The AI stack powering our fractional CMO practice — and why we chose every tool deliberately

Every tool we use is documented, vetted and governed. Here's an honest look at what's in the stack, what each tool actually does, and the principles that guided every decision.

This is that conversation for In Cahoots Co.

We operate a fractional CMO practice serving mid-sized Australian businesses. AI tools are central to how we work ... not as a novelty, and not to replace human judgement, but because the right tools, used responsibly, let a small consultancy like ours deliver at a standard that would otherwise require a much larger team. We've published a full AI Transparency Statement on our website, aligned to the AS ISO/IEC 42001:2023 international standard and the Australian Government's Voluntary AI Safety Standard. This insights piece is the plain-English version of what's actually in the stack and why.

Every AI-generated deliverable — whether a blog post, a line of code, or a synthetic voiceover — is reviewed, edited and approved by an In Cahoots Co before it reaches a client. Human judgement is not optional in our workflow. It is the workflow.

The stack, layer by layer

Infrastructure & Admin

The foundation everything runs on. Gmail, Drive, Docs, Sheets, and Meet for day-to-day operations. Gemini is our deep research engine, used for synthesising large volumes of information quickly, from competitive landscapes to industry trend analysis. NotebookLM sits alongside it as a document intelligence tool, allowing us to interrogate lengthy reports, transcripts, and reference materials without reading every word manually.



Why this: Data stored in Australian data centres (Sydney and Melbourne regions). Enterprise-grade security. Gemini's integration with Workspace means research flows directly into working documents without copy-paste friction.

Google AppSheet + Google Studio

AppSheet is where we build our internal custom apps — client onboarding workflows, project tracking, and operational dashboards that are built to our process rather than someone else's. Google Studio (Looker Studio) handles internal reporting visualisation, connecting data sources into dashboards that give us a real-time picture of client performance without exporting spreadsheets manually.

Why this: No-code app development means our operational tools evolve with us quickly. Staying within the Google ecosystem keeps data flows clean and within our approved privacy architecture.

Manus AI

Our most specialised tool. Manus handles custom research reporting, client-facing dashboard production, and Meta campaign management analysis. Where Gemini is our broad research engine, Manus is our precision instrument — used for tasks that require structured output, deep data synthesis, and the kind of detailed campaign reporting that would take a human analyst hours to produce manually.

Why this: Manus operates as a true agent — it can execute multi-step research and reporting tasks autonomously, then hand the output to a human for review and editorial judgement before it goes anywhere near a client. It can even assist with fault finding obscure issues with websites like Wordpress.

Claude (Anthropic)

Our primary writing tool for everything from long-form articles and blog content to strategic frameworks, client communications and document drafting. Claude is specifically chosen for writing work because of its ability to maintain nuance, follow detailed editorial briefs, and produce output that requires minimal structural editing. The articles in this blog series are Claude-assisted and human-edited ... not the other way around.

Why this: Claude's context window and instruction-following capability makes it the most reliable tool for extended writing projects where consistency of voice and argument matters. Anthropic's enterprise data commitments align with our privacy obligations to clients.

ElevenLabs (Professional Tier)

Used for synthetic audio generation — voiceovers, audio content, and voice-based deliverables for clients where applicable. ElevenLabs is an ISO 42001 certified provider, which means their AI management practices meet the same international standard we align to.

Critically: we never clone a voice without the express, biometric-verified consent of the owner. No exceptions.

Why this: ISO 42001 certification was the non-negotiable. In a space where voice cloning is being misused at scale, we only work with a provider whose governance commitments match ours.

The four principles that govern every tool decision

The stack didn't happen by accident. Every tool addition goes through the same four-question filter before it gets approved for use with client work.

1. Where does the data go?

Every tool in the stack has a documented data residency and retention policy. Client data stays within our approved architecture. We don't experiment with tools that have opaque data handling — especially in a post-Privacy Act reform environment.

2. Is there a human in the loop?

No AI output goes to a client without human review. Not because we don't trust the tools, but because the judgement call about whether something is right for a specific client, in a specific context is a human responsibility, not a machine one.

3. Can we explain it honestly?

We don't use tools we'd be uncomfortable disclosing. If we couldn't put a tool on our AI Transparency Statement and stand behind it publicly, it doesn't belong in the stack. This is our internal AI washing test.

4. Does it make us better, not just faster?

Speed is a byproduct, not the goal. Every tool in the stack is there because it improves the quality of thinking, research, or output ... not just because it reduces the time to produce something mediocre.

Why we publish this at all

The honest answer is that we think transparency about AI use in marketing is going to become a baseline client expectation — and sooner than most agencies expect. Even it our day-to-day conversations there is an expectation that these tools are being used, as most people in their own businesses are using in one way or another.

The businesses that get ahead of that expectation, that can point to documented governance and clear principles rather than vague reassurances, will be the ones clients trust with their most sensitive commercial work.

We also publish it because the conversation about AI in Australian marketing is still dominated by either hype or fear — and neither is useful to a business owner trying to make practical decisions. Showing exactly what a responsible AI-augmented marketing practice looks like is, we think, more useful than adding to the noise.

If you're a business owner who wants to understand how AI is being used in the marketing work being done on your behalf — by us or by anyone else — you're entitled to ask. And you should.

Our complete AI Transparency Statement — including data handling principles, human accountability commitments, and your rights as a client — is published at incahootsco.com.au/ai-transparency-statement and reviewed annually.