Which SaaS is dead? Ask the VC who funds it.

Josh Farrugia calls himself a failed founder and a non-technical AI engineer. He has built Black Nova's own AI-native operating system on Claude Code, replacing tools the fund paid tens of thousands of dollars for, while keeping LP data, bank details and personal information well away from it.

So the sharper question becomes:

Which software should you build yourself, and what should AI never touch?

In this episode of Agents After Dark, Matt Doughty sits down with Josh Farrugia, Principal Investor at Black Nova VC, to talk about building real systems without a technical background, and why the best AI builds start with structured data rather than a model.

Together they discuss:

• Why "SaaS is dead" is true for some software and nonsense for the rest, and how to tell which is which
• How Josh learned to build: hackathons, workshops and borrowing best practice from builders who know more than he does
• Drawing the line on data: why LP data and bank details stay out of the operating system, and how encryption tools like CipherStash fit in
• How a team of five core staff and 26 venture partners supports 60 portfolio companies, with fund three adding more
• Measuring ROI through response times, intro tracking and backlogs cleared
• Why staff building 15 dashboards to watch their agents is a warning sign, and why enterprise adoption will follow the same path as the cloud
• Why, if we stopped making new models today, the world would still change overnight
• How deterministic to be before an agent touches anything, and when that starts to hold better models back

Josh's view is that the gap is no longer the technology. It is knowing what AI should do, what it should not touch, and being able to see what it is doing once it is running.

This conversation is for founders, investors and operators deciding what to build, what to buy and how far to trust agents with their data.

About Josh
Josh Farrugia is a Principal Investor at Black Nova VC, an early-stage fund now investing from its third fund. A former founder with a product background, he leads Black Nova's work on AI-native systems. Black Nova is an investor in Prefactor.

About Prefactor
Prefactor makes AI agents better.

Prefactor learns how your agents are supposed to behave, evaluates what they actually do on every run, and shows your team how to improve them. Engineers, product managers and the subject matter experts who know what good looks like all work from the same picture, so the agent gets better with every release.

Learn more at prefactor.ai

What stops an agent paying a $20,000 invoice in $499.99 chunks?

Hari Om Garg designs agentic systems for Fortune 500 companies at Wipro, where an agent might approve any invoice under $500 from a known vendor and send anything bigger to a person. Rules like that only hold if the agent can't learn its way around them, and the target is still a 15 to 25% return on investment in the agent's first year in production.

So the sharper question becomes:

How do you give an agent real authority without handing it a loophole?

In this episode of Agents After Dark, Matt Doughty sits down with Hari Om Garg, Managing Consultant at Wipro, to break down how he builds guardrails, confidence thresholds and human review into agentic workflows from the start.

Together they discuss:

• Why the $499.99 trick is a fraud detection problem, and how conditional access rules borrowed from cloud sign-in stop it
• How a confidence wrapper took one agent from 72% to around 90% accuracy, and why anything under 83% goes to a human instead of executing
• Why every engagement starts with a matrix of quick wins and high-impact use cases, each forecast to return 15 to 25% in its first year
• Why agents need their own identities, the way cloud platforms already give services managed identities
• Whether AI ROI is really headcount reduction, and why Hari measures what people do once the agent takes the repetitive work
• Why new use cases should come before rebuilding legacy RPA bots
• What a CEO under pressure should do in the next 30 days: cut the noise, spend two weeks finding real use cases, and ship quick wins inside three months

As agents move from answering questions to approving payments, the rules around them matter as much as the model. Hari treats guardrails as part of the architecture, designed with the use case rather than added after something goes wrong.

This conversation is for CIOs, CTOs and AI leaders who need to show a return on agents this year.

About Hari
Hari Om Garg is an AI solution architect and managing consultant at Wipro, where he designs agentic AI solutions for Fortune 500 companies. He has previously worked at Microsoft, HCL Technologies, LTIMindtree and PowerObjects, and writes for CIO.com on guardrails and intelligent automation for the enterprise.

About Prefactor
Prefactor makes AI agents better.

Prefactor learns how your agents are supposed to behave, evaluates what they actually do on every run, and shows your team how to improve them. Engineers, product managers and the subject matter experts who know what good looks like all work from the same picture, so the agent gets better with every release.

Learn more at prefactor.ai

Only 16% of an engineer's time is spent writing code.

That's the number Adam Witanowski keeps coming back to. Hand your engineers the best coding agent on the market and individuals accelerate 120% to 400%. But you've optimised 16% of the pipeline. Measured end to end, he saw between zero and 40% velocity gain, with everything bottlenecking at the PR.

So the sharper question becomes:

If the coding agents aren't the constraint anymore, what is?

In this episode of Agents After Dark, Matt Doughty sits down with Adam Witanowski, AI architect behind one of Australia's most comprehensive enterprise AI transformations, to break down what it takes to move an organisation up the agentic maturity curve.

Together they discuss:

• Why rolling out a coding copilot and calling the job done just bloats the middle of your delivery pipeline
• The harness he built across the full SDLC, and why rework and incidents beat adoption and token spend as metrics every time
• His audit of 10,000 historical PRs: roughly a third could have been automated, and if you're not ready to automate any of yours, your maturity is lower than you think
• Governor: policy as code injected into coding agents at invocation time and reasserted on CI, so cyber, legal and compliance shape every build without slowing the dev loop. Human in the loop, not human in the way
• Skynet: 3,600 repos ingested into a knowledge graph with decision history, so agents see the downstream impact of a change instead of breaking the team next door
• Why engagement went up while productivity went down, and what job anxiety and a weak front of funnel do to an AI rollout
• Winning over non-engineers: sit with the call centre, find the pain, and sell teams a solution rather than announcing their replacement

His test for where any organisation really sits: maturity isn't about the model. It's about how much you trust the agent when no one's watching it.

This conversation is a practical playbook for any engineering or AI leader trying to turn coding agents from an individual productivity toy into an organisational capability.

About Adam
Adam Witanowski is an AI Executive who spent the past year building one of Australia's most comprehensive enterprise AI transformations from the inside: 11 integrated agentic tools reshaping how engineering teams build software, reimagining the SDLC into an AI-DLC with full executive backing. He shipped 18 products to production in 12 months without writing a meaningful line of code himself.

He is now a consulting Chief AI Officer and keynote speaker on AI-powered engineering, with a background in bioinformatics and, formerly, the best job title in the industry: Chief Officer of Creative AI and Novel Engineering.

About Prefactor
Catch your agents' mistakes before they reach customers.

We evaluate every step your agents take, live. Mistakes get caught as they happen. Not in a Slack thread at 1am on a Tuesday.
Learn more at prefactor.ai