The $10 AI download that raids your finances

Somewhere out there is a $10 skill pack that thousands of people have cheerfully downloaded without reading. A couple of hidden lines in one of those, and it could be raising invoices through your connected Xero before anyone notices. Nyssa Waters would like a word before you hit download.

Produced by Europa Creative Partners (europa.nz).

In this episode of Slideshow, host Dave Hayward sits down with Nyssa Waters, CEO and co-founder of Possibl AI and RCRT (Right Context Right Time), and one of the most brilliant minds working in AI right now. She led AI adoption at Google Cloud before building her own company to over $4M in revenue in under two years, and she moves from security architecture to the future of the working week in the same breath. The conversation covers serious ground. What an AI agent actually is, for a start (most of what people call agents is automation with better PR). Then the risks that come with wiring AI into everything your business owns: skill poisoning, tool drift, downloaded skill packs nobody reads. Every connection is a point of vulnerability, and most businesses have no idea what they’ve granted access to.

The warnings are only half the story. Nyssa has a hefty bet running with a US venture capitalist that he’ll stop opening his email within two years, and she’s building the case for why she’ll win: her company is replacing 90% of its 42 SaaS subscriptions with a system her own team built. Their first mission was, in her words, destroying Jira. From there it ranges wide: purpose-led teams, why nobody should write a three-year AI strategy, and her belief that AI can restore us to being human again. If you want to know where all this is heading, an hour with Nyssa Waters is a very good place to start.

Links:

Nyssa Waters, LinkedIn: https://www.linkedin.com/in/nyssawaters/
Possibl AI: https://www.possibl.ai/
RCRT: https://www.possibl.ai/rcrt/
Possibl AI, LinkedIn: https://www.linkedin.com/company/possiblai/
Dave Hayward, LinkedIn: https://www.linkedin.com/in/haywarddave/

Chapters

Introduction (~0:00)
AI Agents Defined (~1:42)
The Problem with SaaS (~5:26)
AI Security Risks (~7:56)
Future of Work & Purpose-Led Teams (~35:10)
Backend as a Service & RCRT (~47:37)
AI Restoring Humanity (~1:02:28)

FAQ

What is an AI agent and how is it different from automation?

Nyssa’s definition comes down to one word: agency. An agent can functionally perform a task without human input, which is a step beyond a chatbot conversation or a workflow that runs on autopilot. Most of what gets marketed as an agent today is really automation. True agents act autonomously within the scaffolding you set up for them, ideally with human-in-the-loop checkpoints where it matters.

What is skill poisoning and why is it a risk for businesses?

Skill poisoning (sometimes called tool drift) is when a popular downloadable AI skill or tool is altered to include hidden instructions. Someone launches something genuinely useful, it gets widely adopted, then a couple of lines are added that can do things your business would never know about. If that skill has access to your systems, your data, or your accounting software, the consequences can be serious. Nyssa’s advice: read what you download, and understand what it has access to before you connect it.

Will AI replace SaaS software like HubSpot and Jira?

Nyssa believes most SaaS won’t serve us in the future. Fragmented subscriptions mean fragmented data, stacked costs, and a vulnerability point at every connection. Her own company is replacing around 90% of its 42 SaaS subscriptions with custom applications built on a single backend, and their project management tooling now updates itself as work happens. Her view is that the industry is battling for a single pane of glass: one context-aware system in your language, built for you.

What is backend as a service?

Instead of every SaaS product bringing its own front end, database, hosting, authentication and billing, backend as a service flips the model. The backend (your data, connections, authorisation and context) becomes the core, and you build custom applications on top of it. Nyssa argues this is the architecture the agentic era needs, because agents are only as good as the context they can access, and fragmented backends starve them of it.

How long should an AI strategy be?

Twelve months, maximum. Nyssa will help clients set a North Star for 2030, but she won’t build anything longer than a 12-month AI strategy, because the technology is moving too fast for anything else to survive contact with reality. A build that cost one of her customers $2 million two years ago would cost $200,000 to $300,000 today. Prototype, build quickly, and get fast ROI before the next capability shift arrives.

Can AI actually make work more human?

This is Nyssa’s core belief: AI can restore us to being human again. Humans aren’t made for mass data analysis or endless email ping-pong, and the mental load of modern work crowds out presence with our families and teams. If the right information is served at the right time, the noisy and boring work falls away, leaving the creative, insight-driven, human parts. She even suggests the disruption could rebalance households, with shorter working weeks becoming genuinely possible.

Keywords

AI agents, agentic AI, future of SaaS, backend as a service, AI security, skill poisoning, MCP security, AI strategy, context engineering, single pane of glass, future of work, purpose-led teams, AI transformation, New Zealand business, AI agency