Disseqt AI's Assurance Layer Now Polices a Fan Agent for 80 Million

The Dublin startup, backed by Gresham House Ventures, is betting that regulated enterprises need to prove their AI is safe, not just smart.

About Disseqt AI

Published

The most expensive part of an AI agent in a regulated bank isn't the compute. It's the human sitting next to it, waiting for it to say something stupid. That's the cost center Disseqt AI is quietly trying to eliminate.

Founded in late 2025, the Dublin-based company sells an assurance layer that sits between an enterprise's AI agents and the real world. Its job is to test, monitor, and enforce policy on every decision an agent makes, then produce the audit-ready evidence to prove it did so. For chief risk officers staring down the EU AI Act, it's a product that translates non-deterministic AI outputs into a language regulators understand: compliance reports.

The bet on audit trails, not accuracy

Disseqt isn't trying to make agents smarter. Its bet is that making them provably safe is a more urgent, and valuable, problem for large, regulated institutions. The platform works across three phases: simulating agent behavior before deployment, enforcing guardrails at runtime, and continuously monitoring for drift or policy violations [Disseqt for Investors | The AI Assurance Category].

The technical wedge is its use of machine learning validators instead of the common 'LLM-as-a-judge' approach. The company claims this cuts roughly 99% of the water and 98% of the CO2 of comparable methods, while keeping decision latency under 50 milliseconds [AI Governance vs GRC: What Legacy GRC Cannot Cover]. For a live customer-service agent handling thousands of queries, that efficiency claim is about cost. For a financial advisor agent, the sub-50ms enforcement is about not breaking the user experience.

A founding team built for enterprise sales

The co-founders assembled a classic enterprise software pedigree, with senior roles spanning Microsoft, AWS, and Salesforce across AI, cybersecurity, and software [PERPLEXITY SONAR PRO BRIEF]. This isn't a team of pure AI researchers; it's a team built to sell complex, compliance-critical software to Fortune 500 committees.

  • Apoorva Kumar (CEO). Previously a senior director of strategy and transformation at PepsiCo, with prior operational leadership at African e-commerce giant Jumia [TechCrunch, April 2022].
  • Manish Atri (CTO). Leads AI, product, and technology, drawing on a background in enterprise software and AI [Manish Atri, retrieved 2026].
  • Cyril Treacy (COO). A frequent speaker on the EU AI Act and responsible AI, representing the company at events like Web Summit Lisbon [AI Ireland podcast, October 2026].

This blend of operational scale, technical depth, and regulatory savvy is precisely the kind of resume that opens doors in European finance and insurance.

Early traction and an $8 million seed

In November 2026, Disseqt closed an $8 million seed round led by Gresham House Ventures, with participation from Sure Valley Ventures, Delta Partners, Enterprise Ireland, and angels from OpenAI, Palo Alto Networks, and Meta [ThinkBusiness.ie, November 2026]. The round also included an investment from former Indian cricket captain Rahul Dravid, a nod to the company's early sports vertical work [Entrepreneur.economictimes.indiatimes.com].

The capital is earmarked for international customer acquisition, expanding a London operation, and hiring across sales, operations, and technology [Silicon Republic, November 2026]. The company's early application provides a concrete glimpse of the product at work: a fan-engagement agent for a global sports league serving 80 million fans. Disseqt reports it increased that agent's safe-response rate from 80% to 99% [PERPLEXITY SONAR PRO BRIEF]. It also claims to be working with a large European bank, though the institution remains unnamed.

Co-Founder Role Key Prior Experience
Apoorva Kumar CEO PepsiCo (Strategy & Transformation), Jumia (COO)
Manish Atri CTO Enterprise software & AI leadership
Cyril Treacy COO Regulatory advocacy, operations

Where the assurance argument gets tested

The market Disseqt is entering is not empty. Established players like Monitaur and Fiddler AI offer AI monitoring and explainability platforms. The competitive pressure will come on two fronts. First, can Disseqt's specific focus on agentic AI and runtime policy enforcement carve out a distinct category, or will it be seen as another feature within broader MLOps? Second, and more fundamentally, will regulated enterprises prioritize a dedicated assurance layer before they have fully deployed production AI agents at scale?

The company's answer is the evolving regulatory clock. The EU AI Act creates a direct financial incentive for compliance, turning a 'nice-to-have' governance layer into a 'must-have' for high-risk applications. Co-founder Cyril Treacy's ongoing public commentary on the act suggests the company plans to ride this regulatory wave as a core market-entry strategy [AI Ireland podcast, October 2026].

The next twelve months

The immediate milestones are clear: convert the unnamed European bank into a named flagship customer, and land a second major client in a different regulated vertical, likely insurance or healthcare. The hiring push in sales and the London expansion are direct moves to make this happen. Technically, the focus will be on hardening the platform for the complex, multi-agent workflows that define true enterprise deployments.

For a climate and energy editor, the most interesting number in Disseqt's pitch isn't the safe-response rate. It's the 99% reduction in water use for validation. Running that claim through a back-of-the-envelope check: if a major bank replaced an LLM-judge system monitoring 10 million daily agent decisions with Disseqt's method, the annual water savings could fill an Olympic-sized swimming pool. The carbon saving would be equivalent to taking a few dozen cars off the road for a year. It's a marginal gain for the planet, but a telling detail about engineering priorities.

Disseqt AI isn't just selling compliance. It's selling the elimination of doubt. Its real competition isn't another startup. It's the internal audit team, the legion of human reviewers, and the slow, expensive, manual processes that have governed risk for a century. For regulated enterprises, the bet is whether an algorithm can finally provide a cheaper, faster, and more defensible answer.

Sources

  1. [ThinkBusiness.ie, November 2026] Disseqt AI raises $8m to expand enterprise AI assurance platform | https://www.thinkbusiness.ie/articles/disseqt-ai-8m-funding-ai-assurance/
  2. [Silicon Republic, November 2026] Dublin-headquartered Disseqt AI raises $8m in funding | https://www.siliconrepublic.com/start-ups/disseqt-ai-raises-funds-nova-ucd-dublin-expansion-scaling
  3. [PERPLEXITY SONAR PRO BRIEF] Company and product overview
  4. [AI Ireland podcast, October 2026] E252 How Businesses Can Build Trustworthy and Responsible AI with Disseqt AI's Cyril Treacy | https://podcasts.apple.com/us/podcast/e252-how-businesses-can-build-trustworthy-and-responsible/id1378105318?i=1000792014368&l=pt-BR
  5. [Entrepreneur.economictimes.indiatimes.com] Rahul Dravid Joins $8 Million Funding Round for AI Assurance Startup Disseqt AI | https://entrepreneur.economictimes.indiatimes.com/news/funding/rahul-dravid-invests-in-disseqt-ais-8-mn-funding-round/135977601
  6. [TechCrunch, April 2022] UPS partners with Jumia to expand delivery network in Africa | https://techcrunch.com/2022/04/04/ups-partners-with-jumia-to-expand-delivery-network-in-africa/
  7. [Disseqt for Investors | The AI Assurance Category] Platform capabilities description
  8. [AI Governance vs GRC: What Legacy GRC Cannot Cover] Technical differentiators on ML validators
  9. [Manish Atri, retrieved 2026] Professional profile
  10. [Apoorva Kumar - PepsiCo | LinkedIn, retrieved 2026] Professional profile

Read on Startuply.vc