Top online brokerages are competing harder than ever for the self-directed investor’s wallet share. But AI only helps win it when it’s built around the investor, with real-time data, client context and evaluations that measure what really matters to their users.
Investors have access to the most capable AI models ever built – and they aren’t waiting for platforms to catch up. In Canada alone, a 2026 survey of 5000+ investors by the Securities and Investment Management Association and Pollara Strategic Insights found that 4 in 10 have already used generative AI for investment purposes.
The challenge is that those models were never designed for investing. They don’t see live market data, they don’t know an investor’s holdings or goals and they’re scored on benchmarks that say little about investing outcomes.
That said, if the answers they get from general-purpose AI tools feel more useful than the ones your platform and your teams provide, their engagement and eventually their assets will follow, even though those answers come with no guarantee they’re accurate, consistent or traceable to a verified source. That’s a risk no regulated firm can afford to inherit.
At the Communify Intelligence Experience in London, Emil Noordeh, Head of AI, Digital Investing at BMO, gave one of the sharpest takes on how to make AI genuinely useful and trustworthy for self-directed investors. Here’s what he sees as missing – and what it means for your firm.
Why Generic AI Falls Short for Self-Directed Investors
Emil comes to AI from astrophysics, a field built on extracting signal from noise across millions of interconnected variables, incomplete observations and constant uncertainty. Investing, he argues, looks much the same.
He’s clear that today’s foundation models are genuinely powerful and getting more capable with every release. But they’re generic by design. They aren’t trained on real-time financial data, they don’t know an investor’s holdings or goals and they’re evaluated on benchmarks that say little about investing outcomes. His answer isn’t to replace them. It’s to build what’s missing around them.
“What you need is a harness around these large language models.”
What Goes Into the Harness: 3 Essentials
Emil names three things a foundation model needs before it can truly serve self-directed investors.
- Real-time financial data: High-quality, real-time market data isn’t part of how these models are trained, so it must be built in. With it, the possibilities widen. Emil describes AI that spots a policy change in one market, traces how it ripples across supply chains, sectors and geographies, then identifies which companies’ forecasts and risk profiles will shift – in seconds, always on. That only works on a foundation like the Communify Knowledge Base, where market data is unified, verified and ready to use.
- Client context: A general-purpose model doesn’t know an investor’s holdings, risk profile, goals or time horizon, so its insights can’t be personal. Give it client context alongside market data and its guidance becomes relevant to the individual asking. In the Knowledge Base, both sit side by side, so every answer can be correct and personal at once.
- A measuring stick built for investors: This is the point Emil says is often glossed over. Even the benchmarks foundation models are judged on aren’t particularly relevant to investors. What’s needed are evaluations fit for exactly what a self-directed investor is trying to do – in his words, “a measuring stick, a ruler” that confirms the system is driving improved investing outcomes. The best role your experts can play right now, he argues, is setting that gold standard with “golden” questions and answers the system is tuned against. Communify’s Proof of Dimension (P.O.D.) framework applies that same discipline at scale, holding every output to a 99% consensus threshold before it reaches an investor. Deterministic Cache then makes sure an answer that meets that standard is the same answer every investor receives.
From Pilot to Proof
Emil’s framework points to a bigger shift for firms. Choosing a foundation model is no longer the hard decision – every firm has access to the same ones. What sets a platform apart is everything around the model: the data it draws on, the context it understands and the standard it’s held to.
That changes what it takes to move AI from experimentation to strategic use. A model that performs well in a pilot isn’t the same as AI you can stand behind across every investor interaction. Firms need to show that every answer is accurate, consistent and traceable to a verified source. That kind of proof can’t be added after the fact. It has to be designed into the platform from the start.
How Communify Delivers Investor-Ready AI
The data, context, standards and proof described above are already built into the Communify platform.
- The Knowledge Base: Every Communify capability runs on it. The Knowledge Base unifies 4500+ data feeds from 230+ third-party providers – covering market, fundamental, research, news and regulatory data – with your firm’s own client data in a single, verified source of truth. This guarantees that every Communify experience draws on information that’s both trusted and personal.
- MIND™ AI: Communify’s proprietary intelligence engine combines deterministic AI engines with leading LLMs, grounded in a verified Knowledge Base rather than the open internet.
- Proof of Dimension (P.O.D.) framework: P.O.D. validates every MIND™ AI output before it’s delivered, with an effectively zero hallucination rate across 50000+ test runs, so every answer is accurate and auditable. Read more about P.O.D. in this whitepaper.
- Deterministic Cache: Once an answer passes P.O.D., it’s stored and served the same way every time the question is asked – consistent for every investor, auditable for compliance and delivered at speed, while reducing token costs for your firm.
These 4 foundational parts of the Communify platform underpin a library of 200+ AI-powered apps firms use to deliver trusted, investor-ready experiences – and firms can build on them fast. Finextra, the independent fintech newswire, built its Finextra Pro platform on Communify’s Component Tech and Knowledge Base, moving from idea to Alpha in under six weeks.
For your firm, that means differentiated, AI-enabled experiences built in weeks, not years.
Your AI Strategy Is Only as Good as Your Measuring Stick
Emil’s advice is simple: give the model real-time data and client context, and hold it to evaluations your experts define for your investors rather than a generic benchmark. The firms that own that standard, and can prove their AI meets it, are the ones investors will trust with their questions and their assets.
The good news is – you don’t need to start from scratch. Communify brings all of this together in one platform, trusted by 400+ financial institutions. To see how it can help your firm deploy AI that investors can rely on, book a demo with Communify.