01 Investing AI

The Investing AI Agent Playbook

A strategic map for launching and scaling AI across digital investing experiences. It helps wealth, product, CX, and AI transformation leaders deploy faster, scale with confidence, and turn investing AI into measurable business value.

  • SpareBank 1

    odeabank

    CIO GROUP

  • SpareBank 1

    odeabank

    CIO GROUP

Playbook

Explore this topic with AIOpen ChatGPT

5.1 Brand and tone

When a customer asks an investing assistant a question, they judge more than the answer. They notice whether it sounds like their bank, whether it uses familiar product language, whether it gives the right amount of detail, and whether the interface feels native to the app they already trust.

That is why brand and tone are not just finishing touches. In wealth and investing, tone is part of the product boundary. It helps the user understand what the assistant can do, what it cannot do, when it is being educational, when it is using portfolio data, and when a human should be involved.

Charlie is built to be configured around the institution. The name, tone, language, topic scope, answer length, supported tools, visual treatment, and deployment model can all be adapted. If Charlie is embedded through the SDK, the experience can be fully white-label and shaped in detail inside the bank's own product. If Charlie is used through API or MCP, the bank can keep more of its own interface while using Charlie as the investing intelligence layer behind it.

Charlie in FR

For clarity, we left the UI (tools and resources consulted) in English.

Tone is configuration, not a single prompt

A prompt can say "be professional and friendly". That is not enough for a regulated investing product.

A real brand and tone setup needs to define how the assistant behaves across many moments: a calm portfolio explanation, a market-news question, an empty-portfolio onboarding flow, an order step, a blocked request, a missing-data state, a support handoff, and a long educational answer for someone who asks for more detail.

This is also where brand, UX, and compliance meet. The FCA Consumer Duty puts strong emphasis on clear, timely, accessible information and on tailoring communications to customers' financial literacy. ESMA's guidance on AI in investment services reminds firms that AI does not remove MiFID II obligations, including acting in the client's best interest. So the question is not only "does this sound on brand?" It is also "does this help the customer understand the right thing at the right time?"

What should be editable

The useful standard is simple: if the investor can see it, hear it, read it, click it, misunderstand it, or infer it from the conversation, it should be configurable.

Area

What should be configurable

Example

Assistant identity

Name, avatar, greeting style, first-person or institution-led wording

"Hi Kris" versus "Good morning, Kris"

Formality

From very formal and institutional to warm and conversational

Private bank tone versus challenger bank tone

Message length

Short answers, detailed explanations, expandable detail, summary-first answers

"Give me the quick answer" versus "Explain the drivers"

Reading level

Plain-language defaults, glossary support, jargon policy

"Bond prices fell" before explaining yield mechanics

Language and locale

Supported languages, local spelling, date formats, currency formats

en_GB, fr_FR, nl_BE, local market terms

Product terminology

Bank-specific names for accounts, portfolios, plans, mandates, risk profiles

"Investment plan" versus "portfolio strategy"

And the editable surface should not stop at text.

Area

What should be configurable

Why it matters

SDK interface

Colours, typography, spacing, cards, widgets, input states, buttons, loading states

The assistant should feel native in the bank's own app

Conversation structure

Bullets, tables, citations, follow-up suggestions, source display, detail toggles

Different banks have different clarity standards

Topic scope

What Charlie can discuss, what it should decline, what should route to support

Brand voice should not widen the advice boundary

Compliance wording

Disclosures, risk language, advice deflection, source phrasing, handoff language

Controls should be explicit and reviewable

Journey behaviour

Welcome, onboarding, portfolio review, order flow, market volatility, empty portfolio

The right tone changes with the moment

Operating model

Who approves tone changes, when settings are reviewed, how changes are logged

Brand settings become production controls

Formality is not one slider

Formality is often treated as one setting. In practice, it is a group of decisions.

A financial institution might want Charlie to sound very precise when explaining risk, warmer when helping a beginner understand an empty portfolio, and short and factual when an order flow is in progress. Another institution might prefer a more coaching style throughout the journey, but still require strict language around suitability, execution, and portfolio losses.

Here is the same idea in a few different styles.

Setting

Example response style

Short and formal

"Your portfolio is down 1.8% today, mainly due to equity exposure. I can show the main contributors."

Friendly and concise

"Your portfolio is lower today, mostly because several equity positions moved down. Want to see the biggest drivers?"

Detailed and educational

"Your portfolio is down 1.8% today. The main driver is equity exposure, especially holdings in the technology sleeve. This does not necessarily change your long-term plan, but it is useful to look at the contributors before drawing conclusions."

Compliance-forward

"I can explain the portfolio movement and the main contributors. I cannot tell you whether to buy or sell unless that is part of an approved advice or mandate flow."

