Share:
Lets say: Click → Сhoose → Pay logic.
But recently, in internal discussions at Portalio, we caught ourselves repeating the same thought: “The user is not the one who is clicking anymore. And in several years this situation will be significantly shifted to AI Agents”
But if we look on it from the inside:
• AI model → the brain (understands, reasons)
• Agent → the actor (does things)
• Agentic route → how decisions and actions happen
• Agentic commerce → when agents actually buy/sell things
Before: User opens app → searches → compares → decides
Now/Shert term perspective: User says what they want and agent figures out the rest.
Andrii said it pretty directly in one of our calls with the clients: “People don’t actually like choosing. They just want the right and the best outcome.”
That’s where agentic routes come in.
It’s not a feature. It’s a different way of building logic.
Instead of:
• Step 1
• Step 2
• Step 3
You get:
• Understand intent
• Decide what to do
• Execute through APIs
Example (very real one)
User: “Plan me a weekend in Tuscany under €400”
Old way:
• 5 tabs
• 40 minutes
• 3 decisions and several solutions from different providers
Agentic route:
• Finds transport
• Picks stay
• Books
• Pays
Done. No UI/UX gymnastics and hours saved on that.
This is where it gets interesting and were we thik we will focus.
Agentic commerce = agents don’t just suggest… they execute transactions.
Not: “Here are 5 options”
But: “I’ve booked it accoridng to your confirmation.”
Inside Portalio, we don’t treat this as theory. We’re actively exploring how this plays out across what we’re building. In Pastaebasta we are creating API layer for future Ecomerce in Italy. We we plan to start onboard loca local comerce very soon for testing.
Pastaebasta started as something emotional: City stories. Photographers. Moments you take home.But then a question came up in the team: “What happens when Personal AI travel agent becomes the customer in two year?”
User tells their AI:
“I want something authentic from Bergamo. Not tourist stuff.”
The agent:
• Understands style
• Connects to Pastaebasta
• Propose variations of intersting stuff avaliable from different local suppliers.
• Picks a present
• Orders it
• Ships to your hotel or your friend
No browsing. No scrolling.
We’re not just building a shop or marketplace. We’re thinking about: creating an API layer where AI agents can discover and buy local products. That’s a very different and new product. And honestly – still early, but very real.
With Kursago it becomes even clearer.
Travel is full of friction:
• searching
• comparing
• timing
• payments
So we asked:“What if Kursago is not used by people… but by agents very soon?”
Scenario
User: “I need a transfer from Milan airport to Como at 18:00”
Agent:
• Connects to Kursago API
• Finds driver
• Selects best car option
• Books
• Pays
User just gets: “Your driver will be waiting.”
That’s an agentic route: No forms. No filters. No UX overload.
Everyone is talking about AI agents. We are builidng not an AI agent we are building systems ready for AI Agents in Italy.
Because for this to work, you need:
• APIs that are actually usable by agents
• clean data structures
• real execution (payments, confirmations, status)
Not just: “Here is our REST API, good luck.”
From what we’re testing and observing:
1. API-first is not optional anymore. If agents can’t connect → your product doesn’t exist for them.
2. You need execution, not just data. Agents must be able to:
• book
• pay
• confirm and handle logistic
3. Permissions matter. Agents need boundaries:
• limits
• approvals
• trust layers
4. UX/UI becomes almost invisible. The best experience is: no experience. Just result.
At Portalio, this is exactly the layer we’re focusing and building around. Not just websites. Not just apps.
We work on:
• agentic routes first (logic + UX thinking)
• AI-ready API layers
• payment + identity integrations
• custom AI agents for real use cases
And yes — we have and constantly building a team that can actually build this AI logic end-to-end:
• product logic
• UX flows
• backend/API architecture
• AI agent layer
Andrii said it best during one of the discussions: “Soon, your main user won’t be a person. It will be personal AI agent.”
So in case you have traditional business and you are interested to cooperate we are open.
Just to be very clear. This is not about adding AI to your product. It’s about asking: “What happens if no one ever opens our app again in two years?”
Share: