Learn how to move from AI pilots to production with real integration, transparency, and value — featuring insights from the latest Frends webinar.
The era of AI-powered integration is no longer about what’s possible — it’s about what’s practical.
That was the core message of the latest Frends webinar, where CTO Asmo Urpilainen unpacked Frends' latest AI feature, answered live questions, and walked through what “agentic AI” really means in the context of business processes.
Asmo took a practical look at what it takes to go from isolated AI experiments to scalable, production-ready automation — introducing the new Intelligent AI Connector and real-world use cases.
“What we want to enable for our customers with the Intelligent AI Connector is basically an easy, safe way to start bringing AI into core business processes,” said Asmo. “To take that leap from the pilot that failed… to sustainable AI solutions that provide long-term value — month after month, year after year.”
Missed the live session? Here’s what you need to know — and why you’ll want to watch the full recording.
Why AI pilots fail — and what comes next
According to Asmo, it’s not the models that are failing — it’s the missing connection between those models and the systems, processes and data that actually run the business. The problem is more foundational, as he explains:
“AI cannot operate in a vacuum. Even if you have an amazing solution on paper, if you can’t connect it into your existing business processes in a safe, transparent way — it’s basically useless.”
Integration is the missing link. And it’s why Frends created a feature that brings AI into workflows, instead of just next to them.
What the Intelligent AI Connector actually does
In short? It lets you add AI into your processes like any other integration step — securely, transparently and with minimal complexity.
“You tell the AI what you want to happen, what’s the goal you want it to achieve,” explained Asmo. “The development time is just a fraction of what it used to be.”
One demo featured a tone-of-voice classification use case, where the AI determined whether a customer email was angry, calm or satisfied, and automatically routed it accordingly or generated an appropriate response.
But the real magic lies in the reasoning log:
“We don’t trust what we don’t understand,” said Asmo. “That’s why every AI execution includes a step-by-step reasoning log, so you can audit and reproduce exactly how the AI reached its conclusion.”
“You Don’t Need to Be a Prompt Engineer”
One audience member asked a common question: How can teams rely on AI when answers can vary?
“We’ve created a curated catalog of pre-prompted, pre-tested models,” said Asmo. “So users don’t have to be superstar prompt engineers — we’ve done the heavy lifting.”
Still want to bring your own model? That’s supported, too — whether via Azure AI Foundry or open-source models deployed with Ollama.
Common use cases, real business value
So where should companies begin?
Asmo offered several ready-to-deploy scenarios, including:
- Email classification and routing (e.g. support, sales, IT)
- Tone analysis for ticket prioritization
- Data extraction from PDFs, emails, Excel or free-text
- Equipment maintenance prediction based on IoT data and manuals
“Start small. Try it in one process. If you see value, add another. Then a third. Before you know it, you’re on your way from pilot to production”
- Asmo Urpilainen, CTO at Frends
He also addressed one of the biggest fears: costly re-platforming.
“We didn’t want customers to have to go through a massive, expensive re-platforming just to adopt AI. You can keep your existing systems and gradually push in AI where it makes sense.”
Trust, transparency and just enough human-in-the-loop
“You’ll never get 100% accuracy — not from AI, not from humans,” said Asmo. “But now, we can prove why something went wrong. We can learn from it. We can improve.”
The Frends platform allows users to build in human-in-the-loop checks only where needed, ensuring autonomy without losing control.
What’s Next?
The webinar also teased what’s coming next: the MCP protocol and the evolution of Frends into a full AI toolbox, enabling agentic AI to operate across systems in real time — and even letting users interact with the platform through natural language.
Watch the Full Webinar
Whether you're just starting to explore AI-powered integration or trying to scale existing pilots into production, this session is full of real-world advice and practical takeaways.