The gap between organisations experimenting with AI and those actually getting results is widening fast. The numbers tell a tough story.
42% of companies abandoned most of their AI initiatives in 2025.
A full 80% of AI projects fail outright, which is double the failure rate of non-AI technology projects. And only 5% of organisations successfully scale AI into their day-to-day operations.
Recent research from BCG backs this up.
Just 5% of companies generate measurable value from AI in the form of revenue increases, cash flow improvements, and workflow enhancements at scale. 60% aren’t achieving material value at all.
The remaining 35% are scaling up, but they’ll admit they’re not moving far or fast enough.
So what separates the 5% from everyone else? After working with organisations across mining, health, government, and logistics, we’ve seen the pattern.
The successful ones aren’t implementing AI because everyone else is. They start with the problem, not the technology. They know exactly what they’re trying to fix and why it matters.
What AI actually does (and what it doesn’t)
Here’s what’s worth understanding about AI before you invest time and money into it.
It’s really good at four things:
- Analysing large volumes of information to find patterns
- Generating content and documents
- Predicting outcomes and identifying anomalies
- Automating repetitive tasks at scale
How leading enterprises approach AI
The companies that are winning with AI aren’t following some secret playbook, but they are doing three things consistently.
First, they prioritise action over perfection. Rather than waiting for the perfect solution, they act quickly, learn from experimentation, and iterate. This mindset shift is critical because analysis paralysis kills more AI initiatives than technical challenges ever do.
Second, they establish a dedicated AI capability. This isn’t just a team, it’s a capability centre focused on upskilling staff and collaborating with technology leaders. Building internal expertise matters more than buying tools.
Third, they embed AI across operations. Successful organisations don’t treat AI as a side project. They weave it into how the business operates, enabling thousands of decisions every day.
The 60/40 rule of transformation
Here’s something that might surprise you. Research from McKinsey, Deloitte, and Gartner shows that 60% of digital transformation value comes from improving existing operations. Only 40% comes from new products and business models.
This matters for AI strategy. Start with what you already do, then expand from there. Get good at using AI to improve your core operations before chasing entirely new business models.
The principles that work for digital transformation apply equally to AI: take small steps, be nimble and adaptable, deliver value early and often, and focus on learning and rapid experimentation.
Where AI is actually delivering value
BCG’s latest research reveals something important. The real AI value isn’t coming from flashy customer-facing projects. It’s coming from technology and operational functions.
The share of companies scaling or fully deploying AI in their tech operations tripled from 9% in 2024 to 28% in 2025. The productivity gains are substantial and measurable.
Companies using AI in software development are already seeing a 25% productivity boost, with expectations of 44% at full scale.
Data management shows similar patterns with 25% gains today and 45% or more expected at maturity.
Even compliance monitoring, traditionally a manual grind, is seeing productivity improvements exceeding 20%.
Here’s what the companies generating real value are doing differently. They’re not spreading AI thinly across dozens of experiments. They’re concentrating 60 to 70% of their AI budgets on deep implementations that handle complex, end-to-end workflows.
They define ambitious targets, like reducing approval times by 50%, and work backward to determine what’s required to achieve them.
This focus on depth over breadth keeps showing up. They’re not trying to ‘do AI’ everywhere. They’re identifying specific processes where AI can fundamentally change how work gets done, then executing relentlessly on those use cases.
Why humans remain critical
Even with all this automation capability, the human element matters more than ever. The organisations getting the best results from AI share common traits
- They thoroughly understand the business problem before seeking a solution.
- They ensure AI initiatives align closely with business goals.
- They involve stakeholders early to ensure relevance and adoption.
- They start small with iterative pilots, gathering feedback before scaling.
The most sophisticated AI system will fail if it solves the wrong problem or if the people who need to use it weren’t involved in building it.
Why adoption breaks down
Here’s the uncomfortable truth. Giving people access to tools like Copilot delivers modest individual gains, perhaps 5% time savings, but it rarely translates to enterprise-level impact.
Without structure, you get ‘shadow AI’ where staff use personal tools without governance or measurable business outcomes.
People will use AI tools anyway. The question is whether they’ll use them in a way that creates value for the organisation.
Success requires embedding AI into workflows, not just providing tool access. It requires governance that enables adoption rather than blocking it. And it requires clear policies, monitored usage, and iteration as you learn.
