Skip to main contentSkip to footer
Logo IT4LIA AI factory
Resources

AI idea, or still looking for where to start? How to recognise which stage you’re at

The AI landscape in Italy

In 2025, only 8% of Italian SMEs launched at least one artificial intelligence project, compared with 71% of large enterprises. Meanwhile, the Italian AI market grew by 50%, reaching €1.8 billion, and 47% of workers say they use AI tools at work. Generative AI systems now produce code with ease, and tools, platforms and databases are easy to integrate. Together, these have caused an explosion of startups and small product companies, each promising to brilliantly solve a problem for every market niche. The risk of becoming the first unwitting testers of an immature solution is, however, a real one.

For those running a business or heading a company function, navigating this complexity means being able to answer some very concrete questions. Where can AI be genuinely useful? What data is needed, and what data is available? What skills and investment does it require? What should be considered when deciding whether to build or buy? And how do you choose a solution that is reliable and sustainable over time?

Experimenting with a tool can be simple. Embedding it in a process, however, means tackling technological, organisational and economic issues at the same time. Running models locally involves assessing infrastructure, computing capacity and maintenance. Using cloud services requires clarity on how data is handled, who can access it and what guarantees the provider offers. Choosing between established products, startups and in-house development means evaluating their maturity, integration, costs and service continuity. Even a promising prototype built by a team member needs a path to become a solution the organisation can manage and use.

AI consulting for SMEs: how to work out the best path

With these figures in hand, it is easy to see why the gap between SMEs and large enterprises is so wide. In an SME, managing this complexity is often a significant barrier to entry, one that training can reduce but cannot fully remove.

In this context, we can distinguish two main scenarios.

“I see the potential of AI, but I don’t know where to start.”
Business owners, managers and heads of function may recognise the value of AI without yet having identified the most useful applications for their own organisation. The available information and product demos are not enough to understand which opportunities are actually viable in their specific process or production context.

“I’ve identified an opportunity, but I don’t know how to make it happen.”
Process, innovation and IT managers may already have a defined problem, a solution idea or an initial experiment. What they need is to sharpen objectives and requirements, check the available data, compare technology alternatives and define resources, timelines and success criteria.

Both needs can arise in companies of any size and can coexist within the same organisation: having already adopted AI in one area does not mean knowing how to apply it to the next. These two needs call for different paths, with dedicated approaches and support methods. Confusing them is the most common way to waste time. Those who already have a clear idea risk starting over with an analysis that is too broad. Those who don’t risk jumping straight into the technical planning of a project they haven’t actually chosen yet.

Where should I start?

You understand that AI has potential, but you struggle to see concretely how it could help you. You are probably not yet starting from a specific problem, but are focused on AI technology itself. You will recognise this situation if you find it hard to answer these questions:

  • Can you name a single business process, among the many possible ones, that you would like to improve with AI?
  • Do you have an idea of which data in your organisation is already available and usable?
  • Do you know whether your in-house technical skills are sufficient to get started?
  • Have you already assessed whether it makes more sense to develop a solution, adapt an existing one or look for a technology partner?

If your answer to these questions is “I don’t know” or “it depends”, the issue is one of orientation. What you need first is an impartial snapshot of where the organisation stands in terms of data, infrastructure and skills. From there, you can build a list of realistic opportunities based on actual internal pain points and ranked by impact and feasibility.

This is exactly the purpose of the AI Assessment: an audit that evaluates the quality and availability of data, the infrastructure in place and internal skills, in order to identify the use cases that best fit the organisation’s real needs.

The goal is not to produce a generic list of possible AI applications, but to identify the most solid priorities, weighing expected impact against ease of implementation.

This is the path that enabled a logistics company to move from a general intention, “we want to optimise operations with AI”, to a specific use case: improving fleet management.

How do I make the opportunity happen?

The situation is different when the organisation has already identified an idea or an area of intervention but does not yet know how to turn it into a project.

You will recognise this situation if you are more comfortable with questions like these:

  • Do you already know which process you want to improve, but want to understand whether the data you have is enough?
  • Do you have in mind the result you want to achieve, even without the technical steps to get there?
  • Do you want to understand whether you need a model built from scratch, an adapted existing model, or something else entirely?

