The Elephant in the Room Finally Speaks

last updated:

The Elephant in the Room Finally Speaks

last updated:

Kofax unveils its Generative AI plans

Way back at the dawn of the “real” AI age in December 2022 (tongue firmly in cheek), we predicted that the technology behind ChatGPT – generative AI – would disrupt the intelligent document processing (IDP) software market and usher in the 4th Wave of IDP. In the face of overwhelming hype, Deep Analysis advised our clients to be cautious given the controversies swirling around ChatGPT. We urged them to invert Zuckerberg’s credo of moving fast and breaking things. Instead, “move slowly and don’t break anything – yet”.

By the end of July, the vast majority of IDP and RPA vendors had announced a genAI “something”. But casting an unavoidable shadow over every announcement was the ultimate elephant in the room: Kofax, the dominant IDP market leader when measured by revenue. What would Kofax do, and when would they do it?

Yesterday the elephant spoke. Kofax presented a briefing on its generative AI product plans. When the elephant speaks, everyone else in the room has no option but to listen. I covered it for Deep Analysis and here follow my first impressions.

Kofax went all-in on the genAI theme. The webinar started with an introduction from Brianna, a suspiciously perfect-looking and shiny person. So I counted the fingers; she has ten. Having worked at Kofax* I thought, okay maybe she is just another perfect Californian from Orange County. Then Brianna confirmed my initial suspicion by identifying as an AI-generated presenter. An impressive one too. Nicely done, Kofax marketing team.

Thankfully, humans took over. Chris Huff, chief growth officer, presented a positive overview of the genAI marketscape and underlined the urgency for every enterprise to have a strategic plan for managing the technology before chaos reigns. To drive home the point, he compared the potential impact of genAI against what mobile did to Web 2.0 companies in the 2000s. Remember? Some embraced mobile and grew faster (Facebook); new companies emerged and grew into giants (Uber); but some were too slow and disappeared (MySpace). Evolve or die.

Next up was Adam Field, SVP of Products, broadcasting live without blurring the background from what seemed to be his man cave. I noted the obligatory guitars, standard kit for today’s cool tech exec. Nothing against guitarists – all my bands had at least one, and I love the Traveling Wilburys – and Adam seems a cool cat. But have you ever seen a keyboard, trumpet, or cello in anyone’s office? Consider for a moment what that reveals about tech execs; then we’ll move on.

Adam is a developer at heart who compared his first experience with Github’s GPT copilot for coders to the life-altering amazement of a young boy in the 1950s seeing a television for the first time. My first experience was considerably less earth-shattering than Adam’s but I think his exuberance makes him the perfect ambassador for Kofax’s genAI roadmap. After all, there are plenty of skeptics who will need convincing, beginning with the enterprise developers who are the backbone users of Kofax Total Agility (KTA) and have the most to gain (or to lose).

To date, most of the focus on AI’s business value has been about creating efficiencies by reducing or eliminating human computing tasks such as data entry or reading documents. Adam elevated the discussion by saying genAI is not just about driving efficiency. He called out RPA, which he said promised to accelerate efficiency and unburden humans but fell short. He cited an Ernst and Young study that 50% of RPA projects have failed. 

The Kofax view is that genAI is really about proficiency : elevating the quality of human work, enhancing our creativity, and enabling deeper insights. The Deep Analysis team couldn’t agree more; we’ve preached that from the beginning.

Adam interviewed Simjees Abraham, Head of Global Automation & Digitalization at DB Schenker, a Deutsche Bahn subsidiary. Simjees’ team uses KTA as part of the company’s automation initiatives; they also have experience working with large language models (LLMs). DB Schenker are passing KTA data extraction results into an LLM to power data analysis, summarization, and data insights. This has also enabled KTA to process more document types without the need for long training cycles or templates. Simjees was quick to point out that human oversight and transparency are two core principles for DB’s use of AI. Wise counsel to us all.

David Sentongo from the Kofax product team then presented the Kofax genAI roadmap. Kofax’s first move is the KTA Azure OpenAI connector, available now on the Kofax Marketplace. Using this, KTA customers can build workflows to add runtime functions such as data insights. In the following bank customer onboarding example, GPT has analyzed the documents in a loan application file and offered insights on the risk factors so the case worker can quickly focus in on the highest risk areas.

Next on the roadmap will be a co-pilot for KTA developers, enabling them to converse naturally with AI to create new workflows, document processing models and forms. As we’ve noted in our other 4th Wave coverage, replacing coding and arcane regular expression language with natural language prompts saves development time as well as dramatically expanding the ranks of “citizen” developers with no prior programming skills. Kofax showed this example of how a new workflow could be created:

What about data extraction and document classification, the two functions most associated with a Kofax capture solution? Other IDP companies have already released their data extraction and classification tools powered by genAI. These tools are reducing the long and arduous model training process associated with older IDP product like KTA down to a matter of minutes.

We’ve written extensively on the benefits of the 4th Wave, and Kofax acknowledged this by showing a comparison table pitting the current state of KTA data extraction tools against a future release using genAI. To summarize, genAI leaves the current state in the dust:
• genAI automatically generates a suggested list of data fields so you don’t have to.
• genAI can find information from any place on a document with a simple instruction. No more lassoing, regex, voting methods, fuzzy matching, or need to collect thousands of samples for training.
• genAI can handle the inevitable data and layout variability within a document type (think invoices) without the need for constant retuning to every variation.

Kofax announced that it will release a “zero-training extraction” function for KTA, which will certainly be welcomed by its large developer community. No release date was given.

Other genAI projects underway at Kofax include:
• an Insights co-pilot that will add summarization and ad-hoc queries for the data produced by KTA.
• An LLM Ops methodology to fine-tune a customer’s private LLM using data extracted by Kofax products.
• genAI functionality across the Kofax product portfolio including some interesting applications for PowerPDF.

The elephant has spoken.

Our Opinion

After months of market speculation about its plans, Kofax has finally joined the 4th Wave with a well-thought out and articulate strategy for its customer base. And none too soon: nimbler 4th Wave competitors such as Hyperscience, UiPath, Rossum, Instabase, etc. are coming hard after the Kofax installed base.

We think Kofax has done the right thing by moving cautiously and consulting its large customer base. Unlike many other IDP companies, Kofax has an extensive product line including RPA and has acquired several IDP products with accompanying technical debt. It takes a long time for an aircraft carrier to change course, and when it does you hope the crew has rigorously checked and double-checked every facet of the maneuver.

The most compelling genAI features announced yesterday may not come to market for 12 months, giving some daylight to competitors. But Kofax has loyal and long-standing enterprise customers who are loathe to rip and replace an installed production system with years of development and training invested.

We predict that before it can launch its own genAI features, Kofax will encounter some customer churn especially for new use cases inspired by genAI such as front office insurance work. Yet its legacy back-office business of capturing and processing claims, invoices, and other business documents should fare well, as those projects are the most difficult to displace.

If you are a Kofax customer looking for independent, third-party advice on when, how and where to adopt generative AI, Deep Analysis is happy to help. Sign up for a free “counseling” session with an expert in process and document automation.

*Dan worked for Kofax from 2009-2012.

Leave a Comment

We use cookies to ensure that we give you the best experience on our website. If you continue to use this site we will assume that you are happy with it.

Work Intelligence Market Analysis 2024-2029