Artificial intelligence has entered businesses very quickly in recent years, both through tools used directly by people, such as ChatGPT, Gemini, Copilot and many others, and through AI features that are now being integrated directly into management software, CRM systems, email platforms and many other applications.
Because of this rapid adoption, one of the questions companies ask most frequently today is which artificial intelligence they should use and which tools they can introduce to improve the way they work.
In reality, this is a valid question, but it is being asked at the wrong stage, because before deciding which technology to use it is necessary to understand which problem needs to be solved, how the work is currently carried out and, above all, which processes can actually be improved.
AI should therefore not be the starting point of a project, but one of the tools that can be used after analysing the company and identifying its real needs.
THE PROBLEM BEFORE THE TECHNOLOGY
When analysing a company, before talking about artificial intelligence it is necessary to understand how that company actually works.
Even within small organisations we can find dozens of different processes, some very simple and others more complex, involving people, software, documents, emails, phone calls, management systems, spreadsheets and many other sources of information.
It is therefore necessary to understand where time is being lost, which operations are being repeated, which data is manually copied from one system to another, where errors occur and which activities still depend entirely on the intervention of a person.
To give a very simple example, a company may receive a request from a customer through a form on its website; someone reads the request, manually enters the data into the CRM, creates a folder for the customer, sends an email to a colleague and later enters some of the same data into the management system.
All the tools being used are digital, but the process itself remains almost completely manual.
In a situation like this, the first problem to solve is not which AI to use, but why the information has to pass manually through so many different systems and whether those systems can communicate with one another.
Only after this analysis can we decide where to use a simple automation, where to create an integration and where artificial intelligence can actually add something useful.
- Problem
- Process
- Information
- Integration
- Automation
- AI
DIGITALISATION AND AUTOMATION ARE TWO DIFFERENT THINGS
Most companies today already use many digital tools.
We find accounting software, CRM systems, ERP platforms, email, cloud services, document management platforms, order management software, ticketing systems, vertical management applications and often a large number of Excel spreadsheets used to manage specific activities.
The problem is that very often all these tools do not communicate with each other.
We therefore have a digitalised company, because it uses IT systems in practically every department, while at the same time people are still manually moving information from one piece of software to another.
And it is precisely within these steps that a large part of the work that can be improved is often found.
A piece of data entered into a CRM could automatically be transferred to the management system, an order could generate a series of activities without someone having to start them one by one, and a request received through a website could automatically create a case and notify the person who needs to handle it.
All of this can also be done without using artificial intelligence.
In fact, one of the mistakes that can be made today is to consider AI and automation as if they were the same thing, when in reality they are different technologies that can work together but perform different functions.
BUSINESS INFORMATION
Another fundamental element is understanding where the company’s information is located.
Very often a company already has all the data it needs, but that data is distributed across different systems.
Commercial information may be stored in the CRM, administrative information in the management system, documents in a cloud service and communications in email, while some company procedures may not even be written down anywhere and may only be known by the people who carry out that work every day.
This creates an important problem, because having a large amount of data does not automatically mean being able to use it.
Before building a system based on artificial intelligence, it is therefore necessary to understand:
- where the information is located;
- who can access that information;
- who updates it;
- how reliable it is;
- whether duplicate information exists;
- which systems need to communicate with each other;
- which data can be used automatically.
Once this structure is clear, AI can become much more interesting, because it no longer works only on a request manually entered by a user, but can be connected to the company’s real information and processes.
NOT EVERYTHING NEEDS AI
Artificial intelligence is a very powerful tool, but this does not mean that it should be used for every activity.
If a process requires the same operation to be performed every time a specific condition occurs, we probably do not need AI, but simply a rule or an automation.
If two pieces of software need to exchange data, we probably need an integration.
If registering a new customer must automatically create a folder, a CRM record and a series of tasks, we need a workflow.
Artificial intelligence becomes particularly useful when the system needs to work with information that cannot simply be managed through predefined rules.
For example, an email received from a customer may contain a request expressed in hundreds of different ways. A standard automation may have difficulty understanding the content of that request, while an AI system can interpret the text, classify it, extract the necessary information and then pass that data to the automated process.
In this case we therefore have artificial intelligence and automation working together, each being used for what it does best.
WHERE AI BECOMES REALLY USEFUL
There are many business activities where artificial intelligence can provide a real advantage, especially when large amounts of unstructured information need to be managed.
Documents, emails, customer requests, images, reports, contracts and communications contain a large amount of information that normally has to be read and interpreted by a person.
