Success by a Thousand Paper Cuts

In our last episode, the AI promise I wrote about AI amplifying people, not replacing them. But it left a big gap: where do we start?

It's a fair question. The narrative is still dominated by "one big AI transformation" - a single initiative that'll change everything: a new platform, a sweeping automation project, or a slide deck and a Gantt chart that helps people feel like they're on the front foot.

The people I've seen get the most from AI aren't making one big bet, they're making a lot of small ones. Each improvement is tiny on its own. A few minutes shaved off a process. Something caught that would've been missed. Someone gets time back to think. Individually, none of them will make a board presentation but collectively they change everything.

I've started calling this "success by a thousand paper cuts". Paper cuts are supposed to be bad but here, each one is a small, positive incision into how you work - a micro-improvement. And like compound interest, they stack. Twenty small improvements across five teams is a hundred cuts. Give it six months and you've shifted how your organisation operates, without ever betting the farm on a single initiative.

Three domains

As we've worked with clients, a framework has emerged for where those cuts can happen. This article introduces it, and over the coming weeks I'll go deeper on each part. We think about AI adoption across three domains: inward, outward, and forward.

Inward — where your team meets AI

Inward is the internal stuff: how your people work, communicate, make decisions, and manage their cognitive load day to day. This is the most applicable because every organisation has it. You don't need to be a tech company. You don't need a product. You just need people getting shit done.

The example I keep coming back to is call recording and transcription. Someone on your team is running a client call. They're presenting, listening, trying to take notes, mentally flagging follow-ups. Give them a tool that records, transcribes, and summarises that call with action items, and you've freed up a significant chunk of their mental bandwidth. It's not revolutionary. It won't make the news. But multiply it across every person who takes calls in your organisation and the compound effect is real. That's an inward cut.

Outward — where your customers experience AI

Outward is the touchpoints where customers interact with your systems and your service - and most of the time, they have no idea AI is involved.

Next time you're at a Coles or Woolworths self-checkout and put tomatoes on the scales, notice what happens. Instead of navigating an A-Z menu, "Tomatoes" is right there on the screen. AI vision sees something red, roughly round, in a bag, weighing about this much - probably tomatoes. Pick something different and it learns from that too. The customer doesn't think about any of this. They just notice the checkout is a bit less painful. Still painful (you're at the supermarket after all) but less. That's an outward cut.

Forward — where your product becomes AI

Forward is the narrowest domain but the one with the highest differentiation. It's about building things that wouldn't exist without AI.

Think about how Canva went from a design tool to an AI-powered design platform. The templates are still there, but now the AI generates layouts, suggests copy, removes backgrounds, resizes for every format. Take the AI out and Canva still works - but it's a much less capable product. That's the forward domain: AI becomes the product. This domain is mostly relevant to product and technology companies, but for those organisations, it's where the real competitive edge lives.

Each domain gets its own deep dive in this series. For now, the framework: inward, outward, forward. Three lenses for finding where the cuts are.

Where to start

Almost always start inward.

It's the safest domain because the blast radius is contained - you're changing how your team works, not how your customers experience you. It's the fastest because inward improvements tend to be tool adoption and process tweaks rather than system integrations. And it builds AI fluency. Your people develop intuition for where AI helps and where it doesn't. They learn to evaluate output, spot when something's off, understand what makes a good use case. That fluency is the foundation everything else sits on.

Move outward when you've built that confidence. When your team understands the tools well enough to trust them in customer-facing contexts. When you've got a sense of what good AI-augmented work looks like internally, and you're ready to extend that to your service layer.

Go forward only when there's genuine product-market fit for an AI-native offering. Not because a board slide says you should. Not because a competitor announced something. Because you've identified a real problem that AI uniquely solves, and you've got the organisational maturity to build and maintain it.

Here's the compounding argument in its simplest form: fifty small inward cuts, each saving a few minutes or improving quality by a small margin, will create more value than one ambitious forward bet that takes six months to ship and might not land. The small cuts are low risk, high certainty, and they compound. The big bet is high risk, low certainty, and binary.

How we work

This is how hps.gd works with clients. We don't walk in with a transformation roadmap and a six-figure platform recommendation. We find the cuts. We help teams identify where AI can make their work a bit better, a bit faster, a bit less crap. We stack those cuts. We watch them compound.

The AI promise I wrote about last time isn't a single moonshot. It's patience and persistence. The willingness to make a hundred small improvements instead of waiting for the one big one. The understanding that most organisations who win with AI won't be the ones who made the boldest bet - they'll be the ones who made the most incisions.

Over the coming weeks, I'll go deeper on each of the three domains: inward, outward, and forward. What the cuts look like in practice, where we've seen them work, and where they tend to go wrong. If you're figuring out where to start, or you've started and it's not compounding the way you expected, I'd love to hear about it.