The same tool, three different problems
When someone asks «is AI any use to me?», the honest answer isn't yes or no: it's depends where it hurts. And that changes with company size in a fairly predictable way.
Small business: the bottleneck is you
In a business of one to ten people, the owner handles enquiries, quotes, invoices and also sells. Here AI doesn't replace anyone, because there's nobody to replace: it gives hours back. Answering repeated questions, booking appointments and logging contacts are tasks that can run on their own and that currently eat the morning.
The typical mistake at this level is wanting a complete system from the start. What pays off is one agent on one channel, the one that gets the most messages, and nothing else.
Mid-sized company: the problem is coordination
Between ten and a hundred people another problem appears: there is a team now, but information doesn't travel. Sales doesn't know what support promised; the lead who wrote on Monday falls between two people who each thought the other had it; nobody has the full monthly number.
Here AI pays off on two fronts: answering with a single voice, so the answer doesn't depend on who gave it, and logging everything so follow-up actually exists. The conversation stops living on somebody's phone.
Large company: consistency, volume and control
With several locations or business lines, the problem stops being time and becomes consistency. A hundred people answering well in ninety different ways is a brand problem. And the volume no longer allows anyone to read everything that comes in.
At this level what matters isn't that the system answers, but how it's governed: who can change what it says, how it connects to what already exists (the ERP, the CRM, the phone system), what gets audited, and what the system does the day it gets something wrong. The conversation looks more like an integration project than a subscription.
What doesn't change with size
- Someone has to own the process. A system with no owner goes stale in three months.
- AI doesn't fix a broken process. It speeds it up, errors included.
- Data rules. If your prices, hours and terms aren't written down somewhere, the agent has nowhere to get them from.
- Limits are declared up front. What the system doesn't do, and when it hands over to a person, is defined at the start, not after the first complaint.
How to tell which level you're at
Not by headcount, but by the question that keeps you up at night. If it's «there aren't enough hours in the day», you're at the first. If it's «I don't know what happened with that customer», the second. If it's «I don't know whether the other branch is saying the same thing», the third. Each one has a different first step, and confusing them is the most common way to spend a budget badly.
At Intellectum the conversation starts the same way whatever the size: half an hour to look at your actual processes and say where it makes sense to start, with no commitment. Sometimes the recommendation is not to automate yet.