AI Pathfinder
The model specifications
Who we are
Artificial Intelligence
firm
Amplifying human potential through AI — for banks, asset managers and other regulated enterprises.
Belief 01 / 06
Human amplification is key to unlocking AI value
We combine human and artificial intelligence to boost human performance
Belief 02 / 06
Co-creation brings together the best possible AI
We combine your industry & process knowledge with our deep expertise in AI
& process
knowledge
AI expertise
Integrated
Amplifying AI
Belief 03 / 06
Combine strategic vision with employee involvement
We create an AI journey driven by trust, excitement, and meaningful collaboration
Belief 04 / 06
A successful AI journey starts with employee engagement
Embracing AI together, transforming scepticism into enthusiasm
Belief 05 / 06
Incremental innovation is very powerful
Custom plug-ins that enhance the business step by step
Focus
Belief 06 / 06
Not every AI should be built internally
Non-strategic, non-aligned AI can be bought off the shelf — Gemmo can guide this decision
Human amplification through AI
AI journey
The Gemmo AI journey is built on three pillars
Identification of AI opportunities tailored to your organisation, grounded in real team needs.
Co-implementation of AI agents with your team, ensuring alignment and ease of integration.
AI agents must be continuously monitored, maintained, and governed.
Track record
Gemmo in numbers
Track record
Three areas of AI capability
Computer Vision
Automate complex processes that require high-level understanding of an environment and interacting objects.
Computer vision powered by deep learning pushes the boundaries of product innovation across industries.
Predictive AI
Support data-driven decisions and save resources by looking into the future.
Deep learning combined with proprietary, unique and large-scale datasets unlocks value like nothing else.
Generative AI
Boost productivity with the power of custom large language models.
Assistive technology that speeds up everyday activity across every department.
Track record
Knowledge workers we amplified
Case study · Financial services
Financial services — client retention
An AI copilot for every wealth manager: client health, sales signals and next best action
Manager
Wealth Manager
Case study · Life sciences
Pre-clinical trials
Assessing animal response to new drugs with automated behaviour tracking
Researcher
Outline
Today's discussion
Start with the end
in mind
WHAT DO I GET FROM AIP?
Scoreboard for a typical Asset Man. Client
The view from the top
The whole portfolio on one chart
The view from the top
Every use case carries a score
The view from the top
The score becomes a rank
The view from the top
The rank becomes three paths
The view from the top
Path 1 sits where the pattern is strongest
THE AI PATHS
Path 1 — the foundation
THE AI PATHS
Path 2 — scale on the same rails
THE AI PATHS
Three paths, 33 months
End of Path 1
Path 1 — the final list, ranked
End of Path 1
Re-sorted by department
End of Path 1
Department is the first grouping
End of Path 1
Sub-department is the second
End of Path 1
Themes tell us what to build once
End of Path 1
Four delivery tracks fall out of the grouping
View from the top
What has been done
Three paths, each with its own risk and return profile
Foundation
Low effort, low risk. Builds the shared data, MLOps and tooling groundwork, and delivers the early standardisation wins that fund the paths behind it.
Scale
Growth-oriented; medium effort; balanced benefits; enables scaling without a proportional headcount increase.
Optimise
High effort, high reward; maximum efficiency and cost reduction; strong standardisation; little focus on basis points.
How it
works
Use case selection
Pillars of the use case selection
Grounded in your priorities: ROI, risk, and regulatory compliance
Use case selection
Pillars of the use case selection
Grounded in your priorities: ROI, risk, and regulatory compliance
+
⅓ efficacy A deliberate bias to lower-risk delivery
Use case selection
Pillars of the use case selection
Grounded in your priorities: ROI, risk, and regulatory compliance
+
⅓ efficacy A deliberate bias to lower-risk delivery
proof Explainable, human in the loop
What will be your
AI legacy?
You can tune the AI ROI
Tune the AI ROI
Three types of return, three dials
Where you set them decides which use cases come first
Tune the AI ROI
Three types of return, three dials
Where you set them decides which use cases come first
Tune the AI ROI
What each dial is made of
The signals we score a use case against
Tune the AI ROI
The three dials add up to your ROI
Our task is to find the optimal mix for Mediolanum
Risk definition & factors
How we define risk
ROI assessment
How we define ROI
Implementation risk
How we break down risk
risk = Norm ( AI modelling risk + Data risk + Adoption risk , 1, 10 )
Government & regulation risk
How we score the regulatory side
risk = Norm ( Risk score , 1, 10 )
Risk calculation
How we calculate and weight risk
Reputation and trust are more important than anything else
risk ×
risk
ROI assessment
ROI becomes the vertical axis
Standardisation, cost savings and basis points are scored, summed and normalised onto a single 0–10 return scale.
Risk calculation
Risk becomes the horizontal axis
Two axes
Axis 1 — return
Return is the first thing we score. On its own it tells you what a use case is worth — not what it costs you to get there.
Two axes
Axis 2 — risk
Two axes
Combine the two axes
Two axes
And this is what your portfolio looks like
Running the
Pathfinder
Introduction
Luca Marchesotti, Ph.D.
AI maturity
Accelerate the AI journey
Human amplification through AI
AI journey
The Gemmo AI journey is built on three pillars
Identification of AI opportunities tailored to your organisation, grounded in real team needs.
Co-implementation of AI agents with your team, ensuring alignment and ease of integration.
AI agents must be continuously monitored, maintained, and governed.
AiP final recommendations
AI opportunities qualified for the next phase
Phase 2: filter AI opportunities through the Pathfinder pipeline
AiP final recommendations
Aggressive focus on business impact prioritised 21 opportunities
Phase 2: filter AI opportunities through the Pathfinder pipeline
AiP final recommendations
Aggressive focus on business impact prioritised 21 opportunities
How impact is scored — every opportunity gets one of three grades
AiP final recommendations
Aggressive focus on business impact prioritised 21 opportunities
Only "good" and "great" impact qualifies for the next stage
AiP final recommendations
Two AIOs discarded because too onerous
Phase 2: filter AI opportunities through the Pathfinder pipeline
AiP final recommendations
One AIO discarded because too risky
Phase 2: filter AI opportunities through the Pathfinder pipeline
AiP final recommendations
What is the AiP output?
Three outputs to start implementing AI
Working together
From your perspective: a 5-step process
The interviews
What is a workgroup?
A group of people sharing the same pains, gains and aspirations when it comes to AI
The interviews
Who to invite?
Mix perspectives on purpose — the interviews work best on contrast
The calendar
Interview and catch-up weeks
The interviews
The workgroup interviews: part 2 — new use cases
AiP deliverables
AI Pathfinder: what do you get?
Clarity on the AI path
Change management
Guidance on AI deployment
Let's pick the
first path together
Block C, Dublin, D02 VK60
22100 Como