrow's prospect, briefed: latest news, recent moves, public LP comments.

Operations

Personal task assistant.

Calendar, to-do, follow-ups — chased and re-surfaced before they slip.

We give you a working agent. You decide what it does next. Each function takes it home and pushes — that's where the real value compounds.

Opening
Concept
Break
Pick use case
Break
Hands-on build
Demos
Closing
3-hour session flow
10
Workshop 3 · Week 3

Build your own tools.

By Week 3 the team understands prompting and has a personal agent running. Now we move from using tools to building them. Claude Code lets non-developers commission small, specific tools that solve specific recurring problems.

Track A · Everyone · 30 min
Understanding what Claude Code is, what it isn't, and how to scope a request well. The grammar of a good build prompt. When something should be a tool vs a one-off chat. The "prototype → production" path inside ECAL's governance.
Track B · Hands-on prototyping · 2 hours · public / synthetic data only
Research
A-share filing summariser
Compliance
Regulatory handbook lookup
Risk
Scenario stress narrator
Operations
Reconciliation flagger
BD
RFP question parser
Marketing
Factsheet commentary drafter
Prototype → Production path
Sign-off › Validation › Monitoring
Opening
Track A · Scope
Break
Track B · Build your tool
Demos
Closing
3-hour session flow
11 Per-function tailoring

You bring the expertise. AI puts wings on it.

These are illustrative use cases — possibilities, not prescriptions. Midas plays tech enabler: we surface what AI can do; you connect it to your subject-matter judgement. The leap from idea to action shrinks from days to seconds.

You are the expert.

Your domain knowledge — markets, regulations, clients, ops — is what AI doesn't have.

×

We bring the capability.

What AI can do today, and where the boundary sits. Three weeks of structured exposure.

=

Turbo brain.

Ideas convert into actions in seconds, not days. Productivity uplift compounds.

Illustrative starting points — every row is a conversation, not a contract.
FunctionWeek 1 · anchor exerciseWeek 2 · sandbox agentWeek 3 · capstone tool
Research A-share filing risk extraction Morning briefing agent A-share filing summariser
Business Development Pitch hook drafting Meeting-prep agent RFP question parser
Compliance Risk matrix on AI adoption Regulatory news monitor Regulatory handbook lookup
Risk Scenario planning on a portfolio shock Market-stress watcher Scenario stress narrator
Operations Process-flow audit using 5 Whys Personal task assistant Reconciliation flagger
Marketing Multi-channel content for a fund launch Content idea generator Factsheet commentary drafter
12 Tools shopping list

What you'll need, week by week.

One simple checklist. Mix-and-match where we offer choice. We'll help everyone get set up before Week 2.

Week 1

AI superpowers — pick your daily driver.

Manus Primary · agentic browser AI
+ one of
ClaudeAnthropic
GeminiGoogle
ChatGPTOpenAI
Free tiers work for the workshop. Paid tier (USD 20/mo) recommended if you plan to keep using them.
Week 2

AI Assistant — VPS + an LLM API.

Hostinger VPS Where your agent lives
+ one LLM API
Claude APIBest quality
Gemini APIMost generous free tier
Moonshot AICheaper · CN-friendly
We help everyone get the VPS + API spun up during the session. ~USD 10/mo VPS · pay-per-token API.
Week 3

Build your own tools — Claude Code + GitHub.

Claude Code Required · for everyone
+
GitHub repo Where the code & materials live
A personal GitHub account is free. Each participant gets a repo for their capstone.
▲
A VPN is strongly recommended for the whole program. Some services and APIs are restricted from HK / mainland networks — a VPN keeps the workshop friction-free.
13 Between sessions

What we expect from you. And what we're here for.

The value of three weeks compounds only if the team pushes between sessions. Our job is to remove every friction we can — yours is to stay curious.

What we ask of you

Push beyond the workshop.

  • ↗
    Stay curious. Explore beyond what we present in each session — try AI on the tasks that bother you most.
  • ↗
    Bring questions. The dumbest question is the one you don't ask. Surface what isn't clicking.
  • ↗
    Apply your judgement. AI is the engine; your subject-matter expertise is the steering wheel. Don't outsource that part.
  • ↗
    ~30 min between sessions. Light, function-specific homework. Wk 1: try C.R.A.F.T. on three recurring tasks. Wk 2: run your agent for a full week. Wk 3: iterate the prototype once.
Always-on support

WhatsApp group · the full three weeks.

