Short answer: in a mid-market business, AI pays for itself fastest on expensive, repetitive processes you already pay for, on coordination work that lives in people’s heads, and on decisions slowed down by scattered information. It rarely pays when there’s no named process, owner and measure behind it.
Every board I speak with is asking some version of the same question: “What’s our AI plan?” The honest answer in most mid-market businesses is a handful of pilots, a few enthusiastic individuals and no clear line to the P&L.
That’s not a failure of ambition. It’s a failure of focus. Here’s where I’ve seen AI pay for itself, and where it tends not to.
Where the return is
1. Expensive, repetitive processes with a clear cost today
The best place to start is a process you already pay a lot for, that follows a repeatable pattern, and where “good enough plus human review” is acceptable.
Translation is a good example. At a global consumer brand I moved translation and localisation onto an AI-first platform covering 14 languages, with people reviewing rather than translating from scratch and specialists brought in through a marketplace when needed. It saved around US$1M a year. The return was obvious because the cost was already visible.
Look for the equivalent in your business: document processing, first-draft content, data entry between systems, routine customer queries.
2. Coordination work that lives in people’s heads
A lot of operational cost is hidden in coordination: chasing approvals, copying information between systems, reminding people what comes next. Agentic workflows, where software takes a sequence of actions across tools, are well suited to this.
At the same consumer brand, one intake form now triggers 450 launch tasks across teams, regions and agencies. That’s automation rather than AI in the narrow sense, but it’s the same principle: take the work that depends on memory and make it systematic. AI makes this kind of workflow far easier to build and maintain than it used to be.
3. Decisions that are slowed down by finding information
Many decisions are slow not because they’re hard but because the information is scattered. Pulling it together, summarising it and putting it in front of the right person is something current AI does well.
Where the return usually isn’t
- “AI strategy” with no process attached. A strategy document that doesn’t name a specific process, owner and measure is a wish list.
- Pilots without a production path. If nobody has decided who will own the tool, how it connects to your systems and how you’ll measure it, the pilot will stay a pilot.
- Replacing judgement where the cost of error is high. AI is excellent at first drafts and pattern-matching. Keep people accountable for the decisions that really matter.
- Buying a platform before defining the problem. It’s easy to spend six figures on licences and still not know what you’re trying to fix.
A simple test for any AI initiative
Before you fund it, ask four questions. (If governance is the gap, start with my AI governance checklist for Australian boards.)
- What does this process cost us today, in money or time?
- Who will own it once it’s live?
- How will we know it’s working, and by when?
- What happens when it’s wrong, and who catches it?
If you can answer all four, you have an initiative worth funding. If you can’t, you have a pilot that will probably stall.
Where to start
If you’d like a structured view of where your business stands, try my free AI Readiness Scorecard. It takes about five minutes and looks at strategy, data, process, people and governance. If you’re wondering who should own this, read does your company need a Chief AI Officer?, or see how I work as a fractional Chief AI Officer. Or book a 30-minute call and we can talk through your specific situation.
