It’s easy to talk about the benefits of AI in the abstract. It’s harder to show what it actually looks like inside a running enterprise workflow — who it helps, what problem it solves, and how it works alongside the humans who still need to be in the loop.
Jaam automation, a Nintex premier partner, has been integrating AI into K2 solutions for several years and across a wide range of industries. Andrew “Murph” Murphy, Co-founder and Chief Strategy Officer at jaam, recently walked through three of those implementations at a Nintex Connect session.
From invoice processing to agents: A brief history
Before diving into the use cases, it’s worth understanding how jaam’s use of AI with K2 has evolved. It didn’t start with agents or large language models. It started, as most things do, with documents.
“We began in 2023 with custom intelligent document processing,” Murphy said. “And we are actually still using that today.”
We think it’s impressive (and telling) that the IDP solutions jaam built three years ago haven’t been replaced, but joined by successive layers of capability. Custom AI services arrived in 2024. Now, agentic approaches are entering the picture. Each wave has added to the stack rather than disrupting it.
The consistent thread throughout has been Nintex K2 and its evolving capabilities. K2’s SmartObjects and integration-first architecture provide the connective tissue between whatever AI service is in play and the workflows, forms, approvals, and downstream system integrations that make a solution actually work.
Three real-world examples of what jaam built, and how they built it
Welbeck Health Partners
Welbeck Health Partners, a health practice in the UK, wanted to launch a premium advanced health check service. But when patients’ tests were sent to multiple specialist analysis clinics, the results come back in a completely different format from each one, and codifying extraction rules for every possible variation was impossible.
jaam’s solution used IDP for the PDF results and additional AI services for JSON files, training models to extract and unify the data into standardized scores that clinicians could act on. K2 handled everything else, from routing results to the right clinicians across different clinic configurations and managing approvals to generating the final patient reports.
As a result, report turnarounds dropped from 10 days to just two. Half of that gain came from AI. The other half came from K2 — specifically the visibility and routing capabilities that simply hadn’t existed before.
Global Construction Firm
A major construction company with 65,000 employees needed to process incoming CVs in dozens of different formats, standardize them, and integrate them into SAP SuccessFactors — both one at a time and in bulk uploads of thousands at once.
The one-off case was manageable: Extract the data, present it in a K2 SmartForm, highlight where confidence is high and where gaps exist, and let a human fill in the blanks. The bulk case was where things got trickier.
“When people upload 5,000 CVs, you can’t just say, well, if the confidence is less than 90%, don’t do anything,” Murphy said. “That’s not good enough. So we had to train the model and to find ways for K2 to surface those tasks at scale.”
The solution required heavy model training alongside smart workflow design, with K2 surfacing the right exceptions to the right people, at volume, without creating any bottlenecks.
Carpmaels & Ransford
Carpmaels & Ransford is one of Europe’s leading IP law firms and a K2 customer of 15 years. jaam has been working with them on a multi-phase, three-year transformation program covering re-platforming off legacy SharePoint and SQL environments, building a unified work management platform, and introducing AI-driven document and task handling.
“In their world, you cannot miss a deadline,” Murphy said. “If you miss a deadline on filing an IP patent, that’s game over.”
As such, jaam’s approach was deliberately cautious. Rather than leading with technology, the team gathered user stories and requirements first, then decided which processes would benefit from AI and which wouldn’t. Plenty fell into the “doesn’t need it” category, which the team considered a feature rather than a failure.
For those that did benefit, the AI layer handles document classification as correspondence arrives — whether from the European Patent Office API or via email — extracts key data from patent documents, and passes it through to K2, which kicks off the appropriate workflows. Office JavaScript APIs built for both Outlook and Word surface the same K2 task management interface across different working environments, and an upcoming MVP feature will use AI to auto-suggest the right task type and workflow based on incoming content.
The common thread between use cases
In each of these three real-world examples, AI handles the part of the problem that doesn’t scale: Classifying documents across wildly different formats, extracting data from thousands of CVs, and triaging inbound correspondence from dozens of sources.
And in each one, K2 handles everything that comes next: Routing, approvals, exception management, human review, and downstream integration.
No matter how fast AI changes, K2 provides the frameworks to support the work you’re doing, wherever you’re doing it, instead of embedded AI tools that tie you to specific models.
Learn more about automating without losing control, with Nintex K2.