AI: From paralysis to progress

by Jacob Gunzburg, CFO, Fusion5

Balancing the AI books

As a CFO for a technology business, AI is both a compelling efficiency play and a budgeting conundrum. Treat AI like any other capital investment - but with tighter guardrails, shorter payback expectations, and more caution. 

Start small, move (and fail) fast 

The "fail-fast" approach becomes crucial in AI project management. Rather than committing large sums to uncertain outcomes, set clear financial thresholds and kill switches, such as stopping projects that reach 60% of their budget without demonstrating value. 

We’ve discovered that early wins often come from small, targeted tools built in relatively short periods of time. From experience, we’ve found that lightweight builds minimise incremental cost, create quick proof points, and reveal where to scale - or where to stop. 

Yes, ROI matters (but measure it differently) 

ROI is still king – even if you have to bend the ruler! Traditional ROI calculations become complex with AI projects, especially for back-office functions. AI benefits often manifest as increased capacity rather than reductions in headcount. Measure "hours back" for high-value activities and quality improvements instead of immediate OPEX savings. 

The challenge lies in accurately modelling both costs and benefits. My advice:   

• Start with the problem, not the technology. For efficiency goals, quantify released hours and translate to dollars. For accuracy/insight goals, define acceptable error rates or cycle times. 
• Model payback the same way you would for any other capex: upfront cost plus a contingency, versus discounted cash benefits over time. AI solutions can continue learning and improving over time, potentially creating accelerating returns that are difficult to predict.  
• Aim for a short payback window. The rapid pace of technological advancement means that what’s cutting-edge today might become obsolete not long after.

Build real guardrails 

AI is still greenfield. Unknowns are high, and technology can leapfrog at pace.  

That means: 

• Use stage gates and kill switches.  
• Add contingency to both cost and timeline, be conservative on benefits. 
• Track adoption and throughput changes, not just model accuracy. 
• Model the investment like any capex with a DCF: upfront build plus ongoing costs vs. a multi-year benefit stream.  

Prioritise the right problems 

AI shines where grunt work dominates - document review, reconciliations, reporting prep - freeing teams to focus on value-added analysis. In finance, specifically, robust AI solutions are still emerging – so remain open and experiment while resisting big bets without clear outcome visibility. 

My recommendation to fellow CFOs: treat AI as capital investment with stricter controls. Start small, measure relentlessly, and scale only when math and adoption prove successful. This approach protects downside risk while capturing upside potential in this rapidly evolving landscape. 

by Shannon Moir, AI Director, Fusion5

Three AI challenges every CIO should anticipate – and how to stay ahead of them

When I put on my CIO hat and sketch the first AI-enabled transformation, three warning lights start flashing long before any code is written: organisational change, data and security, and integration.  

Data and security traditionally get top billing, followed to a lesser degree by integration, so I’ll acknowledge them briefly and then focus on the less visible – but most dangerous – threat to successful transformation: people and process. 

1. Data and security 

Yes, you need clean, consented data and a zero-trust security posture. But the twist with modern agentic AI is scope. By limiting each agent to a narrow, “purposeful and proactive” slice of data, you can shrink the blast radius of a breach and simplify compliance with your organisation’s trustworthy AI guardrails. 

2. Integration

AI agents are designed to make decisions and act on them. However, to do so effectively, they need good integration with endpoints such as your ERP, CRM, and ticketing platform. So, look early at API maturity, message queues and identity management. If integration is brittle, the most sophisticated model in the world will remain idle. Budget for modernisation sprint(s) before you unleash any autonomous capability. AI can reuse all of that hard work that you’ve already done. 

3. Organisational change – the silent killer

Despite the headlines, most AI projects stall not in the cloud but in the corridor conversations outside it. Employees fear “the AI robot that takes my job,” while business leaders struggle to articulate which processes are suitable for automation.  

