Blog ·
What AI can really do for your business and what it can't
Artificial intelligence in a company can already answer customers from your own material, write and translate product descriptions, sort emails, evaluate sales, track competitor prices and plan deliveries. It cannot make decisions for you, and without source material it makes things up. This is a practical overview for a business owner: what pays off, what not to entrust to it (yet) and how to start.
AI is not a robot that takes over your company
First, let's be clear about what we are talking about. The artificial intelligence we use in companies today is essentially a very capable assistant that can read, write, compare and sort. It has read an enormous amount of text, which is why it understands a customer's question, can write a product description or pull the due date out of an invoice.
It knows nothing about your company, however, until you tell it. And it has no judgement of its own. Give it poor source material and you get confidently written nonsense. In our experience, when AI fails in a company it is rarely the technology – it is that nobody gave it proper source material and rules.
Once it has the material and the rules, it handles things that used to take hours of human work every single day.
What AI can handle in a business today
1. Answer customers – even at ten in the evening
The question we hear most often is whether it can answer your customers. It can, but only from your own material: the price list, product descriptions, terms and conditions. A customer asks whether you deliver to Austria and gets an answer immediately, without waiting until you arrive at work in the morning.
One rule matters: the assistant answers only from the material, and when it cannot find the answer there, it hands the conversation to a person. During the first weeks we check its answers and fine-tune the material so that it does not make things up. Set up this way, the assistant handles a large share of the repetitive questions and leaves you the ones that genuinely need a decision.
It happens that a company wants to launch the assistant straight away, before it has a price list or terms and conditions. Without source material it would answer in generalities and wouldn't help the customer. That is why we always start by sitting down with the questions your customers ask most often and writing out the answers. Only then does the assistant have something to draw on.
2. Write hundreds of product descriptions and translate them
An online shop with thousands of products and descriptions copied from the supplier is a common state of affairs. AI writes an original description for every product in your brand's tone, adds keywords for search engines and translates it into other languages. By hand this is months of work; AI prepares the drafts in a fraction of the time and leaves you the checking.
Human review before publishing is a condition. Not because AI writes badly, but because you know things about your products that are not in the material. And whatever is missing from the material, AI may guess wrongly. So check dimensions, materials and compatibility before publishing.
3. Sort emails and social media messages
Every day a company receives enquiries, complaints, invoices, offers and spam – all to one address. AI reads and files every message: urgent ones go to your phone, routine ones get a draft reply, and spam is set aside. Imagine opening your inbox in the morning and, instead of one big pile, seeing today's complaints at the top, enquiries with a prepared draft reply below them, and the rest sorted where it belongs. In the same way it can watch the comments under your posts and alert you when a reaction is needed.
4. Evaluate sales and say where the room is
You have years of data in your accounting and your online shop, but nobody has time to read it. AI goes through it and tells you in plain language: which products are stopping selling, where the margin is low, what sells seasonally and where it would pay to aim your advertising. Instead of a table with thirty columns, you get a few sentences an owner understands.
5. Track competitor prices
It can be done by hand, but only for a few products and only occasionally. With hundreds of items nobody keeps up. AI goes through your competitors' shops every day, finds the same products even when the competitor names them differently, and matches them to yours. In the morning you have an overview: where you are more expensive, where cheaper and where you have room to raise a price.
6. Plan deliveries for companies with their own vehicles
If you deliver goods with your own vehicles, you know the routine: every morning someone assembles orders, vehicles and drivers into routes. AI does it automatically – it builds routes for each vehicle from the orders and, while doing so, also reads what is written next to each order: when the customer is open, where a large vehicle can get to, who needs a call in advance. The driver gets a route that takes all of that into account, and the dispatcher has a free morning.
What not to entrust to AI (yet)
So that it doesn't sound as if AI solves everything, here is the list of things we would not hand over to it today:
- Decisions about money without oversight. AI can propose a new price or an order to a supplier. A person should approve it.
- Legal and accounting output. Leave a contract or a tax return to a professional. AI prepares the material for you; responsibility for the result stays with the expert.
- Anything without source material. Ask the assistant about something that is not in its material and it should answer "I don't know, I'll connect you with a colleague". If that is not set up, it will make things up – and sound convincing.
- Sensitive data without rules. Customers' personal data, contracts, health information. AI should see them only where necessary, and it must be clear where they are processed. It can be set up, but it should be done by someone who knows what they are doing – it is not an afternoon's job.
What it costs and how to tell whether it pays off
We get this question at every meeting. There are two routes, each with its pros and cons.
The first is AI from the large providers. The company buys nothing, uses it over the internet and pays for actual use – much like a phone: the more messages AI handles or texts it writes, the higher the fee. For an ordinary company this is a monthly running cost, not the cost of another employee, and it can be estimated in advance from how much work AI will be given. The advantage is a quick start with no investment in equipment.
The second is your own in-house AI. It runs on a computer in your company, the data never leaves, and there is no usage fee. You do, however, pay for powerful hardware and its operation, and such an AI is usually weaker than the one from the large providers. It suits companies that work with genuinely sensitive data or have a very large volume of work.
With both routes the larger investment comes at the start: connecting AI to your systems, giving it source material and rules, and testing it. Which route suits you depends on the data you work with and how much work AI is to do – and that is exactly what we can calculate in advance, before the first pilot.
You don't have to work it out yourself
You don't need to know which process in your company is ready for AI, or how AI connects to your stock or online shop. Just describe what slows your company down the most – where things get retyped, where people wait, where the same question gets answered over and over. We propose what AI can take over, prepare the material and the rules, deploy a pilot and measure how much time it saves. Only then do you decide whether to go further.
See what we do in automation and AI, or write to us – the first consultation is free of obligation.
