The Four Investments That Make AI Tools Actually Work

Most CEOs treat AI tool spend as the investment. The real investment is the four things that determine whether any tool ever works inside your organisation.

Your organisation bought the tools. Probably more than one. The contracts are signed, the vendor is onboarded, and someone in IT gave a demo that looked impressive. Six months later, the people who were supposed to use those tools are still doing the work the old way, and the tools are running at a fraction of their designed capacity. You are not behind on implementation. You already made the investment that will determine whether this works — and it was not the software.

The vendor cannot fix what you have not built

The strongest argument against this position is worth taking seriously. Modern AI platforms are not blank canvases. They arrive with embedded workflow logic, pre-built guardrails, and vendor support designed to reduce the internal expertise you need to operate them. Procrewment argues that commercial tools with light customisation often outperform a build-first approach, particularly when internal expertise is scarce. McKinsey Academy's structured external programs exist precisely to fill the skill gap without requiring the organisation to own the full development process. These are legitimate inputs.

They fail at a specific point. An external program delivered into an organisation without leadership accountability and co-owned responsibility does not produce lasting capability — Acorn PLMS is explicit on this. The vendor configures the platform. Nobody configures the organisation.

What the research actually shows

Oxford University Press published a study in 2007 that named the core problem directly: firms consistently acquire technology faster than they build the knowledge to use it well. That was written about enterprise software. The pattern has not changed. The IRSPP academic report from 2015 goes further and treats capability as a distinct management problem from tools — one that requires its own attention, its own investment, and its own sequencing.

BCG's 2013 procurement capability research identified what actually builds durable skill: not training events, but skills developed through real work, pilots, and daily reinforcement. One-off training programs are the organisational equivalent of buying a gym membership and expecting fitness. [Inference — this pattern from procurement capability research maps onto AI adoption by structural similarity, not by direct AI-specific evidence in the supplied research.]

The four investments, in order

The Victorian Government's procurement capability guides from 2018 and 2020 give the clearest list of non-technology levers. They are not glamorous. Governance — who decides what the tool is used for, and what counts as a bad outcome. Skill development embedded in actual work, not classroom hours. Management accountability, meaning someone's performance review reflects whether the team is building capability or just logging in. Workflow redesign, because AI tools inserted into unchanged processes produce AI-speed versions of the same broken output.

The sequencing matters. Gap assessment comes first — you need to know what your people do not know before you design training that addresses it. The Victorian Government's 2018 guide is explicit on this. Governance design comes next, because without it, pilots produce local wins that nobody can scale. Skill development runs alongside pilots, not before them. Workflow redesign is last, because you cannot redesign a process you do not yet understand well enough to change.

What this costs you if you skip it

I have watched organisations spend eighteen months on Copilot rollouts and end up with staff who use it to reformat emails. The tool was fine. Nobody had defined what good AI use looked like in that organisation, nobody was accountable for adoption, and the workflows the tool was dropped into were not designed to take advantage of what it does. That is not a technology problem.

The uncomfortable part of the Oxford study's finding is that the gap between what firms buy and what firms know how to use does not close on its own. It closes when the organisation deliberately builds the internal capability to make decisions about how the tool fits its specific work. Vendors do not do that. Consultants visit and leave. The four investments — governance, embedded skill development, management accountability, workflow redesign — are the work that stays.

Start with the gap assessment. It takes four weeks and produces a list of what your organisation does not know. That list is more useful than another tool.