If you’re following the evolution of AI, you already know how noisy things have gotten. Every platform promises magic. Every tool claims to be smart. And yet when you ask most of them a simple factual question about your own company, the answer often falls apart. While it may sound confident, as our Director of IT Fred D’Silva puts it, modern AI is also prone to being “confidently inaccurate.”
To cut through that noise, we sat down with Fred, who is also the architect behind Chroniqle and has spent years studying how AI handles real business documents. Our conversation focused on the gap between generic AI tools and the kind of intelligence companies actually need: not clever guesses or web-sourced assumptions but verified truth drawn from the documents that make up their histories.
The Real World Is Visual
Think about your own documents over the past decade. Almost none of them are purely text anymore. You have charts, photo spreads, infographics, callouts and layouts designed to guide a reader’s understanding.
If an AI system ignores the visual layer of these materials, it’s ignoring a huge part of your institutional memory. That becomes a real problem when you are trying to make decisions, tell stories or verify events that took place years ago.
“These AI systems are here to help find all that information that you have and give you access to it. But what good is it if it’s only giving you half of the information?” Fred says. “If you only have the text stuff but you don’t have anything about the images, then you’re missing out on a lot of other crucial information to help tell your story and tell it in an authentic manner.”
Chroniqle doesn’t have that problem. Using a combination of techniques—what Fred calls the “secret sauce”—it pulls data from charts, tables and graphics to give you the information and answers you need.
As Fred says, “We always want to make sure that we’re speaking to the history of your company and speaking to it accurately.”
Why Other AIs Fall Short
During testing, Fred and the History Factory team used our 2025 Heritage Gap Report, a visually rich document, to compare Chroniqle to tools like NotebookLM and ChatGPT. The results were clear.
NotebookLM got some answers right when it read the PDF directly but failed when we asked it about specific callouts that were only referenced within an image. Additionally, tests that simulated typical AI enterprise platforms that only capture raw text showed heavily inaccurate answers. One AI tool interpreted a chart that clearly showed a rating of 74 as displaying a 50 instead—simply because it could not understand the table’s visual layout.
ChatGPT struggled too. It was able to correctly interpret some of the references within the embedded image, but it too read the chart incorrectly and provided inaccurate answers. As Fred explains, generic AI models rely on what they have learned from the public web rather than being fully grounded in what is actually in your documents and thus often “quickly hallucinate the wrong information.”
When you’re dealing with brand history, regulatory documentation, executive communications or anything tied to your organization’s reputation and truth, there’s no such thing as “good enough.” The information you’re presenting needs to be correct.
How Chroniqle Approaches the Problem Differently
Chroniqle is built on a closed-loop model that draws exclusively from your organization’s professionally curated archives and research. Every answer is grounded in verified history.
But the real differentiator is what happens before information ever reaches the model.
Chroniqle’s workflow blends machine learning–based text extraction, image interpretation with LLMs and rigorous AI workflows to accurately capture and interpret figures, graphs, captions and more. But just to make sure, our team of experts steps in to confirm that visual information is captured accurately and reflects its true sources.
“Before it makes its way into that knowledge base, we have teams of experts that are essentially comparing the converted log that’s going to make its way into AI versus the original source of truth, which is the original PDF or document,” Fred says. “We’ve basically perfected it to a point where we now have an AI system that is going to always give you the right information, because we’ve put the right information in for these AI systems to be able to read it.”
The Big Picture
Chroniqle does something other AI tools are not built to do: It treats your company’s history as a strategic asset, not as a loose pile of files. It respects the way people actually read, create and remember information.
And because it stands on 45 years of archival expertise, it does all of this with accuracy at the center.
“If you’re looking for an AI that you can use without having to worry about if the information you’re getting is wrong, try using Chroniqle,” Fred says. “Just ask a question and the world is your oyster.”`
Join our waitlist for a CHRONIQLE™ demo.