“We should add AI” is a common ask with an uncommonly vague target. The businesses that get real value from it usually start from a specific, annoying manual process, not from a general mandate to adopt the technology.
Where it genuinely helps
- Sorting and routing incoming enquiries or support tickets based on content, rather than someone manually triaging every one
- Drafting first-pass responses to common questions, with a human reviewing before anything goes out
- Extracting structured data from unstructured sources — emails, PDFs, scanned forms — that would otherwise be typed in by hand
- Summarizing long documents or threads so a human can decide faster what needs attention
Where it tends to disappoint
Fully autonomous decision-making in anything customer-facing or high-stakes, without a human reviewing the output, tends to produce exactly the kind of confident-but-wrong result that erodes trust fast. AI that replaces judgment rather than supporting it is usually the source of the horror stories.
Start smaller than you think
The businesses that get genuine value tend to start with one well-defined, repetitive task, measure whether it actually saved time, and expand from there — rather than attempting a sweeping “AI transformation” that tries to change everything at once and ends up delivering nothing reliably.
The honest filter
If you can’t describe the specific manual task AI would replace, and roughly how much time it currently costs, it’s not ready to be scoped yet — it’s still an idea, not a project.