Every week, I talk to founders and operators who start with the same sentence: “I know we should be doing something with AI, but I don’t know what… or if it’s just hype.”
I run Kraken Automate, a solo automation consultancy built on n8n, and my answer is always the same: ignore the hype and follow the workflows.
Why I ignore AI headlines (mostly)
In August, The New York Times ran a piece titled “A.I. Hype Is Running Into Reality.” That mirrors what I see on the ground: grand promises, thin results.
Investors are starting to say the same thing. TheStreet recently covered a trade “for cash flow, not AI hype.” The market is getting tired of big AI stories that don’t turn into reliable revenue or efficiency.
Inside companies, the pattern has a name now: “AI washing.” BetterUp called out how over-marketing AI creates “workslop” and erodes trust at work. When every product is suddenly “AI-powered,” people stop believing any of it.
When governments and industry magazines are calling for “infrastructure resilience, not AI hype” and “industrial AI without the hype,” they’re telling you something simple: fundamentals matter again.
My starting point: no AI until the workflow is clear
When I come into a business, I don’t start with models, agents, or big ideas.
I start with three questions:
- Where does work get stuck? (Approvals, handoffs, data entry, customer replies.)
- What do people repeat every day or every week? (Copy-paste, file renaming, status updates, manual reports.)
- Where do errors actually cost you money or reputation? (Billing, compliance, customer promises.)
If we can’t answer those concretely, we are not ready for AI. We’re not even ready for automation. We’re still at the whiteboard stage.
“AI without the hype” for me means: no generic solutions, no vague “copilots,” no slideware. Just specific workflows and clear before/after metrics.
Where AI genuinely helps right now
Most of my real wins come from combining plain automation with a thin layer of AI in exactly one or two steps.
Here are the patterns that keep paying off:
- Inbox triage – n8n pulls emails from a shared inbox, an AI step classifies them (billing, support, sales, spam), and routes them. Humans still respond, but they stop playing traffic cop.
- Summarising chaos – long customer threads, vendor PDFs, internal notes. AI summarises and extracts the 5–10 fields we actually care about. n8n puts that into a CRM, ticket, or database.
- Data cleanup – AI normalises messy free text (company names, subject lines, product descriptions) so automation can match, merge, and route reliably.
- Draft, don’t decide – AI suggests a first draft for a support reply, status update, or knowledge base article. A human edits and approves. n8n handles posting, notifying, and logging.
Notice what’s missing: no fully autonomous agents running your company. No “magic” bots that replace teams. Just mundane, well-bounded tasks where the downside is low and the upside is compounding.
My simple rule for AI in workflows
When I design an n8n workflow, I use this rule:
If AI makes a decision that could hurt cash flow, customers, or compliance, a human must be in the loop.
That’s it. This is especially important in areas like anti-money-laundering (AML), where “AI without the hype” has become a serious conversation. In those environments, AI can help surface better alerts or cluster suspicious patterns, but it cannot be the only line of defence. Someone accountable has to see and confirm.
I apply the same logic for small teams:
- Let AI sort, summarise, suggest.
- Let humans approve, escalate, own outcomes.
n8n is the glue in the middle: moving data, enforcing checks, logging decisions, and giving you a timeline you can actually audit later.
What “AI without the hype” looks like in practice
Here’s how I usually structure a first project with a client:
- Week 1: map one painful workflow – We pick a single, boring process: onboarding a client, closing a ticket, sending invoices. We list every step and every tool.
- Week 2: automate the obvious, no AI yet – n8n handles triggers, data movement, and notifications. We stabilise the plumbing first.
- Week 3: add one AI step – We insert AI where humans are clearly wasting time reading, sorting, or rephrasing. We measure error rates and time saved.
- Week 4: decide to scale or stop – If the numbers work, we scale that pattern to similar workflows. If not, we turn it off. No sunk-cost fallacy, no “but it’s AI so we must keep it.”
This is slower than buying a flashy AI tool and rolling it out to everyone in a week. It’s also the only way I’ve seen leaders trust the results.
How to audit your own AI hype
If you’re already paying for AI tools or building internal projects, ask yourself:
- Can I point to at least one process where cycle time is measurably shorter? If not, it’s probably still in the hype zone.
- Do my team members trust the outputs? Or are they quietly double-checking everything because of earlier misfires?
- Could I explain this workflow to a new hire in one page? If not, it’s too complex to be safe and reliable.
Western governments are talking about “infrastructure resilience, not AI hype.” I think the same principle applies inside your business: resilience is boring. It’s also where the real value is.
My own bias is simple: I’d rather help you remove five hours of weekly grunt work from ten people than chase one big “transformational” AI project that never lands.
If you want AI without the hype in your business, start where the work is already happening, not where the headlines are pointing.