The question is no longer whether to integrate Artificial Intelligence into your workflow, but how to do so in a way that generates real value.
AI is now present in the tools we use every day, in conversations with customers and colleagues, and in the newsletters we read. But there is a substantial difference between asking ChatGPT to write an email and building a system that makes your team truly more effective and competitive.
The uncomfortable truth: AI alone is not enough
When we talk about AI in business, we often stop at the surface: automation here, a prompt there, integration with some tool that promises extraordinary results. The real question is: are we solving real problems or are we just following the trend of the moment?
In our case, we had to address this question honestly. And the answer was not at all obvious.
Every technological evolution brings rapid change. AI is no exception; in fact, it is probably the fastest of all. Those who stand still and watch are quickly left behind. Not necessarily by larger competitors, but by those who have understood how to use these tools to do better, faster and with less waste.
Three things we have learned
We don't have a magic formula. We tried, we failed, we started again. But along the way, we learned some fundamental things.
☝️AI amplifies strategic thinking, it doesn't replace it
If you don't know what you want to achieve, no algorithm will do it for you. Our best processes are those where we first clarify the goal, then find the right way to use AI to achieve it more quickly.
✌️The best integration is invisible
When a team member uses AI without even noticing because it's integrated into their usual workflow, you've won. If they have to interrupt what they're doing, switch applications, copy and paste data, you've just added complexity.
🤟Not all hours saved are equal
We've found that sometimes AI saves us time on low-value tasks; other times, it allows us to invest more time in higher-quality output. It's not enough to ask, ‘How much time are we saving?’ but ‘Are we working on the right things?’
What we have achieved...
- More informed decisions because AI allows us to explore more scenarios quickly.
- Replicable processes that we can standardise and scale.
- Less time wasted searching for scattered information, more focus on strategy.
Use case
Project management is one of those areas where integrating AI makes immediate sense. That's why we changed our perspective and approach. Instead of starting from scratch every time, we built a support system that intervenes at two key moments in the process — the generation of the PRD (Product Requirements Document) and the estimation of timelines for quotes.
The image above shows the flow we have built: from initial strategic thinking to operational decomposition.
We start with the brief, which includes the project's objectives, context and technical constraints. From there, the content goes through Google AI Studio with Gemini models. The prompt analyses the information, correlates it and considers priorities and complexity. Then we iterate until the output is aligned with what we want to achieve.
The result: a structured PRD with technical specifications, rough estimates, and decomposition into epics, stories, and tasks. Not perfect, but solid enough to enter the Kanban flow after review.
As we have learned (through trial and error)
It wasn't a walk in the park.
The first attempt was a disaster. So was the second.
Creating effective prompts takes time. You don't just write a prompt and immediately get what you want. To work, a prompt must be thought out, constructed, tested, and continuously improved.
In fact, achieving this flow took weeks of iteration: giving the model the right context, defining clear constraints, guiding the reasoning step by step, providing concrete examples. Each version of the prompt taught us where more precision was needed.
The value of your contribution
The review remains the critical point.
AI can generate a draft PRD in minutes, but it takes an expert eye to understand whether it makes sense in the specific context of the project. It can suggest timelines, but we are the ones who know that that customer needs more attention or that that feature hides complexities that the algorithm cannot see.
The challenge of estimates
This is where it gets complicated. As project managers know all too well, every project is different. You may have similar requirements, but they are never identical: the customer's context, existing integrations, technical constraints and priorities are always changing.
Creating a standard prompt for estimating the timing of a given project has proven to be more difficult than expected.
We therefore had to roll up our sleeves and build a more flexible system, capable of adapting to different types of projects without losing consistency.
🤔 Does AI really improve productivity?
The honest answer? It depends.
It depends on how you use it and what you want to achieve. Some teams may double their output while maintaining the same quality, while others may add an extra layer of complexity without any tangible benefits.
The difference? The former have integrated AI strategically, with clear and precise objectives. The latter have used it because "everyone else is doing it".
AI improves productivity when it allows you to focus on the things that matter: making better decisions, spending time on strategic conversations with customers, experimenting with solutions you wouldn't have had time to explore before.
So the real question to ask yourself is: is AI adding value to your work, or are you just adding another tool to your list?
Did you enjoy this article? Subscribe to our newsletter to receive updates on our activities and our experience with digital innovation.