Adam Field, Chief AI Officer at Tungsten Automation – Interview Series

Adam Field, Chief AI Officer at Tungsten Automation, is a longtime enterprise technology leader with deep expertise in artificial intelligence, intelligent automation, and product strategy. In his current role, he leads the company’s global AI transformation efforts, overseeing the integration of AI across Tungsten’s product portfolio, guiding the Tungsten AI Lab, and establishing governance frameworks for responsible AI adoption. Prior to becoming Chief AI Officer, he served as Chief Product Officer, managing a portfolio generating more than $550 million in annual revenue. Before joining Tungsten, Field spent nearly 17 years at Pegasystems, where he led innovation and client experience initiatives, helped shape emerging technology strategy, and became known for delivering large-scale product showcases and enterprise innovation programs. Earlier in his career, he held technology and consulting roles at Staples, Publicis Sapient, and Fidelity Investments.

Tungsten Automation, formerly known as Kofax, is an enterprise software company focused on AI-powered workflow automation, intelligent document processing, robotic process automation (RPA), and business process orchestration. The company provides automation tools used by organizations across industries such as finance, healthcare, insurance, and government to streamline document-heavy operations and improve efficiency. Its platform combines AI, low-code automation, and document intelligence technologies to help enterprises automate repetitive tasks, extract insights from unstructured data, and modernize business workflows at scale.

You spent years leading product strategy and innovation, including building out innovation labs and scaling a $500M+ product portfolio, before stepping into the Chief AI Officer role at Tungsten Automation. What convinced you that now was the moment to shift fully into AI leadership, and how did your earlier experience shape that decision?

I’ve spent a large part of my career focused on turning new technologies into something that actually works at scale inside enterprise environments. Over the past few years, it became clear that AI is not just another capability to integrate into products. It is reshaping how software is built and how decisions are made across the business seemingly eclipsing and disrupting everything that came before. This shift from experimentation to real expectations for outcomes, coupled with the increasingly undeniable fact that AI is here to stay, made it the right time to step fully into an AI leadership role.

It also became obvious that AI wasn’t turning out to be the panacea that many marketed it as. AI success requires people who combine technical expertise and industry know-how. Tungsten wants to help companies do AI the right way and experience real returns, which is why Tungsten created the AI Office and my role within it.

Tungsten has evolved from early document capture and OCR into a full intelligent automation platform powering mission-critical workflows for thousands of organizations. How do you see that legacy shaping your approach to agentic AI today?

Tungsten’s history is deeply tied to how enterprises actually operate. We have spent decades working with documents and workflows that sit at the center of critical business processes. That means we understand how complex and often unstructured that information can be.

That foundation is very relevant for agentic AI. These systems need to operate within real environments, not just interpret information in isolation. Our background in document intelligence allows us to focus on context and on making sure AI acts in a way that is consistent with how the business runs. It is about building systems that can be trusted in production, not just explored in theory.

That is why this latest AI evolution is so exciting. It takes Intelligent Document Processing to places we never could take it before — solving problems that were too expensive or impossible to tackle in the past.

You’ve emphasized embedding AI across the entire product portfolio rather than treating it as a standalone feature. What does “AI-native” transformation actually look like in a large, established software platform?

It became clear early on that generative- and agentic AI-powered features were quickly becoming table stakes, meaning that customers weren’t always willing to pay extra for them. We also realized that these technologies allowed us to modernize what Tungsten has been doing for years: helping companies make sense of their document data.

We didn’t change our brand promise. We didn’t create one-off products or bolt-on features. We refactored how the product is used, and when that foundation is in place, AI can operate in a way that feels natural within the product rather than separate from it. And the use cases our customers began addressing moved from structured documents to unstructured sources of information. And, we redefined “document” along the way. No longer is a document an image of paper or a digital file. Unstructured data lives in things like claims adjuster notes, contact center call transcripts, social media posts, web articles, and much more.

Taking this approach allows our customers to augment the foundation and open models with their proprietary data, which is the true differentiator.

As the company’s first Chief AI Officer, how are you balancing innovation speed with the need for governance, security, and responsible AI deployment at scale?

There is always a push to move quickly with AI, but in enterprise environments trust matters just as much as speed. Governance and security cannot be treated as an afterthought. They need to be built into the system from the start.

The way we do this is by setting expectations up front by educating our end users. For example, half of my role is focused on internal AI strategy, evangelism, and governance. We brought together a cross-functional advisory council very early on. We encourage sharing, experimentation, and communication. There were times when the technology was ready to roll out to all employees connected to several internal systems. The prototypes were powerful and got everyone excited, but we let our advisory team know whenever we run into potential security or regulatory hurdles. They appreciate the insight and often participate in the solution.

