Most retailers don’t fail at AI because they picked the wrong tool. They fail because nobody decided what kind of tool they actually needed before they started buying.
I’ve sat in enough of these planning sessions to know the pattern by now. A team gets excited about AI, and within an hour there’s a list of twenty things to try. A chatbot here. A new feature inside a tool they already use. Someone signs up for a content platform they saw mentioned in a LinkedIn post. A developer floats the idea of building something custom from scratch.
Six months later, none of it is finished. Nobody owns any of it. And half the team has quietly stopped believing AI is going to do much for the business at all.
The problem was never the tools. It was that nobody made a decision before they started collecting them. Rent, buy or build sounds like a technical question. It’s really a focus question, and most teams only ask it after they’ve already burned through months of budget and goodwill.
The Pattern
Foundations matter more than which tool you pick
A recent MACH Alliance survey of 600 enterprise decision-makers puts a number on this. Organisations with the right foundations in place are six times more likely to see measurable ROI from their AI investments than the ones without. Most businesses, though, are still running scattered pilots on tools that don’t talk to each other and can’t compound over time.
Source: MACH Alliance research, 600 enterprise decision-makers, Shoptalk Europe 2026
The businesses actually getting a return from AI aren’t running more experiments than everyone else. They’re finishing one thing before they start the next.
That’s the real gap. Not access to tools. Not budget. A decision made too late, if it gets made at all.
The Framework
Rent, buy or build: what each one actually means
These three words get used as if they’re interchangeable. They’re not, and mixing them up is exactly how teams end up committed to the wrong option before they’ve even solved the problem they started with.
| Rent | Buy | Build | |
| What it is | Use what you already pay for | Add a specialist tool | A custom system |
| Investment | Low, often already covered | Moderate, ongoing subscription | Substantial upfront, plus ongoing running costs |
| Setup time | Hours to days | Days to weeks | A few months |
| Ongoing upkeep | Low | Medium | High, ongoing |
| Needs a developer | No | No | Yes |
| Best for | Almost every team. Start here. | Teams who’ve hit the ceiling on the rented version | Where nothing on the market does the job and the payoff is clear |
| Examples | Claude, ChatGPT, Shopify Magic, Klaviyo AI, Gorgias Agent | Triple Whale, Rebuy, n8n, Zapier, Nosto | A custom agent built via API |
Under $20M in revenue? Rent first.
Exhaust the rented version before you spend a dollar on anything else. Then buy. Only build once you can write the problem in a single sentence and a developer agrees the build is worth it.
Why Most Initiatives Die
Why starting twenty things finishes none of them
A team walks out of a planning session with twenty AI initiatives and a lot of energy. Half a year later, none of it is working. It’s almost always the same four reasons.
01 · Nobody owns anything. When everything’s a priority, nothing is.
02 · Setup takes longer than the demo promised. The real configuration and data work behind every tool is significant. Multiply that by twenty parallel projects and none of it gets finished properly.
03 · Weak output kills trust. A half-configured tool produces mediocre results, and every AI initiative after that gets a harder internal sell.
04 · The next shiny tool always wins. A tool that’s been properly set up and run for sixty days looks nothing like one that’s just been switched on and left alone.
The Check
Three questions to ask before you touch anything
01 · What’s the single highest-value problem AI could solve for this business right now?
Not the most exciting option. Not the one with the flashiest demo. The one that would visibly move revenue, cost or time within the next 90 days. If you can’t write it in one sentence, it isn’t clear enough to solve yet.
02 · Could something you already pay for solve it within 30 days?
If yes, set it up properly, give it sixty days, and measure what happens before you touch anything else. If it’s a genuine no, or you’ve already hit the ceiling on what your current tools can do, that’s when a specialist tool earns a look. A custom build only earns consideration when nothing on the market does the job, the upside justifies real money, and someone is willing to own both the build and everything that comes after it.
03 · Who owns this, and what does finished actually look like?
Name one person. Write down the number that tells you whether it worked. Put a review date on the calendar today. The initiatives that quietly disappear are always the ones without a named owner and a measurable definition of done.
If you’re looking at a list of AI ideas and can’t tell which one deserves your time first, that’s the conversation worth having before you spend another dollar on tools. Get in touch and we’ll work out what’s worth renting, what’s worth buying, and what’s actually worth building.