The End of the Dashboard? Why Your Next Data Interface Will Be a Conversation.

What if you could just talk to your Factoring system?

Receivables finance has never lacked data. But having the data isn’t the same as understanding it. Documents, processes, and risk indicators can live in different systems, managed by different teams. With regulations tightening and the demand for audibility growing, it’s becoming more important to make sense of your data.

The limit of clicks and data.

One of the most visible innovations in computing was the introduction of graphical user interfaces. In factoring, what really drove the adoption of web-based GUIs was client-facing demand. Online portals had to look modern, work on mobile devices and match what users expected from their banking apps. Factoring companies adapted and client access was updated, also on mobile devices.

Alongside GUIs, widespread adoption of APIs (Application Programming Interface) brought new flexibility. They made it easier to plug into external tools and build automations around core systems without replacing them. But while APIs helped move data between systems, they didn’t help much with interpretation. GUIs could display information, APIs could retrieve it, but neither could provide the holistic overview of a real life client case.

It’s a bit like how music interfaces evolved. Years ago, I had a chaotic MP3 collection. Files scattered across folders, many of them mislabeled. Winamp helped me play those tracks, but it didn’t fix the disorder. iTunes improved things if I did the tagging myself. Then Spotify came along, and the structure didn’t matter anymore. It learned what I liked, picked up on patterns and surfaced the music I wanted without needing me to organize anything. It didn’t fix the files. It simply changed how I accessed them.

And this is where something new might be emerging. I don’t see many people in our industry talking about this yet, but it’s been on my mind:

What if large language models (LLMs) could serve as a new kind of interface?

A way to explore what systems contain through natural language. Instead of digging through dashboards or downloading reports, you just ask a question and get a direct answer, written in plain terms.

In receivables finance we already have all the data we need: invoices, payments, onboarding files, compliance logs and client emails. It exists, but it’s scattered across systems. If you want to understand the story behind a client you often still need to piece it together somewhat manually.

That’s where Large language models might come in as exploratory tools for interpretation.

You might ask, for example:

“Which onboarding files contain missing or outdated documents and are there patterns in how different teams or regions handle these gaps, based on internal comments or correspondence?”

Today, these questions might require someone to export data from multiple systems, compare manually, and synthesize the answer. With the right setup, an LLM could do the heavy lifting, surfacing connections, summarizing patterns to help teams act faster.

Now, I want to be clear: These tools aren’t in widespread use across the industry, and some areas may never be a natural fit. For example, take the EU’s ViDA (VAT in the Digital Age) initiative. It requires structured e-invoicing and real-time reporting. That’s about strict formats and deterministic rules. A probabilistic system like an LLM isn’t a great fit for filing tax reports. Here, we need systems that validate accurately.

Putting conversational AI to work.

Think of preparing for an audit, training staff, or reviewing internal exceptions. In these situations, the ability to synthesize and explain matters more than ticking a box. You don’t need the system to decide alone, you are using it to help someone understand.

Other regulations create clearer opportunities. The AMLA package, introduced in 2024, strengthens EU anti-money laundering rules. Compliance teams now need to demonstrate how risks were evaluated, how red flags were addressed, and how decisions were justified. That requires narrative thinking. Similarly, CRR III is changing how credit risk is calculated. Factoring providers will be expected to explain why they’ve granted certain advance rates, how they handle exceptions, and what methodology underpins their credit logic.

In both cases, LLMs could act as support tools. They won’t make final decisions, but they could help teams prepare reports, compare historical precedents and flag inconsistencies. Used well, they could free up time by surfacing what matters and allow professionals to focus on judgment.

AI as a tool for judgement in everyday work situations.

A relationship manager wants to understand which clients have increased invoice volumes and whether those increases correlate with any emerging risks. Some of that information is structured in systems that track volume or industry codes. But some of it isn’t. Risk-related notes might sit in internal emails, onboarding checklists, or scanned PDFs from a KYC (Know Your Customer) process.

A traditional dashboard can pull the hard numbers. But it won’t summarize patterns in client communication or spot inconsistencies in documentation across jurisdictions. With the right setup, an LLM could bridge that gap, bringing together structured data and scattered context to offer a first layer of interpretation. Something that helps the relationship manager ask better follow-up questions, faster.

The compliance officer could follow up with, “Draft reminder emails for the flagged clients.” The LLM could prepare a first draft and save the team hours.

Meanwhile, a risk analyst might ask, “Show me clients with unusual shifts in payment behavior, like frequent early repayments followed by sudden delays or defaults.” Instead of needing a custom SQL report or BI dashboard, the LLM could generate a list and highlight related notes from client communication.

Of course, this won’t work out of the box. It requires structured data, thoughtful integration and domain-specific prompts. It also raises real concerns about data security and model accuracy. None of this is trivial. But the direction is worth exploring.

Factoring is a business built on judgment. We rely on client behavior, historical nuance and expert intuition. That’s not going away. But what if we could spend less time hunting for information and more time making the right calls?

Maybe we don’t need new platforms. Maybe we just need better ways to see what’s already there.

This article was published in 2025 Almanac of the Polish Factoring Association. This text has been edited and adapted for the LinkedIn platform.

By Daniel Huszár