AI & Automation Consulting

If the phrase "artificial intelligence" brings robots and science fiction to mind, this page is for you. Below is what AI and automation actually mean for a working business, with no jargon and no sales pitch.

What these two words actually mean

Automation is teaching a computer to carry out a repetitive task the same way every time. There is nothing clever about it, and that is the point. If someone in your office copies figures out of an email into a spreadsheet forty times a day, automation is the thing that does that copying instead.

Artificial intelligence is software that recognises patterns and makes a judgement, rather than following a fixed set of instructions. The difference matters. Automation follows the rules you give it; AI makes a call. Automation can file every invoice into a folder. AI can read the invoice, work out which supplier it came from and what it is for, even in a layout it has never seen before.

You do not need to understand how either one works inside, any more than you need to understand a diesel engine to run a delivery company. What you need to know is which jobs they are genuinely good at, and which they are not.

What it looks like in a real business

Take a distribution company in Casablanca receiving around two hundred orders a day. They arrive by WhatsApp, by email and by phone. Someone retypes each one into the stock system. It takes about three hours a day, mistakes creep in when it is busy, and when that person is on holiday the orders pile up.

Automation handles the straightforward part: it reads the incoming orders and enters them. AI handles the messy part, which is most of it in practice. The customer who writes "the usual, but double this time" is understood, because the system recognises the customer, looks up what "the usual" has meant for them over the past year, and flags the order for a human to confirm rather than guessing silently.

The result is not three hours down to zero. It is three hours down to about twenty minutes of checking. And the person who used to do the retyping is not out of a job; they stop retyping and start chasing the late deliveries that were losing the company money. That is the honest version of what this technology does: it removes the dull part of a role, not usually the role.

How we would work with you

We start with your problems, not with the technology. The first conversation is about where the time goes, where the mistakes happen, and what keeps people at their desks after six. No software is mentioned.

From there we look for tasks with a particular shape: repetitive, high in volume, mostly rule-based, and where a mistake is annoying rather than catastrophic. Those are the safe places to start. We deliberately avoid beginning with anything where an error would cost you a customer or breach a regulation.

Then we test small. One process, a few weeks, with the before-and-after measured so you can see whether it actually paid. If it did not, we say so and we stop. Only once something has proven itself do we widen it out.

What changes for you

• Your staff spend less of the day on repetitive typing and copying
• Customers get an answer in minutes rather than the next morning
• Fewer errors, because the same step is performed the same way every time
• Information that used to live in one person's inbox becomes findable
• Decisions get made on numbers you can trust, not on a best guess

Words you will hear, decoded

Generative AI

Software that writes or drafts things, of which ChatGPT is the best-known example. Useful for a first draft; not to be trusted as a final answer.

Machine learning

How software improves by studying examples instead of being given rules. Show it ten thousand invoices and it learns what an invoice looks like.

Workflow automation

A chain of steps that runs by itself once something triggers it, such as an order arriving.

Copilot or assistant

AI built to help a person do their job faster, rather than to do the job instead of them.

Model

The trained part that does the recognising. When someone says "the model got it wrong", they mean this.

Training data

The examples the software learned from. Poor examples produce poor results, which is why this matters more than most vendors admit.