The AI Growth Stack: One Tool Per Decision

I run my consultancy and two fractional growth mandates on three artificial intelligence (AI) tools. Most marketers I meet pay for more than that to do one job. The difference between us has little to do with taste in software. We mostly buy the same things anyway. The difference is that each of my tools owns a decision, and most of theirs are still auditioning.

That is the argument of this piece. An AI growth stack is a set of ownership decisions wearing a software subscription. Map the layers of your growth system first. Appoint one tool per decision. Then apply one test: if two tools in your stack could answer the same question, nobody has decided which one owns it. That redundancy is the most expensive line in your budget, and it never appears on an invoice.

Here is the logic, the evidence, and the exact stack I run.

1. Everyone Is Buying the Same Stack

Singapore’s Digital Economy Report 2025 found that 84% of AI-using firms here rely mostly on off-the-shelf generative AI tools. Only 44% run anything customised. Globally, 28% of enterprises now run more than ten AI applications, and 66% plan to add more within the year.

Read those numbers together and the strategy question changes. When everyone owns the same tools, the tools stop being an edge. Two competitors with identical stacks still get wildly different results. The gap comes from what each company decided the tools should own.

💡 Key Takeaway: So stop asking which AI tool to buy. Ask which decision you are handing over, and to what.

2. A Second Tool in the Same Layer Means Nobody Decided

The sprawl data describes companies with nobody at the wheel. Only 35% of enterprise leaders say their AI tools go through approval channels, 70% have never integrated their tools beyond basic connections, and 31% discover rogue AI tools every month. Meanwhile, the money compounds. Average AI-native spend reached US$1.2 million per organisation, up 108% year on year, and 78% of IT leaders report unexpected charges from consumption pricing.

Every duplicated tool is a decision nobody made. Two writing assistants mean nobody decided who drafts. Two analytics tools mean nobody decided which number is true. The subscription fee is the small cost. The real cost is that every undecided layer keeps a human as the router, copying context between tools that never talk to each other.

💡 Key Takeaway: For every tool, complete this sentence: “this tool alone owns X.” Two tools sharing one sentence is the red flag. That decision has no owner.

3. Klarna Made the Decision After the Deployment

In February 2024, Klarna announced its AI assistant was doing the work of 700 customer service agents. By May 2025, the company was rehiring humans. The company’s chief executive put it plainly: “We went too far… We focused too much on cost. The result was lower quality.”

Note what failed. The tool performed exactly as sold. It handled millions of conversations at speed. What arrived late was the ownership decision: which conversations should a machine own, and which must a human keep? Klarna made that call after the rollout, in public, at the cost of its service reputation.

At founder scale, the same mistake looks smaller and costs relatively more. You point a chatbot at the whole funnel, then discover the demo requests deserved a human reply within the hour. The tool did its job. Nobody had drawn the line it was allowed to cross.

4. Map the Layers, Then Appoint One Owner Per Decision

Here is the stack I run across my firm and two client mandates. One person, all functions.

• Orchestration. Claude Code is the brain. It sits over a plain-text workspace holding my positioning, my playbooks, and a persona skill for each function I run. Context lives in files, so nothing starts from zero.
• Operations. The brain connects to Notion for sprints and tasks, Clockify for time, Google Calendar for the schedule, and Pocket for meeting capture. Four systems, four different decisions, zero overlap.
• Measurement. PostHog, Google Analytics 4 (GA4) and the ad platforms feed performance back through the same brain for reporting and recommendations.
• Build. Supabase and GitHub, where the firm’s product work ships.
• Distribution. Kit and Resend, linked by AI for lead capture and nurture.
• Graphics. ChatGPT renders carousels and images. That is its whole job, and it owns nothing else.

And one hole, published on purpose: content-to-call attribution. I can see which posts run and which conversations start. I cannot yet prove which post started which conversation. That layer has no owner. An honest map shows the empty room.

The sceptical reading of this list is fair: my stack will not fit your company. Agreed, and it does not need to. The tool names are the most replaceable part of the map. Swap Notion for Linear, or PostHog for Amplitude, and the logic holds. What transfers is the layer map and the ownership test. And the capacity case stands on its own: one person runs strategy, delivery, content and product across three businesses, because every layer has a single owner and the brain holds the context.

💡 Key Takeaway: Build the map before the shortlist. Layers first, decisions second, tools last.

Final Thoughts: Decide First, Subscribe Second

The AI stack conversation keeps starting at the wrong end, with the shopping list. Start with the layers of your growth system. Appoint one owner per decision. Kill the duplicates. Then name the layers where you have decided nothing yet, because pretending they are covered costs more than admitting the gap.

If you want a second pair of eyes on your own map, that is a conversation I have every week. Book a discovery call, or connect with me on LinkedIn.

---

A note before you close this tab. If your top of funnel has been thinning for reasons nobody can quite explain, the cause may not sit in your strategy. It may sit in your reporting defaults. That is fixable, and the fix starts with naming what your model cannot see.

Mervyn Chua is a growth-transformation consultant helping founders and CEOs build the strategic clarity and systems to grow in an AI-first world. If this raises questions worth exploring for your brand, let’s talk.