The Problem Every Business Has and Nobody Has Fully Solved
Walk into any modern Irish office and you will find employees using AI tools — ChatGPT, Claude, Gemini, Copilot — as a routine part of their working day. They are drafting emails, summarising documents, writing code, generating reports and solving problems with a speed and fluency that would have seemed extraordinary five years ago. The productivity gains are real, widely reported and increasingly embedded in how work actually gets done.
What makes the situation awkward rather than merely untidy is the distance between policy and practice. Most companies of a certain size have now written something about AI — an acceptable use note, a slide in an onboarding deck, a paragraph bolted onto an IT policy. Far fewer of them can say with any confidence what actually happens at nine thirty on a Tuesday morning when a deal is late, a client wants a two hundred page document summarised by lunchtime, and the person holding the file has a tool that will do it in ninety seconds. The policy says one thing. The deadline says another. That gap is where unmanaged AI use lives, and no amount of internal guidance closes it on its own.
The problem is what goes into those tools alongside the prompts. Every time an employee pastes a client contract into ChatGPT to summarise it, or drops a financial forecast into an AI tool to analyse it, or asks an AI to help draft a response to a sensitive legal query — they are potentially exposing confidential company data to a third-party AI system whose data handling, training practices and retention policies vary significantly between providers and may not meet the requirements of the company's own data governance framework, its regulatory obligations or its client confidentiality commitments.
This is not a theoretical risk. Data protection authorities across Europe have been examining AI tool use in professional settings since 2023. GDPR implications of AI data processing are a live compliance question. And the commercial consequences — if a client discovers their confidential information has been processed through a consumer AI tool without their knowledge — are significant.
Dublin-based startup VizCo was founded to address exactly this problem. The company has built a platform that allows employees to use AI tools — including the major consumer-facing AI assistants that have become standard in professional workflows — without exposing sensitive company data to those tools' underlying models.
The Pre-Seed Round
In early September, VizCo closed an oversubscribed pre-seed round of just over $2 million (approximately €1.75 million) with investment from Entrepreneur First — the talent-first investor that builds companies from scratch, starting with the founders — and a syndicate of angel investors with backgrounds in enterprise technology, data security and AI governance. The oversubscription is notable at pre-seed level, reflecting both the urgency of the problem VizCo is addressing and the quality of the founding team assembled through the Entrepreneur First process.
At this stage of a company's life, an investor is not underwriting revenue. They are underwriting a read on the future — a judgement about whether a problem that looks like a nuisance today becomes a line item on a budget in three years, and whether the people standing in front of them are the ones to solve it. The oversubscription says the read landed on the right side of that question.
The round gives VizCo the capital to build out its product and begin acquiring its first enterprise customers in Ireland and the UK — the two markets with the highest concentration of professional services firms currently navigating the tension between AI productivity and data security compliance.
How VizCo Works
VizCo acts as an intelligent intermediary layer between the employee and the AI tool. When an employee submits a prompt that contains sensitive data — client names, financial figures, personal information, proprietary processes — VizCo's system detects and redacts or anonymises that data before it reaches the AI model, and then reintegrates the AI's response with the original context restored on the way back to the employee. The net effect is that the employee gets the full productivity benefit of using the AI tool, while the AI model itself never processes the actual sensitive data.
The approach is technically demanding — building a system that can accurately identify sensitive information across the full range of document types, communication formats and query patterns that professional employees use, without generating so many false positives that it disrupts the workflow it is meant to protect, requires significant investment in both natural language processing and enterprise-grade reliability. The founding team's background in AI and data engineering is central to the company's ability to execute on that technical requirement.
The harder test is behavioural rather than statistical. A protection layer that blocks too much does not get fixed, it gets routed around — the employee quietly opens the personal account they were told not to use, and the company ends up with less visibility than it had before. A layer that blocks too little protects nothing. The product therefore has to be judged less on a single accuracy number and more on whether it survives contact with an ordinary working week, where speed matters to the person using it and the alternative to using the tool is doing the job slowly by hand.
The Commercial Test
Selling a safety layer into businesses is not the same motion as selling a productivity tool. Nobody buys governance because a piece of technology is elegant. They buy it because a client asked a question they could not answer, because an insurer tightened a renewal, because a regulator wrote a letter, or because a board finally decided that finding out after the fact was not an option. That makes the buying cycle slower and the sales conversation drier, but it also makes the resulting contract stickier — once a supplier is written into how a firm meets its obligations, it is not casually swapped out.
There is a standing risk in this corner of the market, and VizCo's first customers will want it addressed honestly. The companies that build the underlying models are moving up the stack, adding enterprise controls, retention guarantees and compliance tooling of their own. A standalone layer has to justify itself against that. Its case is that it works across every tool a firm's people actually use rather than the one a vendor prefers, that it stays neutral when a provider changes its terms, and that it produces an answer an auditor or a client can be shown. Whether that holds is the question the next twelve months will ask.
The Enterprise AI Safety Market
VizCo enters a market that is growing rapidly as enterprises in every sector grapple with the practical governance challenge of widespread AI tool adoption. The enterprise AI safety and governance market — covering the tools, platforms and frameworks that allow companies to use AI productively while managing data, compliance and reputational risk — has attracted significant investment from enterprise technology investors in the US and Europe throughout 2025 and 2026.
Ireland, as the EU headquarters of most major AI companies and a significant centre for financial services, professional services and pharmaceutical research, has a particularly acute version of the enterprise AI governance challenge. The combination of GDPR obligations, sector-specific regulatory frameworks and the confidentiality requirements of serving global clients makes the Irish professional services market a natural early adopter of enterprise AI safety tools.
There is a second reason the Irish market carries more weight than its size would suggest. This is a product that has to be sold into a procurement process, and procurement wants a vendor it can interrogate — someone who can walk into a room, answer awkward questions about data flows, and stand behind an answer. A company built here, close to the firms it is selling to and the regulators it has to satisfy, has an advantage in that conversation that a distant platform team cannot easily replicate.
The Bottom Line
VizCo has identified a problem that sits at the intersection of two of the most significant technology trends in the current enterprise landscape — the widespread adoption of consumer AI tools by professional employees, and the growing regulatory and commercial pressure to ensure those tools do not expose confidential data. The $2 million pre-seed round gives it the runway to prove the technical approach and acquire the first enterprise customers who will validate the commercial model. In a market where the problem is universal and the solutions are early-stage, the timing is right.
What to watch from here is unglamorous and telling: whether the first logos are paid customers rather than pilots, whether those customers come from the confidentiality-heavy end of the economy where the risk is contractual rather than theoretical, and whether the layer holds up as the model providers push harder to own the whole stack themselves. If those three things break the right way, a small Irish company ends up sitting in the middle of how a very large number of people do their jobs.