I think it’s also important to not allow perfection to get in the way of progress. We set the expectation with our staff that they should expect change, and lots of it. They should expect that we’ll roll tools and features out as they are ready, get feedback, change course if necessary, and then roll out more.

Agentic AI is quickly becoming a major focus across the industry. In your view, what separates real enterprise-grade agentic systems from experimental or overhyped implementations?

The key difference is how systems perform in real conditions. Many experimental approaches work well in controlled environments but struggle when they encounter messy data or complex workflows. Enterprise-grade systems need to handle that variability and still deliver consistent results.

Most systems over the past 30 years were built for human interaction or via very controlled API access. Systems integration needs to be rethought in the agentic era. Everything from how to handle exceptions, errors, and auditing is different when agents are interacting rather than a human through a traditional UI.

Another important factor is accountability. Organizations need to understand how decisions are made and be able to trust the outcomes. That level of transparency is what allows agentic systems to move from interesting demonstrations into real operational use.

You’re leading the Tungsten AI Lab as a hub for research and applied innovation. How do you ensure that experimental AI work translates into measurable business outcomes for customers?

I actually took somewhat of the opposite approach with the Tungsten AI Lab. I let the team know early on that it was okay to experiment, to learn, and to try new approaches even if the outcomes never made it into our products. Often it’s better to learn what not to do. I believe that this has given them the freedom to think freely and experiment new ways of doing things.

As an example, while I can’t disclose the exact feature, one of our current research sprints involves a brand new approach to an existing product component. The researchers found new methods to solve a problem which led to a “lightbulb” moment that we may be able to offer a complete new add-on solution to our customers. If we just researched how to implement what was already in the roadmap, we would have never gotten here.

That said, it’s not a free-for-all. We are thoughtful about where we spend time and how much of it we spend on each research project.

Many organizations are still struggling to move from AI pilots to production. What are the biggest barriers you’re seeing, and how can companies overcome them?

One of the biggest barriers is dark data. Most organizations have access to enormous volumes of information, but a large portion of it lives in documents, emails, PDFs, and other unstructured formats that are difficult for AI systems to interpret. That means even well-designed models are often working with an incomplete and inconsistent view of the business, which leads to unreliable outputs and stalled initiatives.

To move beyond that, companies need to focus on turning dark data into something usable. That involves not just extracting information, but creating structure, context, and governance around it so AI systems can actually act on it with confidence. Once that foundation is in place, AI becomes far more reliable and much easier to scale from isolated pilots into real production environments.

Tungsten works across document-heavy and workflow-intensive industries. How is AI changing the way enterprises think about unstructured data and decision-making?

AI is changing how organizations think about the value of the information they already have. For years, large amounts of enterprise knowledge sat inside documents, emails, PDFs, and other unstructured content that was difficult to access or operationalize. Now organizations are realizing that this data contains the context and business logic AI systems need in order to produce reliable outcomes. The models themselves are commodity, organizations’ proprietary information combined with these models is the differentiator.

At the same time, there is growing awareness around data sovereignty, governance, and where enterprise information is flowing. A lot of companies are racing to pull in more external data or experiment with broad model access, when in reality they are already sitting on massive amounts of untapped intelligence within their own organization. The focus is starting to shift toward activating that internal unstructured data in a secure and governed way so AI can support better decisions without creating unnecessary risk.

You’ve built Customer Advisory Boards and worked closely with enterprise clients throughout your career. How important is customer feedback in shaping AI strategy, especially when the technology is evolving so quickly?

Customer feedback is a gift, especially in a space that is moving as quickly as AI. It helps ensure that the strategy stays grounded in real business needs rather than theoretical possibilities.

It also helps with prioritization. There are many directions AI can go, but customer input provides clarity on where the greatest value can be created. That keeps the focus on outcomes that matter and ensures that innovation stays aligned with how organizations actually operate.

I recall in the very early days of gen AI, a customer on our advisory board telling me that while she loved the product direction, she’d never pay extra for a new LLM-powered feature in our roadmap. That was eye-opening because she was aligned with the rest of the industry.

Looking ahead, where do you see the biggest opportunity for AI-driven automation over the next 3 to 5 years, and what should enterprises be preparing for now?

The biggest opportunity is in connecting AI more deeply into end-to-end workflows. Rather than focusing on isolated tasks, organizations will look at how AI can support entire processes and improve how work moves across the business. Right now, lots of agentic systems are targeted at discrete tasks but businesses operate on compliant end-to-end processes.

To prepare for that shift, enterprises need to invest in their data foundations and in systems that support transparency and control. And they should be thinking about “build vs. partner” rather than “build vs. buy.” We’ve seen AI DIY from scratch fail too often. The organizations that benefit most will be the ones that find the right AI-powered partners to accelerate their solutions rather than trying to rebuild everything from scratch.

Thank you for the great interview, readers who wish to learn more should visit Tungsten Automation.