  • ●
    Any question, any time. Stuck on a prompt? Agent acting weird? Drop it in the group — we triage live.
  • ●
    Between-session nudges. Bite-size tips and worked examples that build on the week's session.
  • ●
    One-to-one if needed. A 15-minute call if something deserves a deeper look than a chat thread.

The team that pushes hardest between sessions gets the steepest curve. We make sure the door is always open.

14 Logistics & pre-flight

What we'll need from ECAL to make this work.

Schedule

Three weeks, ideally consecutive. Each session 3 hours, ECAL's choice of working-hours or after-hours start. We can adapt to a 4-week format if a buffer week between Weeks 2 and 3 is preferred.

Venue & setup

On-site at ECAL's office. A room that holds 7 participants + 1–2 instructors comfortably. Whiteboard or large display. Reliable internet — see below.

Pre-flight · IT & Compliance
  • Per your IT / InfoSec team — are tools like Claude, ChatGPT, Gemini and Perplexity available in your environment, or is there an alternative you'd like us to teach against?
  • For Week 2 — a Hostinger VPS + an LLM API per participant. We handle setup; we just need a green light to register on workshop day.
  • A VPN, ideally team-wide, to keep the workshop friction-free.
15 Investment

Three sessions. One transparent price.

Per-session pricing. No retainer, no surprises. Optional R&D tax credit cuts the effective cost in half — detailed on the next slide.

Session 01
HKD 30,000
3-hour workshop · on site
AI Superpowers for Asset Management
Session 02
HKD 30,000
3-hour workshop · on site
Your AI Executive Assistant
Session 03
HKD 30,000
3-hour workshop · on site
Build Your Own Tools (Claude Code)
Total program fee HKD 90,000
After HK R&D credit · DLRI No. D041 ≈ HKD 45,450 effective −50%
Billing
50% on kickoff · 50% after Session 03
Includes
All materials, deliverables, prompt libraries, six function-specific tool prototypes, summary memo to senior management
Excludes
Third-party tool subscriptions · venue (on site at ECAL)
16 Financial incentive

Engaging Midas cuts cost in half.

The program qualifies for the HKSAR Inland Revenue Department's enhanced R&D tax deduction. Midas Analytics is a certified DLRI — No. D041.

0% tax deduction on qualifying R&D expenditure under Cap. 112.
✓ Req 01

DLRI-certified payee.

Midas Analytics — D041, valid through Jul 2026.

✓ Req 02

R&D in HK, new tech.

AI workflows for ECAL, delivered out of HK.

Program feeHKD 90,000
300% deductionHKD 270,000
Savings @ 16.5%− HKD 44,550
Effective costHKD 45,450
−50% Effective discount
DLRI Instrument of Designation — Midas Analytics Limited, No. D041 Click to enlarge
ITC · DLRI Cap. 112 DIPN 55
17 Open questions for ECAL

Open questions we'd like to align on with you.

  • Are three weeks the right cadence, or would a 4-week version with a buffer feel better?
  • Does the per-function tailoring map capture each role's real top priority, or should we re-weight?
  • Are there sensitivities from the parent group we should be aware of before designing the final detailed agendas?
18 From here to kickoff

Three steps to start.

A simple path from your thumbs-up to Week 1.

Step 01

Your green light.

A short "let's do this" from ECAL confirms the engagement and the scope of the three sessions.

Step 02

NDA & SA.

Midas drafts and circulates the NDA and the service agreement. You sign.

Step 03

Kickoff & Week 1.

We agree the calendar for the three sessions; pre-flight IT items settled; Week 1 lands on site.

Michele De Filippo
Michele De Filippo, PhD
CEO · Midas Analytics
michele@midasanalytics.ai midasanalytics.ai LinkedIn

Thank you.

Looking forward to building, with the ECAL team, an AI-native operating muscle that compounds long after Week 3.

midasanalytics.ai
01/19
←→ navigate