At AlphaSummit 2025, it was no surprise that AI dominated the conversations. Many talks centered on how AI is already helping investors and others save time, surface hard-to-find insights, and make faster, more confident decisions.

But much of the discourse also focused on the future. What does the next phase of AI look like when it’s multimodal, memory-driven, and increasingly personal? And how will these capabilities change the way we search, analyze, and act on information?

Here are some of the key themes that emerged about the future outlook for AI.

Many Models, Many Winners

Keith Weiss, Head of U.S. Software Research at Morgan Stanley, sat down with Patrick O'Shaughnessy, CEO and Founder of Positive Sum. One question they explored: Would there be one single-purpose model that dominates? Weiss suggested that gains would be more dispersed.

“What we’ve seen historically is typically not one giant general purpose solution that solves all of the equation,” said Weiss. “There’s a reason why Amazon is not the totality of retail. There’s a reason why we still have Wayfair.”

And when it comes to AI, Weiss noted that Claude is a great coding tool, but when Weiss wants to write something well, he goes to Gemini. When he wants to understand “what are people thinking right now,” he goes to Grok.

The risk, according to Weiss, is that too much of the capital goes to too few vendors where it should be more dispersed.

This echoes the sentiment of an AI expert from a Tegus Expert transcript in September:

“I don't really think we will get to the point where you'll have one model that does everything really well… Instead, you will just have these models get smarter and smarter and you'll end up having models that act like the smartest engineer in your company, but it's for one very specific task and does that incredibly well, and then you build on top of that.”

Seeing and Hearing with AI

Ehsan Ehsani, Executive Director at Crescendo Partners, predicted that over the next three to five years, genAI will be processing voice and video as fluently as it now handles text.

A CEO’s “No” versus “Mmm…. No” in a conference call may not show up differently in a transcript, Ehsani explained. However, that difference could be picked up from voice. In addition, body language and facial expressions could be captured as well.

Taking it further, Ehsani said that person-specific insights would also be possible. For example, detecting what’s different about a CEO’s tone or demeanor compared to prior videos of the same CEO.

This capability could reshape everything from earnings call analysis to primary research interviews, giving analysts a richer read on sentiment and subtext that numbers and text alone can’t capture.

We are already seeing early signs of this. Google Gemini, for example, can analyze videos and audio together, detecting subtle details like raised eyebrows, shifts in expressions, and variations in pitch, and can then explain what those changes might convey. It’s a glimpse of how multimodal AI could interpret not just what is being said, but how it’s said — turning tone, gesture, and inflection into new layers of insight.

From Search to Real-Time Insights

Ehsani also predicted a leap in how we interact with information. Instead of waiting for videos to end, we will be able to query them in real time.

“What will happen is everything will become more real time,” Ehsani said. “As the CEO is speaking, you can make a query and say ‘What has changed?’ and get an instant reply so you don’t have to wait for it to be over.”

He also said we’re moving from platforms behaving like YouTube, where users need to search for something, to being more like TikTok, where we won’t have to type anything to get what we want.

“They know us, they know our preferences, and they provide suggestions to us,” he said.

AI systems are already retaining information about users across sessions, learning their preferences, interests, and patterns of interaction over time. That makes this next step feel inevitable: tools that don’t just respond, but anticipate. We’re already seeing early signs with OpenAI’s ChatGPT Pulse, which can proactively research on a user’s behalf based on past conversations and doesn’t wait for a user prompt.

The Rise of Personalized Agents

Along these lines, AI’s growing ability to remember and anticipate user needs is paving the way for more personalized, proactive agents.

Matt Reustle, host of the Business Breakdowns podcast, said we should expect personalized agents that evolve alongside their users.

“Agentic AI is going to be more proactive, and the memory feature that’s built into this technology layer is the biggest thing here,” he said. “With agents, there is the ability to train these to not only learn you, understand you, act in unique different ways, but those are going to get better and better over time, based on your interactions with agents, but also their ability to learn and continuously stay up to date autonomously learning.”

In the future, Reustle added, agents won’t just interact with people, they’ll interact with each other.

Still, there are challenges. One that often gets mentioned: “Agents don’t have all the nuanced thinking of our team.” Capturing that nuance — your personal or organizational way of working — is difficult. He recommended using notetaking tools to document as much as possible. Today, that’s a time saver; no need to take notes. Eventually, this becomes data for your LLM or AI agent.

Ehsani shared a similar view, noting that many funds have a lot of investments that they have already made and have metrics they consider more than others. They can train the system so agents will give them outputs and reports exactly the way they like.

The Road Ahead

Across these conversations, a picture of AI’s future comes into focus. It’s a landscape defined by specialized models, multimodal understanding, interactive systems, and personal experiences.

As AI evolves, one thing is clear: the human element will remain central. The organizations that succeed will combine human judgment and machine intelligence in ways that build confidence and trust, ensuring that insights are delivered with the clarity and context needed to drive better decisions.

As Jack Kokko, CEO and Co-Founder at AlphaSense, said in his keynote address, “The right insights surfaced when they are most needed can change the path a company is on.”

This idea captures what the AlphaSummit discussions revealed: a future where insight is timely, contextual, and deeply personal.

Missed the event? Catch all the 2025 AlphaSummit on-demand sessions and register to join us for next year’s conference.