Most business owners can name at least one system that never quite stuck. A CRM everyone half-fills in. A shared drive with three versions of the same client list. A project tool the team quietly moved back to email for. The usual instinct is to blame the tool: wrong software, bad fit, too complicated. Two AI executives interviewed this month, working in industries that have nothing else in common, both pointed somewhere else entirely.
The “10% Technology, 90% People” Finding
In separate interviews with HRD Australia, Cathy Doyle, chief people and culture officer at Sydney insurer NobleOak, and Rochana Golani, vice president of learning and enablement at Databricks in California, both pointed at the same underlying gap: readiness and adoption, not the technology itself, decide whether AI transformation actually works. Golani put a number on it: “Every organisation that I talk to keeps bringing this up,” she said. “10% technology, 90% people.”
Golani also drew a distinction worth sitting with: a large share of businesses adopting AI are “scaling” old habits with new tools, doing the same things they always did, just with AI attached. Genuine reinvention, changing how the work actually gets done, is rarer.
Why This Isn’t Really an AI Problem
While Doyle and Golani were talking specifically about AI, the underlying adoption problem will sound familiar to anyone who has introduced a new system or process into a business. Swap “AI tool” for “new CRM,” “new booking system,” or “the process you finally wrote down properly,” and the pattern looks the same: sometimes the tool genuinely is the wrong one, a poor fit, too complicated, badly integrated, or missing something the team actually needed. But plenty of technically sound tools fail for a completely different reason: the business skipped straight from “we bought this” to “everyone should be using this,” with nothing in between. That’s the imbalance Golani’s 90% points to, and it’s consistent with what Doyle’s own interview described about workforce readiness at NobleOak.
What Actually Makes a Rollout Stick
A practical rollout plan should answer at least three questions:
Who’s going to use it first? Golani’s finding that “managers are the biggest change agents” scales all the way down to a five-person team. If the business owner is still doing the old workaround under the table, so is everyone else.
Does the team know why, not just what? “Use this from now on” is an instruction. “Here’s the problem this actually fixes for you” is a reason. Only one of those survives the first busy week.
What happens to the old way? Running the new system alongside the old one “just for now” is how “just for now” becomes permanent. At some point the shared spreadsheet has to actually get archived, not just discouraged.
What This Looks Like at 5 to 30 People
A solo operator adopting a new tool only has to convince themselves. In a 5 to 30 person service business, there often isn’t someone whose sole job is managing change. The responsibility tends to sit alongside someone’s day job, whether that’s the owner, an operations lead or another member of the team. That makes ownership even more important: someone needs to be responsible not just for getting the new system live, but for making sure it becomes the way the work gets done.
The Question Worth Asking Before the Next Rollout
Choosing the tool is only one decision. Before rollout, you also need to know who owns the change, how the team will learn the new way of working, what happens to the old process, and how you’ll know whether the new system is actually being used.
This is one of the reasons Smooth Ops looks beyond the software itself when diagnosing an operational problem: even the right tool can fail to deliver what the business expected if there is no adoption plan behind it. An Ops HealthCheck Audit is built to surface both, what system a business actually needs, and whether the conditions exist for that system to actually get used once it’s in place.
If this sounds familiar, it’s worth reading What We Check When a Business Says “We’re Just Too Busy” for the broader diagnostic picture, or What to Sort Out Before You Touch an AI Tool if the next rollout in question is specifically an AI one.