Are we ready to trust AI agents in decision-making in high-stakes situations? In a world where artificial intelligence is being used to automate critical decisions, this is one of the big questions. In the case of Portia AI, a growing UK start-up launched in 2024, the answer, however, is yes, but only when there is a radical reconsideration of the way in which agents are constructed and implemented.
Portia AI, based in London, aims to deliver predictable, controllable, and secure AI agents. Co-founded by Emma Burrows, former CTO of Stripe UK, and Mounir Mouawad, who helped grow Google Pay across 30 countries, the company’s philosophy emphasises that AI tools should be powerful, governable, accountable, and auditable. Their backgrounds in secure financial infrastructure shape this vision.
The Agent Revolution- Guardrails
AI agents are autonomous software entities that make decisions and communicate via large language models. They are more capable but also more unpredictable than traditional AI. Many startups rush to develop these agents, but few are ready for enterprise use. Burrows and Mouawad identified this gap early, emphasising the need for agents that act responsibly, visibly, and under human control.
Portia provides a solution to this with a developer-friendly offering:
- Foreseeable planning: Agents are capable of creating action plans in sequences that can be checked and altered prior to implementation.
- Human-in-the-loop control: In case of necessity, agents stop to seek clarification or permission, introducing an important safety level.
- Safe authentication: Agents have access to tools and information only when they need them with just-in-time (JIT) access tokens, thus minimising risk and streamlining compliance.
An Enterprise-Ready Framework of AI
Portia isn’t experimenting with AI like competitors. Instead, it focuses on scalability and reliability. Its open-source SDK allows developers to build structured workflows with checkpoints, intervention points for human input, and audit logs.
Portia’s system automates processes like finance, customer service, and operations while avoiding reliance on potentially unpredictable AI.
It doesn’t prioritise LLMs but offers integration with OpenAI, Anthropic, Claude, Google Gemini, and local models like LLaMA, letting companies choose based on security or cost.
Investor confidence and funding
In early 2025, after emerging from stealth, Portia raised a £4.4 million seed round led by General Catalyst, with investors like First Minute Capital, Stem AI, and angel investors from AI and fintech. The appeal wasn’t just its technical team but also its value proposition: creating responsible AI systems enterprises can trust. Juliet Bailin of General Catalyst noted Portia’s focus on not only deploying AI agents but also developing a governance framework. The funds are being used to expand the engineering team, enhance the SDK and developer dashboard, and acquire enterprise clients for early pilots.
A Builder-built Product
Burrows and Mouawad have assembled a group of engineers who are extremely experienced in payments, automation, and developer tooling. The lean organisation, five full-time employees, and a global community of contributors allow it to be fast and yet stay close to the actual needs of the customers.
Portia features include:
- Â Plan learning (the skill to like successful workflows so that they can be used later)
- Â Safe connections with such apps as GitHub, Slack, and Zendesk
- Â A headless browser agent that will allow the use of tools in real-time behind a login wall
- Â Paginated APIs and long document processing are essential in enterprise workloads
The New Standard of AI Automation
Because enterprises are increasingly considering how to apply LLMs in a live scenario (intelligent ticket triage, automated reporting, and customer follow-up, among others), the Portia framework is becoming more applicable. It provides a moderate course between the heavily regulated conservatism and the unregulated experimenting.
The interest in controlled automation is of interest to enterprise product leaders, compliance officers, and CTOs. According to one investor, Portia provides an infrastructure to feel confident about AI in a production setting. That is what makes it outstanding.
Looking Ahead
Portia’s roadmap emphasises expanding integrations, testing planning agents, and supporting enterprises. It’s also releasing an educational cookbook on responsible AI. The team is hiring developers, security engineers, and business leaders in the U.K., EU, and U.S.
Portia AI offers trust, control, and predictability, crucial as business automation grows, rather than magic. Connecting capability with accountability is vital for AI’s future, and Portia AI strives to be that bridge with a thoughtful, enterprise approach.