Modern AI platforms offer built-in controls. Utilise your own knowledge base, establish guardrails and parameters, and operate in modes that safeguard your data. The concerns about accuracy, data privacy, security, and misuse are valid, but they’re solvable.
Building trust through progressive adoption
Trust in AI isn’t built through presentations. It’s built through progressive exposure to real results. We recommend a three-stage approach that balances risk and value while building capability.
The first stage focuses on low-risk automation, transactional tasks where the downside of errors is minimal. This includes document generation and summarisation, knowledge retrieval and Q&A, data extraction and entry, and routine communications. These projects build familiarity and demonstrate value without putting critical operations at risk.
The second stage introduces AI-assisted decisions with humans in the loop. This includes complex document analysis with response drafting, workflow routing with recommendations, and anomaly detection with triage and prioritisation. The AI recommends, humans decide.
The third stage, autonomous execution, only comes after trust is established. This includes end-to-end request fulfilment, dynamic scheduling and resource allocation, and multi-step process orchestration. At this stage, AI decides and acts with minimal human intervention. But you earn the right to operate here by succeeding at stages one and two.
The rise of Agentic AI
The next wave of AI value is already here: Agentic AI. Unlike basic chatbots or simple automation, these systems can observe, reason, plan, and act autonomously. They execute complex, multi-step workflows with minimal human intervention.
According to BCG, agents account for about 17% of company-wide AI value in 2025, expected to nearly double to 29% by 2028.
The top 5% of companies already allocate 15% of their AI budgets to agents.
Where are they delivering value?
- Service desk automation that triages and resolves issues autonomously.
- Software development with automated testing and release management.
- Compliance monitoring with automated remediation.
- Infrastructure optimisation.
This is happening now in leading organisations, widening the gap between those who act and those who wait.
We’ll be exploring Agentic AI in depth in our upcoming series, including what makes these systems different, how to deploy them reliably, and when multi-agent systems make sense for complex workflows.
Getting started: Four principles that work
After delivering AI solutions across mining, energy, education, logistics and more, we’ve seen what actually works. Four principles keep coming up.
- Embrace rapid experimentation. Start small, validate fast. Use AI to prototype and test ideas in days or hours, not months. The cost of experimentation has collapsed, so take advantage of it.
- Get hands-on. Reading about AI isn’t the same as building fluency through practice. Upskill your teams through real projects. The organisations pulling ahead are the ones whose people are actually using these tools.
- Improve core operations first. Automate or augment high-impact existing processes before chasing new business models. Remember the 60/40 rule. Get good at the 60% before chasing the 40%.
- Measure everything. Build feedback loops and track the value you’re delivering. If you can’t measure the improvement, you can’t scale it.
What this looks like in practice
Theory is useful. Results are better. Here’s what we’ve actually built for organisations like yours.
Staff at an aged care provider were spending hours digging through policy documents just to answer basic questions. We built them a policy assistant that delivers accurate, sourced answers instantly. No more hunting through PDFs or second-guessing compliance.
A recruitment firm was drowning in reference checks. Every hire meant hours of phone tag and follow-ups. We automated the workflow, so their recruiters could focus on candidates instead of chasing references.
A transport company was manually coordinating their charter bus fleet. Inefficient routes, wasted fuel, buses sitting empty. We built a system that matches vehicles to routes automatically based on location, timing, and capacity. The result? Less empty mileage, lower costs, better utilisation.
A mining company needed to process thousands of environmental consultation queries. Manual handling wasn’t cutting it anymore. We delivered an AI platform in eight weeks that understands complex questions and searches across massive document sets. What used to take days now happens in minutes.
The real competitive advantage
AI isn’t the competitive advantage. The competitive advantage is in how quickly you can turn AI from potential into performance.
While many organisations are still exploring what AI might do, the leaders have already embedded it into their operations. They’re compounding their advantages while others are still running pilots.
BCG’s research quantifies this gap. Leaders report nearly 50% greater cost reduction and 60% higher revenue growth from AI compared to laggards. They invest twice as much in digital transformation, allocate double the staff, and have scaled twice as many AI solutions.
The gap is widening. The question isn’t whether to start. It’s how quickly you can move from experimentation to execution.
The technology is ready. The question is whether your organisation is.
We help organisations solve complex problems faster through the intelligent use of technology and AI. We design and deliver smart software, systems, and data solutions that improve performance, efficiency, and decision-making.
To discuss how AI can create value for your organisation, contact us at hello@horizondigital.au