To help overcome these challenges, I use four tactics: 

• Choose your processes carefully – be conservative (the 5% rule). In year one, target to help just 5% of your workforce with AI. Not 50%, not "enterprise-wide transformation," just 5%. Why? This seemingly modest target forces you to do something that most organisations skip entirely: build proper foundations. The 5% rule creates natural constraints that drive good decision-making through: choosing your project carefully (forget the vanity projects), building proper guardrails first, measuring what matters, and learning before scaling. 
• Inspire organisational confidence. Build human trust in AI agents (and their ability to make good decisions) through a well-designed onboarding process. Hire an AI agent the same way you hire a human: give it a job description, success metrics, and an engaged human mentor or boss. Why? Because without a feedback loop and correction, an agent will continue unchecked - unaware that it’s making wrong decisions – just as an untrained intern would.    
• Embed observability as culture. Every decision an agent makes must be explainable and traceable. Ensure that you have dashboards displaying where the model deviates, why, and how you retrained it to transform abstract governance into concrete safety nets that your business stakeholders can trust.  
• Install guardrails. With a coherent AI strategy and framework in place, you’ll have guardrails to guide the AI transformation, ensure the project and decisions are auditable, the process is trustworthy, and the social response is thoughtfully considered.  

What are the signals that you’re losing the organisational battle? 

by Kristy Brown, CEO, Fusion5 NZ

Don’t let AI paralysis stop you in your tracks 

From the CEO’s chair, AI governance isn’t optional - it’s the foundation of safe, scalable value. But it mustn’t become a reason for paralysis. The winning approach is top-down intent with pragmatic execution: tighten what matters, move fast where you can, and use today’s tools to bridge the gap. 

Understanding where and why governance matters 

Lockdown, or let loose? 

Modern agent frameworks let you specify exactly what an agent can do, what it cannot, and which data it can touch. That means you avoid broad exposure while still moving forward. Combine this with a rapid data readiness assessment to identify which domains are safe to unlock now and which need phased remediation, and your AI transformation roadmap will start to take shape. 

Move forward, move fast 

The mandate for leaders: set governance as a strategic priority, seek expertise to work smart, and run both motions in parallel. Don’t delay value while you perfect everything. Govern decisively, start narrowly, expand confidently - because in an AI world, speed with safeguards beats slow perfection every time. 

At Fusion5, we’ve always taken pride in being forward-thinkers. As you’ve probably gathered by now, we like to ‘go beyond.’ 

However, over the past year, something fundamental has shifted in how we operate - not just as a business, but as a collective of problem solvers. This change reflects a new leadership conviction, a shared vision, and the genuine excitement of reimagining what’s possible for us and our customers.   

When we first championed AI adoption internally, it was, in part, an acknowledgement that we couldn’t ask our customers to embrace something that we were holding at arm’s length. However, we could also see that AI represents a significant shift in how we can deliver value – it’s a tool to elevate our people, amplify their potential, and redefine service excellence, which is why we are so committed to our internal AI transformation. 

The irreversible shift   

There’s a phrase we use internally: “Once you see AI’s potential, you can’t unsee it.” It’s like trying to reimagine life without electricity or the internet.  

Our executives live this potential daily. We’ve moved from theory to action so quickly that even our meeting rhythms have transformed. Decisions that once took months now happen in weeks. Why? Because we’ve embraced progress over process.   

This cultural shift starts at the top. As leaders, we don’t just endorse AI - we “mandate” it. We’re actively demonstrating that transformation isn’t a distant goal but a present reality. Whether it’s streamlining operations or enhancing customer deliverables, we’re proving that AI isn’t a threat to expertise - it’s an accelerant.  

Something that demonstrates this shift at a grassroots level is that we no longer prepare proposals that don’t feature AI.  

And in the last month, we’ve onboarded four new digital Fusion5 employees (AI agents) to transform workloads and service delivery. For example, Agent Claire is dedicated to reviewing contracts - driving significant internal efficiencies and improving our response times from weeks to days. Agent Stella specialises in user story creation, dramatically reducing the time it takes for customers and our project teams alike to create user stories, and shifting the effort to business alignment, review, and impact. Stella is fundamentally enhancing how we bring value to customers and projects. (It’s worth noting that none of these new digital employees require consolidated integrated data to get going and make an impact on how we do business.)  

Progress over process: Leading with agility