How to Integrate AI into Daily Operations: Redesign the Process First
· Kirsty Dove

Most AI rollouts I get asked to fix didn't fail because of the tool. Someone bought it, switched it on, told the team to have a play, and a few weeks later everyone was back doing things the old way.
So if you're wondering how to get AI into your daily operations, here's what I'd do: sort out the process first, then add AI where it actually helps. That's the whole idea behind how I work. Process first, then AI.
Here's how it goes in practice.
Is this for a team or an owner-run business?
The steps are the same either way. What changes is who's involved and how much there is to untangle.
If you're an owner-led business with one to ten people, you're probably the office as well. The work is mostly about finding the one or two jobs eating your week and fixing those first. It can be done in an afternoon with you, not a fortnight of interviews. That's what my small business work is built around.
If you've got a team of roughly ten to a hundred, the challenge is getting everyone using AI the same sensible way, not just the one keen person. That means talking to more people, agreeing ground rules, training the whole team and leaving someone inside the business to keep it going. That's what my teams work covers.
I'll flag where the two paths differ as we go.
What does "integrating AI into daily operations" mean?
It means AI is part of how the work gets done on a normal Tuesday, not something one person tried once and left alone.
A task moves through a set of steps. AI handles the ones it's good at. A person handles the ones that need judgement, care or their name on the result. It's not about giving everyone a chatbot login and hoping for the best.
It's about the workflow, not the software.
What's the difference between AI strategy and AI implementation?
They get used as if they're the same thing, but they're not.
AI strategy is deciding where AI will help your business most, and what you're going to leave alone. It covers priorities, risks, policy and who's responsible for what.
AI implementation is making it happen. Building it into the work, training the people doing the work, and checking it's actually saving time.
You need both, in that order. Implementation without strategy gets you a pile of tools nobody asked for. Strategy without implementation gets you a nice document in a drawer.
Step 1: Map how the work really happens
Pick one area, like enquiries, quoting, onboarding, invoicing or reporting. Just one.
Write down each step the way it's done today, not the way the manual says. Include the spreadsheet on someone's desktop and the sticky note on the monitor. For each step, jot down:
- Who does it
- How long it takes
- What goes in and what comes out
- Where it gets stuck, repeated or fixed by hand
Ask the people doing the work. They already know where the annoying bits are, and they're usually happy someone's asked.
- Owner-led business: this is mostly you describing your week, from the first enquiry through to the invoice. An afternoon is usually enough.
- Team: talk to leadership and the people doing the work, 30 to 45 minutes each. The gap between how the work is meant to happen and how it does is usually biggest here.
Step 2: Find the steps worth changing
Not every step needs AI. If the honest answer is "none", I'll tell you so.
The ones worth a look are usually:
- Repetitive: the same kind of task, many times a week
- Heavy on words or data: drafting, summarising, sorting, checking
- Mostly routine: with a few exceptions that go to a person
- Low risk: or easy to review before it goes anywhere
Leave alone anything where a mistake would be costly and hard to spot, or where the personal relationship is the whole point. Your best client doesn't want an AI-drafted apology.
Step 3: Redesign the process, not just one step
This is where most rollouts go wrong. If you swap one manual step for an AI step and leave the rest, the time you saved gets lost in handovers and rework.
So ask:
- What can be dropped altogether? Some steps only exist because of an old limitation.
- What can AI draft, and what does a person need to approve?
- What does AI need at the start? Good inputs matter more than clever prompts.
- Where does the result go next? If it lands where nobody looks, it's wasted.
- What happens when it gets it wrong? Decide that now, not afterwards.
Then draw the new process on one page. If it won't fit on one page, it's too complicated for everyday use.
Step 4: Give your team a few simple rules
People are far more confident with AI when they know where the edges are. Keep it to one page that covers:
- Which tools are okay to use
- What should never be pasted in (client details, financials, anything confidential)
- What always gets a human check before it goes out
- Who to ask when they're not sure
Plain words work best. People will actually read them.
If it's just you, this is a one-page "how we use AI here" for you and anyone who helps out, so you're not the only one who knows. If you've got a team, it matters more, because staff need to know what's allowed on day one. It's worth a short briefing so it gets read rather than filed.
Step 5: Start small, then train properly
Run the new process with one team or one workflow for four to six weeks. Keep it small enough that you can see what's happening.
Make the training about your real tasks, like "here's how we handle a new enquiry now", rather than how AI works under the bonnet. Give people time to practise and make it okay to say "this isn't working for me".
- Owner-led business: one job at a time, built and working on your real work. A monthly session with someone in your corner is often enough to keep it moving.
- Team: hands-on sessions built around your team's own tasks, then a few internal champions who the rest of the team can ask. That's what stops it fading once the trainer's gone.
Step 6: Check it's working, then do the next one
Pick two or three things to track before you start:
- Time taken per task
- How often something has to be fixed or redone
- How quickly customers get a reply
- What the team thinks of it
That last one matters. If the team hates it, it won't last, however good the numbers look.
Review after the pilot, tidy up what's clunky, then move to the next workflow. Each one gets easier because the habits are already there.
Mistakes I see a lot
- Starting with the tool instead of the problem
- Automating a broken process, which just gets you the mess faster
- Leaving the team out, so they quietly work around it
- Trying to do everything at once
- Having no one in charge of how AI gets used day to day
Where to start this week
- Pick one process that drives your team a bit mad.
- Spend an hour mapping how it really runs.
- Circle the two or three steps that are repetitive and low risk.
- Sketch the new version on one page.
That hour will tell you more about your AI strategy than any tool demo.
Frequently asked questions
How do I bring AI into my business without disrupting everything? Start with one workflow. Redesign it on paper, try it with a small group and look at the results before going further. Nothing I do takes your team offline.
Do I need a big budget for AI implementation? Not to begin with. Most early gains come from fixing how the work flows and using tools you might already pay for. Spend more once something has proven itself.
How long does it take? One workflow usually takes four to six weeks from mapping to a working pilot. Across a whole business, it happens one workflow at a time.
Should I hire a consultant or do it in-house? Either can work. Outside help is most useful for the discovery, the policy and the training, where fresh eyes save time. Your team still runs it day to day. And if you don't need me, I'll say so.
Why do AI rollouts fail? Usually the process wasn't redesigned and the team wasn't trained, so people drift back to their old habits. It's rarely the tool.
Want a hand getting it sorted?
I work with teams and small businesses across New Zealand to work out where AI fits, build it into the way they work, and train the people using it.
- Run a small, owner-led business? Have a look at how I work with small business owners.
- Have a team to bring along? Have a look at how I work with teams.
- Not sure which fits? Book a thirty-minute call. No slides, just a chat about your business.
Kirsty
