There's a tempting assumption spreading through boardrooms and back offices alike: if AI can do it, AI should do it.
It's understandable. AI tools are impressive, they're improving fast, and every week brings a new story about what they can now automate. But here's the truth that gets lost in the hype — using AI in the wrong place in your workflows doesn't make you more efficient. It makes you slower, sloppier, and poorer.
At XRAY, we build workflow automations for a living. And one of the most valuable things we do for clients isn't adding AI — it's telling them where not to add it.
Here's the framework we use.
First, a distinction worth making clearly.
Tools like Zapier, Make, and n8n are programmatic automation tools. They follow rules. When X happens, Y occurs. Every single time. No surprises.
AI, on the other hand, is fundamentally non-deterministic. It's designed to handle messy, variable inputs and produce intelligent, variable outputs. That's its superpower — and its limitation.
Both can do similar-sounding things: send Slack messages, update spreadsheets, create folders. But how they do it — and when each is appropriate — makes an enormous difference to your business.
Think of programmatic automation like a vending machine. Press B4, get the Doritos. Every time, without hesitation, without interpretation. That predictability isn't a limitation — it's the whole point.
• A new deal closes in your CRM → a folder is created in Google Drive with that client's name. Always.
• A form is submitted → the contact gets a personalised confirmation email. Always.
No drift. No surprises. No extra cost per run.
AI, by contrast, is better suited to situations where there's no B4 to press — where the input is messy, the context shifts, and the right answer requires actual judgement.
• Got a 45-minute meeting transcript? AI can pull out the key decisions and next steps.
• Getting hundreds of customer messages a day? AI can read sentiment patterns and flag who's frustrated before they churn.
Try doing either of those with deterministic rules and you'll waste hours — or just fail entirely.
AI adds power. It also adds cost, inconsistency, and latency. That's worth being honest about.
The upside:
• Handles unstructured, unpredictable data with ease
• Can interpret, analyse, and generate new content — not just move data around
• Produces insights a rule-based system simply couldn't
The downside:
• Responses vary, even with tightly written prompts
• It can hallucinate — produce confident, wrong answers
• It's slower. Milliseconds for standard automation; seconds to minutes for AI
• It costs more per run, which adds up fast at scale
If you're running a high-volume, customer-facing workflow and every step includes an AI call, you'll feel it — in your bill and in your response times.
Before adding AI to any step in your workflow, ask these three questions:
1. Does this step need to produce the exact same result every time?
If yes — use automation, skip the AI. Creating folders with standard naming, routing data between apps, triggering notifications — these are programmatic tasks. Adding AI here is, as we like to say, like using a sledgehammer to hang a picture frame.
2. Would intelligent variability make this step meaningfully better?
Sometimes AI gives you options a rule-based system can't. Auto-tagging content is a good example — you could write rules based on keywords, but AI can surface tags you didn't even think to define. If the added intelligence genuinely improves the output, it's worth considering.
3. Does this step involve unstructured input or require a unique output?
If yes, AI isn't just helpful — it's probably the only realistic option. Analysing customer feedback and drafting a thoughtful reply. Summarising a dense report into three bullets. Generating personalised content at scale. These tasks are AI's natural habitat.
Work through these three questions for each step in your workflow. Some steps will call for AI. Most, frankly, won't — and that's fine. The goal isn't to use as much AI as possible. It's to use the right tool at the right moment.
AI is not a replacement for smart workflow design. It's a powerful component within it — one that works best when it's placed deliberately, not reflexively.
The businesses winning with automation right now aren't the ones who've AI'd everything. They're the ones who know exactly where to put it.
Ready to build workflows that actually work?
At XRAY, we audit your current processes, identify the highest-value automation opportunities, and build the systems that give your team their time back. Whether you want us to handle it end-to-end or want to learn as you go with one of our low-code engineers, we're here to help.