In this case, the problem is not working out what to do, but how to structure it.

The idea needs to be turned into measurable objectives, technical requirements, work phases and a realistic estimate of the time and resources required.

This is the role of AI Project Definition: support that starts from an idea that has already been identified (for example, “we want to automate quality control”) and translates it into an operational plan with clear objectives, technical requirements and a roadmap with defined phases and timelines.

This is the path that enabled a manufacturing company to turn the idea of predictive quality control into a project ready for implementation, with a compliant architecture and the peace of mind of knowing its data is secure.

Why the distinction matters

The two situations call for different kinds of work. The first needs breadth: looking across the whole organisation to find the right priority, identifying synergies between potential use cases to choose the highest-impact route. The second needs depth: going into the technical detail of a single idea that has already been chosen, managing the architectural and technological complexity.

Starting from the wrong point has a concrete cost. A general audit on an idea that is already clear risks wasting weeks re-discussing decisions that have already been made. Technical planning on an idea that is still vague risks producing a detailed plan for the wrong use case.

Distinguishing between orientation and project definition therefore helps make better use of time, resources and skills.

Dedicated services help those who lack the time, energy or internal resources to speed up the journey without losing control of it.

Where to go from here

You don’t need to answer these questions with absolute certainty to know where to begin: an initial sense of direction is enough. If the main uncertainty is about what to do, the first step is an assessment of the organisation’s AI maturity. If instead the uncertainty concerns how to deliver an idea that is already clear, the first step is the technical definition of the project.

In both cases, getting started does not require an immediate large-scale commitment. IT4LIA offers consulting expertise, computing infrastructure and a catalogue of services designed to support organisations one step at a time, from the initial assessment through to experimentation. You don’t need to have all the answers to start a first conversation and find out which of the two paths is right for you.

How to access IT4LIA

The path to getting oriented is simple and doesn’t require having every detail worked out: just book a call, explain where you are, and let the IT4LIA team point you in the right direction.

How to access IT4LIA: Start your AI Adoption journey now

Book a call with our AI experts now to find out which service is right for your organisation.

Veronica Brizzi, AI4I

AI idea, or still looking for where to start? How to recognise which stage you’re at
Resources

AI idea, or still looking for where to start? How to recognise which stage you’re at

AI idea, or still looking for where to start? How to recognise which stage you’re at

The AI landscape in Italy

In 2025, only 8% of Italian SMEs launched at least one artificial intelligence project, compared with 71% of large enterprises. Meanwhile, the Italian AI market grew by 50%, reaching €1.8 billion, and 47% of workers say they use AI tools at work. Generative AI systems now produce code with ease, and tools, platforms and databases are easy to integrate. Together, these have caused an explosion of startups and small product companies, each promising to brilliantly solve a problem for every market niche. The risk of becoming the first unwitting testers of an immature solution is, however, a real one.

For those running a business or heading a company function, navigating this complexity means being able to answer some very concrete questions. Where can AI be genuinely useful? What data is needed, and what data is available? What skills and investment does it require? What should be considered when deciding whether to build or buy? And how do you choose a solution that is reliable and sustainable over time?

Experimenting with a tool can be simple. Embedding it in a process, however, means tackling technological, organisational and economic issues at the same time. Running models locally involves assessing infrastructure, computing capacity and maintenance. Using cloud services requires clarity on how data is handled, who can access it and what guarantees the provider offers. Choosing between established products, startups and in-house development means evaluating their maturity, integration, costs and service continuity. Even a promising prototype built by a team member needs a path to become a solution the organisation can manage and use.

AI consulting for SMEs: how to work out the best path

With these figures in hand, it is easy to see why the gap between SMEs and large enterprises is so wide. In an SME, managing this complexity is often a significant barrier to entry, one that training can reduce but cannot fully remove.

In this context, we can distinguish two main scenarios.

“I see the potential of AI, but I don’t know where to start.”
Business owners, managers and heads of function may recognise the value of AI without yet having identified the most useful applications for their own organisation. The available information and product demos are not enough to understand which opportunities are actually viable in their specific process or production context.

“I’ve identified an opportunity, but I don’t know how to make it happen.”
Process, innovation and IT managers may already have a defined problem, a solution idea or an initial experiment. What they need is to sharpen objectives and requirements, check the available data, compare technology alternatives and define resources, timelines and success criteria.

