Bogdan Năforniță is the CEO & Co-Founder at Profluo, a leading company using Agentic AI to automate invoice processing and financial operations. With a strong background in finance and software development, Bogdan has been at the forefront of AI adoption in finance.
Bogdan, tell us a bit about your journey – how did you come to work on invoice automation and intelligent finance tools?
I was a CFO for more than 20 years, but I was an unusual one – always on the lookout for operational improvements, whether through technical solutions like ERP systems and finance automation, or through deeper business process re-engineering. As a CFO, you need the curiosity to understand what truly drives financial results and the courage to act on that understanding.
What was the original idea behind Profluo, and how has the product evolved since?
Profluo was born out of a deep frustration with the state of finance automation tools at the time – they were brittle, siloed, and far from adaptable.
We initially set out to build a finance workflow platform, but as the product evolved, we realized we were straying beyond our core expertise. More importantly, we had underestimated the central role that finance documents play in the automation journey. That insight led us to refocus on AI-driven invoice processing.
Today, we’re doubling down on documents and processes through an Agentic AI lens – bringing modern, intelligent automation to the entire CFO office.
For people unfamiliar with the space, what exactly is “Agentic AI” and how does Profluo use it differently from traditional automation?
Agentic AI is essentially autonomous automation – it understands process goals and dynamically designs the most effective steps to achieve them. In finance, Agentic AI blends advanced AI with traditional automation, delivering the accuracy, efficiency, compliance, auditability, and confidentiality that finance teams require.
What are the biggest pain points finance teams face today when it comes to e-invoicing, especially with mandates like RO e-Factura or the upcoming ViDA changes?
The main challenges are complexity and cost – finance teams face multiple technical integrations and intricate compliance requirements across different jurisdictions. Aligning enterprise ERPs with these demands can be overwhelmingly expensive.
Implementing, deploying, and maintaining such a fragmented infrastructure often isn’t worth the effort or investment. That’s where unified platforms like Profluo come in, streamlining operations and reducing the burden for large and mid-market companies.
Profluo processes thousands of invoices daily. What makes your AI engine particularly good at this task? Is it the training data, the rules, the integrations?
It all starts with our team’s expertise – many of us come from finance roles as accountants and controllers. We deeply understand which problems actually matter to our customers, and we stay focused on solving those first.
Second, we’re obsessed with accuracy and speed. Our AI achieves over 95% accuracy and performs tasks up to 6x faster than manual work. In finance, even small errors can be high-risk. What to a data scientist is a 10% error in column reading
accuracy, to an accountant it can mean the AI mixes quantities and unit prices, therefore the resulting invoice is 100% wrong.
Our third major advantage is the type of data we work with. We thrive on messy, real-world edge cases. Our customers bring us complex problems and documents, which constantly push us to evolve our AI, enhancing accuracy while improving generalization across scenarios.
You’ve worked closely with Romanian companies during their transition to mandatory e-invoicing. What lessons can other countries learn from Romania’s early move?
We’ll dive deeper into this in our upcoming e-Invoicing study, but the key takeaways can be summarized in three main insights:
- For authorities – don’t rush the rollout. Take the time to explain the system, onboard all stakeholders, test the environments thoroughly, and allow space for organizations to absorb the complexity and associated costs.
- For businesses – use this as an opportunity to rethink your automation strategy. Focus on end-to-end workflows, automated accounting, and built-in compliance.
- For multinational companies – the integration complexity across jurisdictions makes it far more efficient to adopt a unified compliance platform that connects with your ERP – rather than adapting the ERP to every local, constantly shifting requirement.
You’re a big advocate for using AI in everyday workflows. Can you share your personal AI stack – what tools do you use day-to-day to stay productive and informed?
Sure, happy to! I firmly believe that AI should be part of everyday workflows, not just experimental projects.
- Deep research
I mostly rely on Gemini. It is great at researching cutting edge technical and financial topics. One caveat: it is much weaker when exploring creative solutions, the perspectives tend to be bland and generic.
Frequency of use: almost daily
- Note-taking
I use Meetgeek because it is solid at understanding low-resource languages (i.e. the ones for which AI training is less available), it integrates well with our CRM and it provides plenty of automation opportunities. Taking notes should be just a first step in a longer journey.
Frequency of use: multiple times during business hours
- Coding
For Excel automation, I usually code in basic Python and VBA. For this task I usually rely on Deepseek or Gemini. I love that they not only give me alternative solutions, but they also anticipate possible improvement scenarios in the same reply.
Frequency of use: around 2-3 times per week, but in longer sessions
- Private AI research
For our own AI LLM engine, I experiment with foundation LLM models in low-resource tech stacks – I experiment with degrees of distillation, quantization, fine-tuning patterns, RAG behavior. For this I use LM Studio. This is not an AI tool, but a toolkit that manages LLM models. As I am using real customer data and documents (with their permission), this whole environment is entirely local, sandboxed, and encrypted.
Frequency of use: weekly (hard to keep up with new LLM models!)
- Transactional Record-to-Report and Procure-to-Pay
Romania is a full e-Invoicing country and our own Profluo agents automatically perform all accounting and approval tasks. They notify my team when there are new e-Invoices to be approved, check the contracts for them, prepare the accounting entries for our ERP and then notify our outsourced accounting company who just logs in, checks and sends the data over to the ERP. This way we use part of the accounting fee to get management analyses, as the compliance effort is automated away.
Frequency of use: agents get triggered every 10 minutes for a close to real-time experience
Many countries are adopting national e-invoicing systems, both formats and transmission protocols – XRechnung in Germany, KSeF in Poland, Chorus Pro in France, PEPPOL in Belgium. How is Profluo preparing for this next wave of change?
We’re actively working toward compliance with all major national e-invoicing platforms and formats, as this is an important part of our full automation promise.
The European e-invoicing landscape is in some ways getting simplified (with standard electronic formats), but in other ways the compliance gets more complicated (with various public systems to be integrated and with varying degrees of public finance reporting requirements). We anticipate that its full impact will be harnessed much more efficiently using Profluo’s AI engine, which is tailored exactly for this kind of complexity.
Where do you see finance automation going in the next 3–5 years? Will AI replace accountants or empower them?
AI will empower accountants, not replace them. Those who embrace these technologies will gain a competitive edge, while those who resist risk being left behind.
AI-enabled accountants will shift their focus to higher-value, client-facing services such as business, tax, and treasury advisory, or commercial controllership. Meanwhile, those who rely solely on traditional methods may find themselves limited to routine compliance and bookkeeping tasks.
I don’t foresee a near future where your accountant is replaced by an always-available AI tool. AI will support and inform, but trusted human advisors will remain essential for judgment, nuance, and strategic insight.
For finance leaders who are still hesitant about AI: what’s one mindset shift you’d recommend?
Just start doing it. Whether you’re experimenting for fun, learning the ropes, or aiming for a full-scale implementation – take the first step.
This is the most exciting time in history to be a CFO. Don’t miss the opportunity to lead the change!
And finally, if someone wants to try Profluo or learn more, where should they start?
We encourage anyone interested to reach out and share their challenges with us. We’ve helped complex organizations solve tough AI automation problems and there’s a good chance we already have a fast, effective solution ready for yours.
