Is Chat Really What Users Want?

Last week, on a client call, something small but telling happened. One of our engineers was listening to feedback while we walked through a use case. In the background, he started building. By the end of the call, we had a working interface to show. Not a mock, not something we'd follow up on later, but something real enough for the client to react to immediately. They could say "yes, that's what I meant," or "this needs to move," or "can this connect to X?" That loop used to take days or weeks. It happened during the conversation.

It got me thinking about how we talk about AI.

Most of the conversation defaults to chat. Ask a question, get an answer. And to be fair, it works. It's a great way to explore a model. But that's not what happened on that call. This wasn't about answering a question. It was about creating the thing the user actually needed.

Natural language is clearly powerful, but maybe not as the destination. More like an instruction layer. Not "tell me something," but "build this for me." If that's right, the implication is significant. The interface stops being fixed. Instead of navigating software, you start shaping it. On demand, based on what you're trying to do in that moment.

You can already see the need for this in how people actually work, especially in finance. Analysts are constantly stitching together data, moving between systems, reformatting outputs, and trying to maintain context across everything. Multiple screens, multiple tools, constant switching. Most users in this world are effectively power users, whether the software was designed that way or not.

And yet most firms are treating chat as their AI strategy. A lot of the new "AI products" in this space are essentially ChatGPT clones with a few finance-specific integrations layered on top. Slightly better data access, maybe some domain tuning, but fundamentally the same interface. So it's worth asking: is that what users actually want, or is it just what they're familiar with and comfortable adopting right now?

Chat is easy to understand. It's low-friction to roll out. It feels like progress. But it doesn't solve the underlying problem. The problem isn't getting answers. It's shaping workflows around messy, fragmented data and making that usable across a team.

The reason more firms aren't taking bigger swings beyond chat comes down to a few things. It's the safest place to start. Anything deeper touches data, systems, and processes. And the next step still feels a bit undefined. So teams wait.

But moments like that call last week suggest something is shifting. We're getting closer to a world where users have direct access to their data, AI understands the context of their work, and interfaces can be created or adjusted in real time to match the task. Not perfectly, and not everywhere yet. But enough to start changing expectations.

If that plays out, the question shifts from "which tool should we buy" to "what should our system look like for how we actually work?" Chat was a good starting point. But it's probably not where this ends.

~Ryan Erickson, Founding Executive

About Shift
Investment research shouldn't be this hard. Shift turns your firm's scattered knowledge into powerful insights with AI built for how you actually work. We're a team of builders and finance experts based in Charlottesville, VA.