Both needs can arise in companies of any size and can coexist within the same organisation: having already adopted AI in one area does not mean knowing how to apply it to the next. These two needs call for different paths, with dedicated approaches and support methods. Confusing them is the most common way to waste time. Those who already have a clear idea risk starting over with an analysis that is too broad. Those who don’t risk jumping straight into the technical planning of a project they haven’t actually chosen yet.

Where should I start?

You understand that AI has potential, but you struggle to see concretely how it could help you. You are probably not yet starting from a specific problem, but are focused on AI technology itself. You will recognise this situation if you find it hard to answer these questions:

  • Can you name a single business process, among the many possible ones, that you would like to improve with AI?
  • Do you have an idea of which data in your organisation is already available and usable?
  • Do you know whether your in-house technical skills are sufficient to get started?
  • Have you already assessed whether it makes more sense to develop a solution, adapt an existing one or look for a technology partner?

If your answer to these questions is “I don’t know” or “it depends”, the issue is one of orientation. What you need first is an impartial snapshot of where the organisation stands in terms of data, infrastructure and skills. From there, you can build a list of realistic opportunities based on actual internal pain points and ranked by impact and feasibility.

This is exactly the purpose of the AI Assessment: an audit that evaluates the quality and availability of data, the infrastructure in place and internal skills, in order to identify the use cases that best fit the organisation’s real needs.

The goal is not to produce a generic list of possible AI applications, but to identify the most solid priorities, weighing expected impact against ease of implementation.

This is the path that enabled a logistics company to move from a general intention, “we want to optimise operations with AI”, to a specific use case: improving fleet management.

How do I make the opportunity happen?

The situation is different when the organisation has already identified an idea or an area of intervention but does not yet know how to turn it into a project.

You will recognise this situation if you are more comfortable with questions like these:

  • Do you already know which process you want to improve, but want to understand whether the data you have is enough?
  • Do you have in mind the result you want to achieve, even without the technical steps to get there?
  • Do you want to understand whether you need a model built from scratch, an adapted existing model, or something else entirely?

In this case, the problem is not working out what to do, but how to structure it.

The idea needs to be turned into measurable objectives, technical requirements, work phases and a realistic estimate of the time and resources required.

This is the role of AI Project Definition: support that starts from an idea that has already been identified (for example, “we want to automate quality control”) and translates it into an operational plan with clear objectives, technical requirements and a roadmap with defined phases and timelines.

This is the path that enabled a manufacturing company to turn the idea of predictive quality control into a project ready for implementation, with a compliant architecture and the peace of mind of knowing its data is secure.

Why the distinction matters

The two situations call for different kinds of work. The first needs breadth: looking across the whole organisation to find the right priority, identifying synergies between potential use cases to choose the highest-impact route. The second needs depth: going into the technical detail of a single idea that has already been chosen, managing the architectural and technological complexity.

Starting from the wrong point has a concrete cost. A general audit on an idea that is already clear risks wasting weeks re-discussing decisions that have already been made. Technical planning on an idea that is still vague risks producing a detailed plan for the wrong use case.

Distinguishing between orientation and project definition therefore helps make better use of time, resources and skills.

Dedicated services help those who lack the time, energy or internal resources to speed up the journey without losing control of it.

Where to go from here

You don’t need to answer these questions with absolute certainty to know where to begin: an initial sense of direction is enough. If the main uncertainty is about what to do, the first step is an assessment of the organisation’s AI maturity. If instead the uncertainty concerns how to deliver an idea that is already clear, the first step is the technical definition of the project.

In both cases, getting started does not require an immediate large-scale commitment. IT4LIA offers consulting expertise, computing infrastructure and a catalogue of services designed to support organisations one step at a time, from the initial assessment through to experimentation. You don’t need to have all the answers to start a first conversation and find out which of the two paths is right for you.

How to access IT4LIA

The path to getting oriented is simple and doesn’t require having every detail worked out: just book a call, explain where you are, and let the IT4LIA team point you in the right direction.

How to access IT4LIA: Start your AI Adoption journey now

Book a call with our AI experts now to find out which service is right for your organisation.

Veronica Brizzi, AI4I