Lorem ipsum dolor sit amet, consectetur adipiscing elit. Ut elit tellus, luctus nec ullamcorper mattis, pulvinar dapibus leo.
What Deloitte’s AI fumble teaches us about leadership, accountability, and why “human at the helm” beats “human in the loop” every time.
When Deloitte – one of the Big Four, armed with thousands of consultants and sophisticated systems – had to repay nearly $98,000 of a $440,000 fee after an AI-generated report contained factual errors (AFR, 2025), it made a lot of service business leaders stop and wonder: could this happen to us?
It was wake-up call for teams using AI without clear oversight.
The headlines made it sound like an AI failure. It wasn’t. It was a leadership failure.
Because when no one’s at the helm – when oversight, context, and accountability are missing – even the most sophisticated AI can steer off course.
💡 If you only read one thing:
AI doesn’t fail leadership does. Human-led strategy beats shiny tools every time.
“Reducing digital skills gaps among businesses could represent a $25 billion economic uplift for Australia by 2035.” – Deloitte
AI doesn’t “hallucinate” in a vacuum. It hallucinates because no one double-checks, questions, or applies judgment before the result leaves the building.
That’s not a software bug. That’s a management gap.
As one Deloitte partner reportedly put it, the issue wasn’t intent; it was “oversight.” In other words, the tools did what they were told. People just weren’t leading the process.
This is why at Dovetail, we’re developing our upcoming SMB AI Playbook – a strategic field guide to help service-based SMBs build structure and safe experimentation around AI.
Like many leaders, we’re frustrated by the easy allure of “free” or flashy AI tools that promise transformation but deliver confusion. The Playbook is designed to help small businesses cut through hype, install guardrails, and begin turning curiosity into capability.
Guardrails don’t slow innovation – they enable it.
We’ve all heard the phrase “human in the loop.” It sounds sensible, until you realise it often means the human comes in after the damage is done.
“Human in the loop” suggests you’re there to check the system. Somehow you’re working for the AI not the other way round.
“Human at the helm” means you lead the system.
That distinction matters – especially for small teams trying to stay compliant, competitive, and credible.
According to McKinsey’s 2024 Global AI Report, only 21% of companies have successfully scaled AI across more than one function – and lack of leadership clarity is a key barrier.
“As organisations increasingly rely on AI (and AI agents), leaders must proactively address decision rights, accountability frameworks and power dynamics — or risk the system governing you instead of you governing the system.”
MIT Sloan / TCS, “Winning with Intelligent Choice Architectures”
If Deloitte can get it wrong, so can anyone.
But that’s not a reason to pause AI – it’s a reason to lead it.
There’s a time for exploration.
There’s a time for play. But if your business is relying on AI in even small ways – for insights, decision-making, or client deliverables – someone needs to own it.
Because drift doesn’t create value. Leadership does.
💡 Quick Note: In small teams, AI drift rarely makes noise. It just quietly misses the mark. Whether you’re on your own or leading across locations, oversight is where alignment really starts.
Before you add more tools, try asking:
1. Who’s accountable for AI oversight in our team?
2. Where are our current AI experiments being tracked and reviewed?
3. What’s one process we could improve with a human-led AI workflow?
AI won’t replace leaders. But it will expose the difference between those who lead transformation and those who delegate it.
👉 Ready to test your digital maturity? Start with our 3 Question Snapshot – it’s free, fast and revealing.
Or explore The Digital Maturity Journey to understand the stages of digital evolution, and how AI fits into the bigger picture.
Curious how smart teams are turning prompts into repeatable outcomes? Read our latest blog Beyond Prompts.
Building AI Capability in SMBs
The error stemmed not from the AI tool itself, but from a lack of leadership oversight. No one properly reviewed the output before it was finalised.
It means someone is leading the AI system with intent, judgement, and responsibility – rather than passively supervising it after the fact.
Start small, install oversight loops, and assign clear ownership. Dovetail’s SMB AI Playbook will offer templates, checklists, and governance tools to help.
Yes – but only if it’s guided by clear parameters and accountable humans. You don’t need to be a data scientist to lead well, just strategically curious.
Start with clarity, not cleanup. Take stock of what’s working, define ownership, and implement a lightweight oversight framework like our Digital Snapshot.