That level of detail matters because different answers carry different risk. A welcome message can be light. A portfolio-loss explanation should be calm and factual. An order step should be exact. A blocked advice request should be kind, firm, and useful.

White-label should mean the full surface is yours

With Charlie, you can go two ways:

  • Use the full SDK if you want deep control over the full experience,

  • or use our widget library if you prefer to build more through APIs.

In both cases, the UI is highly editable in detail, so it can match the bank’s design system far beyond a simple logo swap.

Our library includes ready-made widgets for Flutter, HTML/CSS and React, with Swift and native Android coming soon. We keep that library continuously updated, so teams can move faster without losing control over the look and feel.

Tone should change by moment

A strong tone system does not force every answer into the same shape. It gives each situation the right default.

Situation

Better tone bias

Why

First welcome

Warm, short, clear

The user should understand what they can ask

Empty portfolio

Helpful and educational

The user may need confidence before taking the first step

Market volatility

Calm, evidence-based, not dramatic

Emotionally loaded language can push users toward bad decisions

Portfolio loss

Empathetic but factual

The answer should acknowledge concern without sounding alarmist

Order flow

Precise and low-ambiguity

The user must understand the action, instrument, amount, and next step

Out-of-scope request

Polite, firm, useful redirect

A refusal should still help the user continue productively

Missing data

Transparent and specific

The assistant should say what is unavailable and what can still be answered

Human handoff

Clear and reassuring

The user should know why a human is being involved

This is where many generic assistants feel wrong inside a bank. They can be fluent, but they do not naturally know when to become shorter, more formal, more careful, or more explicit. Charlie's per-bank configuration layer is there to make those decisions part of the product setup.

FAQ

Can Charlie match your brand and tone of voice?
Is the SDK fully white-label?
Can messages be shorter or longer?
What should be editable beyond tone?
How does tone connect to compliance?
Can configuration change after launch?

Meet Charlie. The Investing AI Agent.

Give your clients an entire investing experience, rebuilt from the ground up as a conversational AI agent.

Unlock more insights in
digital investing and reporting
MiFID II Risk Profiles: Is it like hiking Alpine mountain trails?

Debiasing in real time: Experimental research on mitigating the disposition effect among retail investors.

There is no lack of books on the many biases that undermine effective financial decision-making. But how do you actually counter the biased behaviour of an investor -in real time- on a 9:16 inch mobile screen? We did experimental research with University of Leuven’s Mintlab to find out if we could decrease the so-called ‘disposition effect’ using nudging techniques.

Read more

Microlearning moments benefit both investors and investment platforms

Microlearning is the snappy cousin of e-learning and a common feature in many apps. But will a microlearning layer, embedded in an investment app, increase people's interaction with educational content? Will it improve knowledge retention? And will users favor this learning experience? To find out, we partnered up with the ‘Meaningful Interactions Lab’ to conduct experimental research. Here are the results:

Read more

Digital Investor Series - The unstoppable rise of the digital investor 4

This update is part of a series, building upon the perspective at the end of 2020: ‘The Unstoppable Rise of the Digital Investor’. In this topic we highlight: Beijing crackdown triggers Chinese tech stocks sell-off in July / The accelerated growth of direct indexing / Robinhood: from a disappointing IPO to a steep surge / The unstoppable rise of cryptocurrency trading amongst young people / Banks launch neobanks: super apps and investing / New investors are getting younger and younger

Read more
Unlock more insights in digital investing and reporting
MiFID II Risk Profiles: Is it like hiking Alpine mountain trails?

Debiasing in real time: Experimental research on mitigating the disposition effect among retail investors.

There is no lack of books on the many biases that undermine effective financial decision-making. But how do you actually counter the biased behaviour of an investor -in real time- on a 9:16 inch mobile screen? We did experimental research with University of Leuven’s Mintlab to find out if we could decrease the so-called ‘disposition effect’ using nudging techniques.

Read more

Microlearning moments benefit both investors and investment platforms

Microlearning is the snappy cousin of e-learning and a common feature in many apps. But will a microlearning layer, embedded in an investment app, increase people's interaction with educational content? Will it improve knowledge retention? And will users favor this learning experience? To find out, we partnered up with the ‘Meaningful Interactions Lab’ to conduct experimental research. Here are the results:

Read more

Digital Investor Series - The unstoppable rise of the digital investor 4

This update is part of a series, building upon the perspective at the end of 2020: ‘The Unstoppable Rise of the Digital Investor’. In this topic we highlight: Beijing crackdown triggers Chinese tech stocks sell-off in July / The accelerated growth of direct indexing / Robinhood: from a disappointing IPO to a steep surge / The unstoppable rise of cryptocurrency trading amongst young people / Banks launch neobanks: super apps and investing / New investors are getting younger and younger

Read more