Masterclass summary: what we’ve learnt in six months

Six months ago, our first AI at Work Masterclass explored why most AI projects fail to deliver their intended value (you can read the recap here). Since then, we've delivered more transformation projects with Australian businesses of every size and learnt a lot along the way.

In our latest session, our founder and CEO Jessica Trumble shared where AI is delivering, where it's falling short and what to do differently. Here are the key themes and takeaways. 

1. Where AI is delivering real value

Across the projects we've worked on over the last six months, AI is consistently delivering value in six main areas.

1. Workflow automation: AI is removing the admin burden from core people processes such as onboarding, offboarding, remuneration reviews and performance cycles. The critical piece is to automate the process steps and include people at the points that matter.
2. Data synthesis: AI is very good at ingesting large volumes of data and drawing inferences from it. However, controls for handling sensitive data are essential.
3. Research: AI can shape your logic and surface credible resources quickly. But be wary of an echo chamber, as AI models are prone to “people pleasing” and telling you your ideas are great.
4. Personalised learning: We're seeing a move away from mass catalogue content towards targeted, cohort specific learning. Built on an internal knowledge base with an agent sitting over it, this enables real time coaching and feedback on actual work, such as emails, reports and presentations. The result is a measurable capability uplift compared with generic L&D programs.
5. People insights: Reporting is shifting from retrospective to predictive, with real time signals. One example is a Slack based tool that asks employees whether they know what's expected of them, then heat maps the responses into a leadership dashboard. The goal is to get ahead of attrition and disengagement before they surface.
6. Commercial partnering: Agentic tools that pair business context with people frameworks are helping People teams articulate strategy in business language, and link their work to commercial outcomes.

💡 Tip: Before automating a people process, map the points where a human conversation changes the outcome, such as a first day welcome or a difficult performance discussion. Keep those human and automate the rest.

2. Where AI is falling short

Despite the investment, value realisation from AI is uneven. Here are the patterns we're seeing most often.

Capacity isn't being released. In our first AI Masterclass, we talked about moving people to higher value work. Six months on, many teams aren't reporting any material capacity release despite their AI investment. When asked, they can't articulate what they're doing with the time AI is supposed to be freeing up.

Individual gains aren't adding up. Individuals are getting faster, but those gains aren't grossing up into team level productivity.

Communication quality is slipping. AI generated outputs often lack clear messaging. There's a real risk of producing frameworks, roadmaps and policies at a speed the business simply can't absorb or adopt. 

False confidence is creating hidden work. AI output looks polished, but it can be inaccurate. Double handling, rework and starting again are a hidden workload that very few businesses are quantifying.

Engagement is taking a hit. Six months ago, we said we were in the era of AI slop. We're now seeing the consequences, including some businesses seeing staff engagement scores drop.

3. Quality control: the missing layer

Many of these AI shortfalls come from a lack of quality control. Speed without QA creates rework, risk and trust erosion.

We’re working with teams to build in QA, in several ways:

• Set human in the loop accountability. Someone must own and monitor agentic output. Define who that person is, what decisions the output informs and where the buck stops. This is a big mindset shift, from being human on the tools to being human in the loop.
• Define your preferred toolset and stick to it. Constant tool switching prevents meaningful progress. Only review your toolset every six months or longer, and only switch if a new tool shifts the needle by more than 20%.
• Build sign offs and audit points into automated workflows. Delegation authorities now extend beyond financial approvals to AI generated outputs. Decide who can approve what, just as you would for spending.
• Review your agents regularly. Run cyclical reviews monthly or quarterly, depending on complexity, and ask a simple question: is it still doing the right thing?

💡 Tip: Start with your highest risk workflow. Name the owner, set the review cadence and agree what good output looks like before you scale it.

4. The real differentiator is skills not technology

Perhaps the most important learning from this session is that technology isn't what separates the businesses getting value from AI. It's the skill set around it.

In our first masterclass, we flagged that the new age soft skills of discernment, critical thinking, communication quality and human judgment would set people apart. Six months on, we're even more convinced. The skills we see businesses underinvesting in are:

• Writing, proofreading and vocabulary
• Commercial acumen in written communication
• Influencing, presenting and in person impact

These are the skills that make AI augmented workers genuinely effective. AI can produce a draft in seconds, but someone still needs to judge whether it's right, sharpen the message and land it with the people who need to act on it.

That's why our L&D focus for next year is to double down on these human skills, then pair them with a defined toolset and clear guardrails. 

💡 Tip: Build an internal knowledge base of your standards, frameworks and competencies. Then layer an agent over it to personalise learning journeys and provide real time coaching feedback.

5. The main takeaways to act on

Jess closed the session with the following takeaways:

1. Progress is real but value realisation is uneven. Pursue initiatives that are high value and iterable.
2. Aim agentic and automated capability at repeatable predictable work. That's where the returns are most reliable.
3. Build quality control in from the start. Have a clear measure of what good looks like before you scale AI use or automations.
4. Invest in the human skills that sit above the technology. They're what turn AI output into business outcomes.

Six months ago, we said AI transformation is a people transformation, not a technology project and everything we’ve seen since has reinforced that. The tools keep getting better, but the businesses getting real value are the ones investing just as heavily in their people, workflows and quality standards.

We hope you found this article useful. If you'd like to listen to the full session, you can find the recording here:

And if you'd like to discuss AI strategy and implementation, please don't hesitate to reach out and book a time with Jess.

‍