AI can be used to classify this information, extract data, summarise documents, search through content, prepare responses, transform information from one format into another and provide users with systems that can be queried using natural language.
But there is more.
When this system is connected to the company’s software, it can also become part of a more complex process.
Take a support request received by email as an example.
The AI system can read the content, understand which customer it relates to, identify the problem, assign a category and prepare a possible response; at the same time, an automation can open the ticket, retrieve the customer’s information from the CRM, assign the request to the correct technician and include all the necessary documentation.
At that point the technician no longer starts from zero, but finds an organised case containing the information needed to work on it.
This is very different from simply using a chatbot.
FROM A SINGLE TOOL TO THE BUSINESS SYSTEM
The first adoption of artificial intelligence within companies mainly happened through tools used directly by people.
The user opens ChatGPT or another system, enters a request, receives a response and then manually uses the result.
This type of use can already save a considerable amount of time, especially for activities related to writing, research, analysis and information processing.
The next step, however, is to connect these systems directly to the company’s infrastructure.
An AI system can receive information from the CRM, consult authorised documentation, process data coming from a management system, use APIs provided by other services and finally pass the result to another part of the process.
In this way we no longer simply have a person using artificial intelligence, but artificial intelligence becoming a component of the business system itself.
And this is exactly where systems, integrations, automation and AI start working together.
BEFORE AUTOMATING, YOU NEED TO UNDERSTAND THE PROCESS
There is another problem that needs to be considered.
Not everything that is currently done inside a company necessarily needs to continue being done in the same way.
Many processes were created years ago, when the systems being used were different, and were later modified over time by adding new steps, new checks and new tools.
It is therefore possible to find duplicated activities, approvals that are no longer necessary, data being entered several times or procedures that exist only because two software systems have never been connected.
If we take a process like this and automate it completely, we simply obtain an inefficient process that is executed more quickly.
For this reason, before thinking about automation it is necessary to analyse the process and, where possible, simplify it.
The process should therefore begin by observing the work as it is currently carried out, identifying the problems, removing unnecessary steps, connecting systems where needed and only then automating what can actually be automated.
Artificial intelligence is introduced into this process at the points where it is genuinely capable of improving the result.
THE RESULT OF AN AI PROJECT
The result of a project should not simply be the ability to say that the company uses artificial intelligence.
The result needs to be measurable in the real work being carried out.
If a particular activity previously required two hours and later requires thirty minutes, we have achieved a result.
If errors caused by manual data entry are reduced, we have achieved a result.
If a customer receives a response more quickly, if an employee can immediately find the information they need or if the company can manage a greater volume of work without increasing manual activities by the same amount, then the technology has created real value.
It is therefore not important how many AI tools are being used, but what effect they have on the way the company operates.
In some cases, a small automation can produce a much greater result than introducing a complex AI system, simply because it has been applied in the correct place.
PEOPLE AND TECHNOLOGY
The introduction of artificial intelligence also changes the relationship between people and IT systems.
Activities that until recently could only be carried out by a person can now be managed, at least partially, by a system.
This does not necessarily mean removing the person from the process, but understanding which activities should continue to be handled by the operator and which can instead be prepared or carried out automatically.
A system can collect information, organise it and prepare a proposal, while a person keeps the final decision.
It can check hundreds of documents and highlight those that require attention.
It can prepare a response based on the available information, leaving the operator in control before it is sent.
In this way, the person’s work gradually moves away from repetitive operations and towards handling exceptions, control and those activities where experience and decision-making ability are genuinely required.
WHERE TO START
When a company decides to begin using artificial intelligence, therefore, the first thing to do should not be to open a list of products and choose the one that looks most interesting.
The work itself needs to be analysed first.
It is necessary to understand the main processes, where information is produced and stored, which systems are being used, which operations are still carried out manually and, above all, which problems need to be solved.
After this analysis it becomes possible to understand whether the company needs to modify a process, integrate two systems, create an automation or actually use artificial intelligence.
Very often the final solution will consist of all these technologies working together.
AI will in fact become increasingly similar to a normal component of IT systems and will progressively be integrated into the software we use every day.
The real difference will therefore not simply be knowing or using the latest available tool, but being able to understand where to place it within a business structure and which result it needs to produce.
This is why AI in business doesn’t start with AI.
It starts with the company, its processes, the people who work within it, the systems it uses and the problems that need to be solved.
Only then comes the technology.