Insights

    Read all the insights from our team.

    Based on questions we receive each week.

    AI strategyMeasurementLeadership

    How do you actually measure AI ROI?

    Only 29% of executives can confidently measure AI ROI. The other 71% are running programs they can't defend at the board. Here's the framework that works for mid-market businesses, and the CFO test that filters real value from theatre.

    26 April 20267 min read
    Read the insight
    AI governanceAI leadershipRisk

    AI governance for mid-market businesses (without an in-house lawyer)

    75% of executives admit their AI strategy is theatre. Most of the gap is governance. Here's the lightweight four-pillar framework that fits a 200-person business without pretending to be enterprise.

    26 April 20267 min read
    Read the insight
    AI governanceRiskOperations

    Shadow AI: 45% of your AI adoption is invisible to IT

    Half your business is using AI for real work, on personal accounts, outside policy. Banning it makes the shadow harder to see. Here's the operating model that channels demand and blocks the genuinely dangerous fraction at the technology layer.

    26 April 20267 min read
    Read the insight
    AI strategyMeasurementPlanning

    How long until AI actually pays back?

    Executives expect 12 months. Reality averages 28. The gap kills more AI projects than the technology does. Here's the realistic shape, and how to align the board before the work starts.

    26 April 20267 min read
    Read the insight
    Data platformArchitectureAI foundations

    What is a data lakehouse, and do you actually need one?

    A lakehouse is one storage layer serving both BI and AI, built on cheap object storage with warehouse-style transactions. When it pays, when it doesn't, and what it costs in 2026.

    25 April 20267 min read
    Read the insight
    Data platformArchitectureProcurement

    Build vs buy: the data platform question for 2026

    For a mid-market Australian business in 2026, the honest answer is buy the core, build the opinionated layer on top. Here's why that beats both extremes, and when each end of the spectrum genuinely wins.

    25 April 20267 min read
    Read the insight
    Data platformDeliveryPlanning

    How long does it take to build a production-ready AI data platform?

    The honest timeline for a mid-market Australian business is 12 to 20 weeks, not six. Here's where the weeks actually go, and which of them you can genuinely compress.

    25 April 20267 min read
    Read the insight
    SalesAI solutionsRevenue

    Your sales team is calling the wrong leads. Here's the fix.

    20-40% of your pipeline value is sitting in leads your reps never call. Not because they're bad at their job, because the default behaviour of any rep under pressure is to work the warm list, not the high-value one. Here's how to change what's in front of them.

    25 April 20266 min read
    Read the insight
    Customer successAI solutionsRevenue

    Churn prediction that changes behaviour, not just dashboards

    Most churn models ship a beautiful dashboard and move retention rate by zero. The fix isn't a better model. It's a prediction tied to a specific play, assigned to a specific human, with a specific SLA. Everything else is decoration.

    25 April 20266 min read
    Read the insight
    Supply chainAI solutionsOperations

    Why your demand forecast is 40% wrong

    The median mid-market demand forecast runs at 30-50% MAPE at the SKU-week level. 15% is achievable. Here's where the error actually lives, and what it takes to close the gap without blowing the budget.

    25 April 20267 min read
    Read the insight
    AI solutionsAI agentsSemantic layer

    Can an AI agent actually answer questions about your business?

    Yes, reliably, if it has clean data access, a real semantic layer, and cited grounding. Without the three, you ship a demo that fools a board meeting and fails the first real Monday question.

    25 April 20267 min read
    Read the insight
    OperationsAI solutionsSupply chain

    Predictive maintenance: when it pays, when it doesn't

    PM isn't a universal upgrade on time-based maintenance. It pays when assets are expensive, failures are gradual, and data exists. When any of those is missing, the economics collapse.

    25 April 20267 min read
    Read the insight
    AI agentsAI solutionsArchitecture

    AI agents vs traditional automation: what's actually different?

    Agents reason. Automation executes. Most failed projects in 2026 come from picking the wrong one for the task. Most successful ones use both, carefully divided, in the same system.

    25 April 20267 min read
    Read the insight
    AI strategyDeliveryLeadership

    Why do AI projects fail?

    Most AI projects don't fail because the model was wrong. They fail because the business couldn't, or wouldn't, change the decision the model was trying to improve.

    24 April 20267 min read
    Read the insight
    AI leadershipAI strategyHiring

    Fractional Head of AI: what does it actually mean, and when do you need one?

    A fractional Head of AI is a part-time, senior AI leader who owns the strategy, the roadmap, and the vendor choices, without a $300k-plus salary. Here's when it fits, and when a full-time hire is the right answer instead.

    24 April 20266 min read
    Read the insight
    AI strategyProcurementPricing

    How much does an AI opportunity assessment cost, and what should you get?

    A proper AI opportunity assessment runs three to six weeks and either gives you a prioritised roadmap or tells you not to start. Here's what that scope actually buys, and how to interrogate a vendor proposal before you sign.

    24 April 20266 min read
    Read the insight

    Cookie preferences

    We use essential cookies to run the site. With your permission, we also use analytics cookies to understand what content helps visitors make better data and AI decisions.