Implementing AI in your business – beyond hype and prompt experimentation

For the past one and a half weeks, I’ve been setting up my OpenClaw installation. I call him Theo. He has no body, no face, no memory beyond the next session – and yet I notice something surprising: he forces me to think more clearly than many meetings ever have.

Not because he is brilliant. But because he simply cannot operate without structure. He has to think in structured ways. And that transfers to me.

Who I am – and why this matters

I work on positioning, design systems and digital communication for SMEs and self-employed professionals across Switzerland and beyond. Communication, brand, web – always with a focus on the system behind it.

AI does not interest me as a gimmick. It interests me as infrastructure – something that accompanies my workflow, supports it, and ideally learns how I actually work.

OpenClaw is an agent framework. It orchestrates models, tools, data sources and workflows – turning isolated prompts into structured processes. What I feel most strongly: for it to work well, I have to be precise. What does Theo need to act pragmatically, empathetically and purpose-driven? That is the same question I must ask myself.

A concrete example: For Theo and me to operate on the same level, I rebuilt my ClickUp workspace. I now work more clearly with tasks, deadlines and priorities – elements Theo can read, interpret and potentially execute on. We need the same language and the same information. Otherwise, even AI talks past you.

Context drift – why long chats fall apart

Most people know this phenomenon. You start solving a problem with a chatbot. At first, it is precise. Over time, it becomes vague. Responses contradict earlier answers. Eventually, you open a new chat and start from scratch.

That is not a bug. It is context drift.

Large language models can only process a limited context window. As irrelevant history accumulates, critical signals are diluted. The result: more noise, less quality – and more tokens consumed in the process.

And this is where things become economically relevant.

Tokens are not abstract. They cost money.

If you seriously consider deploying AI agents or automating marketing with AI, you cannot ignore token economics.

‍

‍

For context: 1 million tokens equal roughly 700,000–800,000 words. That sounds like a lot. It isn’t – once an agent calls multiple tools, analyses documents, stores intermediate results and reflects iteratively.

A single complex task can easily consume several hundred thousand tokens.

At 50 to 100 such tasks per day, this becomes strategically significant.

AI automation is resource design.

Databases as the underestimated foundation

What has become increasingly clear to me: the true leverage does not lie in the model itself. It lies in the structure that prepares it.

We have known this from automation for years. We connect platforms, sync data, build workflows – because manual aggregation is expensive and error-prone.

For humans. And for AI agents.

If I send Theo into chaos, he wastes tokens on orientation. If I provide a clean structure – defined tables, clear fields, structured knowledge – he can focus on execution.

A practical example: I connected ClickUp with MOCO (my agency time-tracking software). Instead of letting Theo query all tasks and manually map them to projects every day, a small background sync bridge maintains a local database – mapping lists to projects, marking active and archived items. Theo queries this local database without a single API call. He gets exactly what he needs – and nothing more.

That is not magic. It is architecture.

And the market signals confirm this. According to Gartner, more than 40% of agentic AI projects will be scrapped by 2027 – often not because of poor model performance, but due to unclear objectives, weak data quality and missing process structure.

MCPs – a new integration layer

More and more platforms are building structured access layers specifically for AI agents. These so-called Model Context Protocols (MCPs) – introduced by Anthropic in 2024 and increasingly adopted – allow agents to retrieve data, execute actions and navigate systems in a structured way instead of crawling blindly.

This represents a new meta-layer of digital infrastructure.

For SMEs in the DACH region, this means: If you document processes and structure your data today, you create a system that will be directly compatible with tomorrow’s AI agents.

Without major rebuilding.

Why this feels like an investment

I am investing time and resources into something that has not yet fully paid back. And still, it feels right.

Working on Theo is, at its core, an investment in documentation – of my thinking, my workflows, my decision logic. What I structure today can later be transferred to AI agents. Not just personality, but capability.

Gartner forecasts that by 2027, around 50% of business decisions will be augmented or automated by AI agents. Not strategy. Not judgment. But repeatable, data-driven execution – yes.

What I think clearly today can be scaled tomorrow.

Where AI truly helps in marketing and communication

The question I currently hear most from SMEs: “How can we meaningfully use AI in marketing?” “Can we automate our communication?”

The honest answer: Yes – but only where clarity already exists.

AI cannot fix unclear positioning. It cannot clean up chaotic processes. It amplifies what is already there.

That is why meaningful AI implementation does not start with buying tools. It starts with a simple question:

What is clear enough in our organisation to be automated?

That is the real design task.

What’s next?

I am still in the middle of this process. Theo evolves, and I learn as much about my own workflow as about AI architecture. It takes time and patience – but it is worth it.

If you are exploring AI and sense the potential, but feel unsure about direction: Which processes make sense? Where are resources wasted? How can AI be integrated meaningfully into communication and marketing?

Then a structured conversation makes sense.

Not to rebuild everything overnight. But to understand where you truly stand.

Feel free to reach out.

‍

Who We Are

Greg.design is a creative studio that works with a network of creative professionals to provide solutions for Swiss start-ups and SMEs.

We support future-proof rebranding, tackle creative challenges and promote the development of digital products – from strategy to modular style guides.

Sound interesting? Then let's talk!