ENTERING THE COPILOT ERA

By Syril Smith, Head of Product for AI & Andrae Allen, Lead Engineer for the Aladdin Copilot Ecosystem and AI Platforms

While AI itself is not new, the acceleration of generative AI has put us on the cusp of seismic change—analogous to the invention of the internet and the launch of the iPhone.

LISTEN NOW:
THE WAY FORWARD

Tarek Chouman, Kunal Khara,
Lance Braunstein, Syril Smith Garson

Since 1950, passing the Turing Test has been the “moonshot” of computer science. Alan Turing famously devised a test for machine behavior being indistinguishable from a human. We are now at that tipping point 74 years later. Executives and technologists have a spectrum of answers about where the emerging technology is headed—and what it means for our lives and our businesses. But the reality is that the promise (and peril) of AI is just beginning to be understood. And the technology’s future boils down to continuous, rapid adaptation; responsible governance; risk management; and ensuring humans are always a non-discretionary part of the overall calculus.While AI itself is not new, the acceleration of generative AI (GenAI) has put us on the cusp of seismic change—analogous to the invention of the internet and the launch of the iPhone. In part, its role (already) is to augment our lives, the way a search engine or a navigation app augments our knowledge. We believe it’s going to increasingly become woven into our day-to-day processes to help better inform our own decisions. It will allow for more cross-domain skill acquisition, enabling everyone to be a developer conceptually.

And the popularity of generative AI has boomed, in part, through the accessibility offered by copilots. Copilots are conversational interfaces that use large models to answer questions, pull information, inform actions, and help users. In some ways “copilot” is akin to saying “website” when the internet was founded, in the sense that copilot will likely become a ubiquitous industry standard and a table-stakes user experience paradigm. Think of this as generally going from a mostly one-sided computer interaction previously—where a machine executes commands but doesn’t respond back to prompts—to now a bi-directional dialogue in natural language.

ALADDIN INNOVATIONS
From automation to machine learning and natural language processing, artificial intelligence has been a part of BlackRock’s DNA for years—and continues to be applied across Aladdin® and eFront® technology. Robust governance and controls, client-centric product development, and the user experience approach are at the core.

With the launch of the eFront Copilot, we’ve integrated GenAI capabilities into our private markets technology platform—helping clients unlock new efficiencies and uncover actionable information instantaneously while ensuring they work in an environment where their data is safe. Similar efforts for the Aladdin platform are currently underway.
Our strategic partnership with Microsoft has enabled us to leverage Azure OpenAI Service to further develop AI-powered capabilities, including Azure Machine Learning and GitHub Copilot for developers.
On the Aladdin Wealth™ platform, we’re exploring ways to use GenAI to arm advisors with a “first draft” of key insights about their customers’ portfolios, guiding them to the most relevant analytics to help them explain their data in a way that end investors can understand.

As copilots proliferate across different domain specialties and use cases, we have the opportunity to stitch together workflows through a “copilot aggregator” concept. Like smartphone apps, however instead with copilot apps. When your third-party vendors deploy their own copilots, the collection of aggregated copilots creates a whole new copilot ecosystem and personalized user experience. In the future, instead of needing to know how to navigate to which app to perform what task, imagine a world where you can “just ask the question” and the AI automatically figures out which copilot app is best suited for that task. But all of this raises questions about controls and security, governance, and regulation. On that, there are four foundational principles.The first is to ensure that expert humans are always in the loop with the appropriate level of oversight for any workflow. The second is to take a walled garden approach, where the entire AI stack is deployed in your own secure environment and data is not shared outside of the environment. The third is to supplement the copilot models via a retrieval/augment/generate approach to feed them live knowledge from within the walled garden. The fourth and overarching principle is to never lose sight of the fact that outcomes depend on input and only information from within the garden walls will be processed.Overall, the more we interact with machines via copilots and other advanced interfaces, we—as operators—will inherently change our client experience expectations, like when we first used touchscreens on smartphones and tablets. This will fundamentally change how we do our work, our behaviors and habits, and the way we interact with other people. We are now officially entering the “Copilot Era.”

As copilots proliferate across different domain specialties and use cases, we have the opportunity to stitch together workflows through a “copilot aggregator” concept.

While nearly all firms agree AI will likely become commonplace among institutional investors in 3-5 years, just 8% have started integrating it into their applications.

^Source:^{Institutional Investor’s Tech Futures Survey 2024}