Stop Building Apps. Start Building Agents.
Low-code was supposed to be the revolution. Drag and drop, no developers needed. But the landscape has shifted again. Today, you don’t even need to drag and drop. You describe what you want and an app appears. And every company in the low-code space is now navigating a collision with Anthropic, OpenAI, and a wave of vibe coding startups all converging on the same territory.
Tiago Azevedo, the CIO of OutSystems, sits right in the middle of that collision. In this episode of Talking AI, he sits down with Matt Paige to share how he’s leading through the chaos, why he refuses to pump the brakes, and what OutSystems is building to stay ahead.
The CIO’s job isn’t to say no. It’s to manage the risk.
Tiago is candid about the pressure CIOs face right now. Every 24 hours, something new drops and everyone loses their mind. His response isn’t to lock things down. He says saying no is the easiest solution, but it’s also the wrong one. The real job is opening the doors while learning how to manage the risk that comes with it. That’s the hard part, and it’s where the leadership challenge lives.
Everything that isn’t agentic is already legacy.
One of the most striking moments in the conversation is Tiago’s declaration that everything non-agentic is now legacy work. The old playbook, where you layer foundations, re-architect systems, and modernize over months or years, is dead. Today, you bypass the mess. You layer agents on top that fill the gaps between broken systems, and you solve for the specific business problem rather than trying to fix the infrastructure underneath.
Use AI to figure out where AI fits.
Tiago’s practical advice for leaders stuck on where to even start is refreshingly simple. Stop brainstorming use cases. Just describe your biggest business problem to an LLM, give it all the context about your systems, data, and processes, and ask it to help you design an agentic solution. Iterate from there. He says you’ll be amazed at what comes back, and that approach is exactly how his team built Deal Mate, a team of agents that prepares sales reps for meetings with deep research, account briefings, and follow-up nudges.
From drag-and-drop to “just describe it.”
The episode also covers OutSystems’ latest launch: Mentor, which brings natural language vibe coding directly into the platform. Users can now describe what they want to build conversationally, and the platform generates the application. The key architectural distinction: from the generated application to the running code, everything is deterministic. The only non-deterministic layer is the conversational input. That means all the security practices, compliance guardrails, and enterprise-grade code generation that OutSystems has built over 20 years remain intact.
The wow factor is not enterprise-ready.
Tiago is also honest about the competitive noise. He acknowledges the market is flooded with impressive demos, but draws a sharp line between prototyping something in an afternoon and actually maintaining, evolving, and running it at enterprise scale. His advice to CIOs: try everything, but be careful with divergence. At some point, you have to choose a path.
Think big, start small, scale fast.
The episode closes with Tiago’s framework for AI transformation. Identify the biggest problem your company needs to solve, not your team’s productivity, but the real business problem your CEO cares about. Use a scout (one person) or a strike team to prove the solution. Then scale fast to the full dimension of the problem.
Watch or listen to the full episode: https://hatchworks.com/talking-ai/
Key Moments
- 00:00 – Matt’s intro: low-code meets vibe coding, the competitive collision
- 01:17 – Tiago on the pace of change and what makes this moment unlike anything before
- 06:20 – Saying no is the easiest solution, managing the risk is the hard part
- 07:49 – Bypass the mess: why agents fill the gaps legacy modernization never could
- 09:10 – Everything that is not agentic is literally legacy work
- 10:15 – Use AI to figure out where AI fits: the meta approach to use cases
- 11:30 – Deal Mate: the team of agents that prepares sales reps for meetings
- 15:07 – Why adding more fields to Salesforce was the wrong move, building agents is the right one
- 16:25 – Mark Zuckerberg building an agent to do his job
- 17:23 – OutSystems’ 20-year journey from visual development to agentic systems engineering
- 19:58 – The deterministic magic behind OutSystems Mentor
- 22:04 – One platform: infrastructure, integrations, UIs, agent skills, and deployment
- 30:19 – 40% customer service automation with AI, versus under 10% before
- 33:48 – How AI is augmenting, not replacing, engineering and product roles
- 39:41 – That’s 2008 and this is 2026, you have to change
- 41:27 – The wow factor vs. enterprise reality: why prototyping isn’t the hard part
- 46:17 – Tiago’s advice: identify the biggest problem, feed it to an LLM, build the solution
- 48:42 – Think big, start small, scale fast
