By Karthi Sepulohniam (Subramaniam), Head of Asia Pacific & Australia, Partners for Growth
Partners for Growth AI Series – Episode 1
Featuring Mark Woodland, CEO & Co-Founder, and Mat E., CTO & Co-Founder, Kismet Healthcare
There’s no shortage of AI content on LinkedIn. Most of it is noise.
So rather than add to it, we decided to go straight to the source: the founders and technical leaders actually deploying AI inside their businesses – not in theory, but in production.
This is the first installment of our AI Series, where we sit down with technology leaders across our ecosystem to understand how they’re using AI in ways that are practical, non-obvious, and genuinely changing how they operate.
We’re kicking things off with Kismet Healthcare.
Mark Woodland, CEO & Co-Founder, and Mat Ellis, CTO & Co-Founder
About Kismet
Kismet is building the digital infrastructure for Australia’s National Disability Insurance Scheme (NDIS). The platform connects participants with over 35,000 vetted providers, handling everything from plan management and invoicing to payments and fraud detection. The founding team, including CEO Mark Woodland and CTO Mathew Ellis, previously built and scaled Xplor Technologies.
2.5x Engineering Output – But Output Isn’t the Point
Kismet’s engineering team has seen roughly a 2.5x increase in output since adopting Claude Code as a core part of their workflow. But Mat was quick to reframe what that actually means.
The bottleneck in their business has shifted. It’s no longer about how fast code gets written – it’s about deciding what’s worth building in the first place. Tasks that used to take six weeks can now be completed in two days, which means the constraint has moved upstream to product thinking and prioritization.
Mat described it as roles collapsing: engineers now need to think like product managers, self-directing toward the highest-value problems rather than waiting for a layer of business analysts and PMs to hand them specifications. He expects this shift to play out across the industry within the next 18 months.
The Ferrari Analogy: Why More Output ≠ More Value
We asked Mark and Mat about the debate playing out in tech: should you hire fewer engineers and supplement with AI, or hire more because each one is now 10x more productive?
Mat’s answer was sharp. He estimates roughly 10% of engineers are genuinely strong – people who think in terms of business outcomes, not just code. AI is making that top cohort dramatically more effective. For everyone else, it may actually be widening the gap.
His analogy: giving someone a Ferrari is great if they already know how to drive. If it’s their first time behind the wheel, they’re probably going to crash.
Mark added a pointed observation about the trend of non-technical founders claiming they’ve “contributed to the code base” using AI tools. His view: if your definition of contribution is prompting an AI to write some code, you’re probably being put up with, and there are better uses of your time. The real skill isn’t generating output – it’s knowing what the outcome should look like before you start.
The Senior-Junior Paradox
One of the most interesting dynamics Mat described is what’s happening across experience levels.
Senior engineers – the ones with the judgment to evaluate whether AI output is actually good – are the slowest to adopt the tools. They’ve built their careers doing things the hard way and are hesitant to change.
Junior engineers are the opposite: rapid adopters, high output from day one. But they haven’t done the reps to know whether the AI’s output is correct.
Kismet’s response has been to push both groups in opposite directions: slowing juniors down to build foundational skills, while pushing seniors to pick up the tools and leverage their existing expertise.
It’s a problem Mat readily admits they haven’t fully solved – and one he thinks the broader industry will be grappling with over the next year.
The Non-Obvious Use Case: AI-Generated Support Documentation
When we asked about non-obvious applications, Mat shared one that stood out for its elegant simplicity.
Kismet has built a system that lets Claude reads their codebase, monitor code changes, and automatically update customer support documentation to stay in sync with the product. Instead of a documentation team manually updating help articles – which inevitably fall out of date – the AI keeps everything current in near real-time.
But the downstream effect was even more interesting. Non-technical team members now query Claude directly about the codebase to understand how business rules work inside the product. Rather than asking an engineer to read through code and translate, they go straight to the AI and get a clearer explanation than a human could typically provide on the spot.
It’s become an internal communication tool as much as a documentation tool.
“Claude is better at OCR than OCR”
One of the most counterintuitive things we heard was that Kismet found AI-based invoice processing to be more accurate than traditional optical character recognition (OCR) tools.
As Mat explained, the AI pipeline actually breaks the task into phases. The initial extraction phase – pulling raw data from a document – turns out to be remarkably precise. The hallucination risk people worry about tends to emerge later, when a model is asked to interpret the data, not when it’s reading the data. The extraction itself is highly reliable.
For a business processing massive volumes of NDIS invoices, that accuracy difference compounds fast. Fewer errors mean fewer manual corrections, faster payment cycles, and more trust in the system.
What’s Next: The AI Concierge
Looking ahead, Kismet’s most ambitious play is rethinking the product interface entirely.
Today, NDIS participants navigate a traditional UI – clicking through screens to check budgets, find providers and book services. Kismet’s vision is to collapse that entire experience into a conversational interface: an AI concierge that understands your healthcare data and can act on your behalf.
Want to check your home and living budget? Ask it. Need to book someone for home maintenance? Ask it. The underlying product infrastructure stays the same, but the interaction layer becomes a conversation rather than a sequence of screens.
The channel becomes flexible too – it could be delivered via WhatsApp, Siri, voice, text, or email. For elderly participants or people with disabilities who may struggle with traditional app interfaces, this is a fundamental shift in accessibility.
As Mat put it, it’s about engaging people on their terms instead of asking them to engage on yours.
The Bigger Picture: Stop Layering AI on Legacy Processes
Perhaps the most thought-provoking insight from the conversation was Mat’s view that most businesses are applying AI to processes designed decades or even centuries ago. Invoicing, for instance, hasn’t fundamentally changed in hundreds of years – you write some numbers on a document and send it to someone to pay.
The real unlock, in his view, won’t come from making old processes faster with AI. It will come from rethinking those processes entirely – designing from the ground up with AI, APIs, and modern infrastructure as the foundation, not a layer on top.
That’s a shift he thinks the healthcare sector, and many others, haven’t made yet. But it’s coming.
—
This is Episode 1 of AI: Where Innovation Meets Execution, a series where we profile how technology companies and investors across our ecosystem are deploying AI in practical, non-obvious ways. If you’re building something interesting with AI and want to be featured, reach out to us at karthi@pfgrowth.com.
Featured companies may be clients or portfolio companies of PFG. No compensation has been provided for inclusion. This content is for informational purposes only and does not constitute investment advice or a recommendation.
The views expressed are my own and do not necessarily reflect those of my employer.
This content is for informational purposes only and does not constitute an offer to sell or a solicitation of an offer to buy any securities. Any such offer will be made only to qualified investors through confidential offering documents. All investments involve risk, including the possible loss of principal. Past performance is not indicative of future results.
