Faster staff, same messy business. David Torgerson of Lucid Software breaks down why individual AI wins rarely add up to real ROI.
AI tools have quickly become standard issue across Australian SMEs, yet converting tool usage into genuine operational value is proving much harder. National AI Centre data shows SME AI adoption climbing to 44% in February 2026 – the highest adoption rate recorded in several months.
But as more employees adopt AI tools, a growing gap has emerged between staff using AI and businesses seeing a meaningful return from it. When I sit down with business leaders, the same question keeps coming up: We’re paying for AI tools, and our staff use them daily, so where is the actual return on investment?
Lucid’s AI at Work research points to a clear disconnect between individual productivity and organisational impact. While 66.8% of workers report AI making knowledge sharing faster and more efficient, 50.8% describe their company’s documentation as functional but incomplete. And for 21.6% of workers, those documentation gaps frequently result in inconsistent AI outputs. In other words, AI can accelerate how quickly people work, but it can only be as effective as the information and processes behind it. Making individuals more productive does not automatically make the organisation more productive.
The reason most 2026 AI strategies are stalling isn’t the underlying technology. True AI transformation occurs across three distinct phases, and many companies are skipping the most important one. Trying to make the impossible leap from Phase 1 straight to Phase 3 of AI adoption without the right context will prevent scaling it beyond individual use.
To understand why AI investments fail to deliver economic value, you have to look at how technology actually scales within an organisation.
Most organisations are in Phase 1. Employees use AI tools individually to research information, draft documents, analyse data, or automate routine tasks. This creates local speed, but zero organisational cohesion.
Phase 2, centred on visibility and mapping, is the critical operational foundation where processes, systems, dependencies and implicit rules are brought out of people’s heads and mapped visually into a shared source of truth.
Finally, Phase 3 represents institutional AI, the destination where human intelligence and autonomous AI agents coordinate seamlessly across end-to-end business workflows, operating at scale by coordinating people, processes, and systems.
The mistake leaders make is attempting to jump straight from Phase 1 to Phase 3. When you bypass visibility and mapping in Phase 2, you do not get transformation, only more output and less coherence. One employee drafts a proposal in seconds, while another uses an agent to summarise client notes, but neither output connects to the same business logic, context, or workflow.
You haven’t modernised the factory, but instead have simply put high-powered motors on individual workstations inside an unmapped warehouse. The result is faster activity at the local level, but duplicated effort and systemic friction.
Every business has a factory floor, even if it’s entirely digital. For a retailer, it may be the way a customer issue moves from an online enquiry to the warehouse and then to finance. For a growing SME, it may be the series of unwritten approvals and judgment calls required to ship a product update or close a contract.
The core challenge is that most enterprise knowledge is tacit or undocumented. It sits hidden in people’s heads, buried in messaging threads or scattered across spreadsheets. When organisations try to deploy institutional AI or autonomous agents into that environment without a visual blueprint, the AI is essentially operating blind. If an AI agent enters an unmapped process, it will guess. That ambiguity breeds inconsistent outputs, security risks, and confusion among employees. You cannot automate what you cannot see. If you layer sophisticated AI on top of invisible, broken processes, you simply accelerate chaos.
The businesses that will capture real, compounding economic value from AI are those taking a step back to execute Phase 2: making the invisible visible and documenting how work flows. Historical precedents prove this point. When factories first replaced steam engines with electric motors in the late 19th century, productivity did not jump for decades because owners simply swapped the motor while keeping the old steam-era floor plan. It was not until managers completely redesigned the factory floor around electricity that productivity exploded.
We are at that exact inflection point with AI. Before turning on automated agents or demanding 10x the output, leaders must map their current processes as they actually happen. This requires bringing cross-functional teams onto a shared visual canvas to explicitly clarify where human context starts the work, where AI proposes a draft or handles execution, and where human judgment must apply guardrails and make the final call. By explicitly mapping workflows, you establish a clear rule of engagement: humans lead, AI assists, humans finalise.
