Today, 26.03.2026
Good morning, Human. The phrase "computer agent" has been floating around AI circles for months, but this week it landed somewhere concrete: the CFO's inbox. London-based Zalos just closed a €3.1 million seed round to build AI agents that log into enterprise finance systems and operate them the way humans do – clicking through screens, entering data, running reconciliations. The mechanism matters more than the money: this is automation that doesn't require replacing the systems companies have spent years configuring.
The Lead: When AI Learns to Use Your Software
The Zalos raise, announced this week, represents something more interesting than another fintech seed round. The company, founded by William Fairbairn and Hung Hoang after their Y Combinator batch, has built what it calls "Computer Agents" for finance operations – software that converts screen recordings of finance workflows into autonomous agents capable of repeating those processes indefinitely.
The round was led by 14 Peaks Capital, with participation from Cohen Circle and 20VC. But the angel list tells the real story: Mike Lenz, CFO of FedEx; Ian Sutherland, CFO of Tide; Long Dinh, CFO of Ada. When the people who actually run finance departments at major companies back a seed-stage startup, it says something about how real the problem is.
Here's the mechanism hiding under the headline. Modern finance teams operate across a fragmented stack of ERPs (Enterprise Resource Planning systems), CRMs, spreadsheets, and banking platforms that were never designed to work together. According to Zalos, APIs between these systems are often missing or incomplete, which means finance teams become the "human API" themselves – manually stitching data across systems to complete billing cycles, close the books, and produce reporting.
Zalos sidesteps the integration problem entirely. Its agents log into platforms like NetSuite, Sage, and SAP S/4HANA with a username and password, navigate screens, enter data, and check against controls – exactly the way a human would. Every action is captured in an auditable log, and the platform holds SOC 2 Part II certification alongside enterprise single sign-on and role-based access controls.
The timing matters. 2026 is increasingly being called the year of "agentic AI" – systems that don't just generate outputs but take actions. OpenAI and Anthropic have both moved into the space with generalist computer agents, but Zalos is purpose-built for finance operations, where the stakes of getting it wrong are categorically higher. Finance teams cannot operate on 90% accuracy; the agents need finance-specific skills, and they need every automated action logged in a format auditors can follow.
Watch the calendar: as the EU AI Act's high-risk system requirements approach their enforcement date, the question of how AI agents interact with enterprise systems – and how those interactions are documented – becomes increasingly material.
The Funding Picture: Deeptech Captures a Third of European VC
The Zalos raise lands against a backdrop of structural shift in European venture capital. The 2026 European Deeptech Report, published this week by Dealroom, Walden Catalyst, and Lakestar, shows VC-backed deeptech companies now valued at $690 billion, with investment in the sector capturing a record 32% of total venture capital – more than double its share a decade ago.
The resilience is notable. While broader venture markets remain well below their 2021 peak, European deeptech funding reached $20.3 billion in 2025, just 4% below its all-time high. Defence and resilience now account for 43% of all European deeptech funding, up from 20% in 2022. The Future of Compute segment saw funding more than double to $2.5 billion.
But the report also surfaces a persistent structural gap. European deeptech companies face an estimated $4 to $24 billion annual funding shortfall, particularly at the growth stage. As a result, 70% of late-stage capital comes from non-European investors. The consequences are clear: companies scale more slowly, strategic control often shifts abroad, and economic value is captured outside Europe.
Yoram Wijngaarde, founder and CEO of Dealroom, put it directly: "Deeptech is where the puck is heading. Europe needs the conviction of capital to make sure we are in the game."
The Platform Play: Spotify Builds a Wall Against AI Slop
Meanwhile, a different kind of AI problem is forcing platform responses. Spotify announced this week that it's piloting a new opt-in feature called Artist Profile Protection, allowing artists to review and approve eligible releases before they go live on their profiles.
The company says protecting artist identity has become "a top priority for 2026," noting that "the rise of easy-to-produce AI tracks has made the [misattribution] problem worse" across streaming services. The problem has affected artists across genres and career stages – last year, AI-generated tracks mimicking Tyler, the Creator's album flooded Spotify ahead of its official release, with a fake version briefly holding the number two spot.
The mechanism is straightforward: artists who opt in receive notifications when music is delivered to their profile and can choose to approve or decline it ahead of release. If declined – or if the artist takes no action – the release won't appear on their Spotify profile, though it may still go live on other streaming services. To avoid slowing down legitimate releases, each artist will also be assigned a unique "artist key" they can share with trusted distributors to enable automatic approval.
This arrives a week after Sony Music revealed it had asked streaming platforms to take down more than 135,000 songs it says were created by fraudsters using generative AI to impersonate artists on its roster. Deezer recently disclosed it was receiving approximately 60,000 fully AI-generated tracks per day – around 39% of all daily deliveries.
The broader signal: as AI-generated content scales, platforms are being forced to build new verification and attribution infrastructure. The question of who controls identity – and who bears the cost of defending it – is becoming a design problem, not just a policy debate.
The Regulatory Calendar
The EU AI Act's implementation timeline continues to shape compliance planning across the ecosystem. According to the European Commission's AI Act Service Desk, the majority of the regulation's rules come into force on , including requirements for high-risk AI systems listed in Annex III, transparency rules under Article 50, and the full market surveillance framework.
Member States should have at least one AI regulatory sandbox per country established and operational by that date. Finland became the first EU member state with fully operational AI Act enforcement powers on ; other member states are expected to follow throughout the first quarter.
One complication: the European Commission missed its legal deadline of to provide essential classification guidelines for high-risk AI systems. The delay is primarily due to the "Digital Omnibus" proposal, a legislative package introduced in late 2025 aimed at simplifying compliance. The proposal could push back certain high-risk compliance deadlines by up to 16 months – but planning for delay is not a strategy.
The Numbers That Matter
- $690 billion – Total value of European VC-backed deeptech companies, per the 2026 Dealroom/Walden Catalyst/Lakestar report
- 32% – Share of European venture capital now flowing to deeptech, more than double the share a decade ago
- 43% – Portion of European deeptech funding going to defence, security, and resilience, up from 20% in 2022
- 70% – Share of late-stage European deeptech funding coming from non-European investors
- 135,000 – AI-generated songs Sony Music has asked streaming platforms to remove for impersonating its artists
- 60,000 – Fully AI-generated tracks Deezer receives per day, representing 39% of all daily deliveries
- 2 August 2026 – Date when EU AI Act high-risk system requirements and transparency rules become enforceable
The Week Ahead
The European Council meets in Brussels on , with AI competitiveness and digital sovereignty expected to feature in discussions. The AI Office continues work on the General-Purpose AI Code of Practice, with the final code expected by . Spain's Agency for the Supervision of Artificial Intelligence (AESIA) has released 16 guidance documents to support organizations in complying with the EU AI Act – worth reviewing for anyone building compliance infrastructure.
The Thought That Lingers
There's something quietly significant about the Zalos approach: automation that works by learning to use existing systems rather than replacing them. It's a bet that the fastest path to transformation isn't building new infrastructure but teaching AI to navigate the infrastructure that already exists. The same logic applies to the broader European AI ecosystem – the question isn't whether Europe can invent the future, but whether it can build the capital structures, governance frameworks, and operational discipline to own what it creates. The deeptech report's $690 billion valuation is impressive. The 70% of late-stage capital coming from outside Europe is the number that should keep policymakers awake.
Human×AI Daily Brief is compiled from EU-Startups, Tech.eu, Dealroom, Music Business Worldwide, the European Commission's AI Act Service Desk, and verified funding announcements. This is meant to be useful, not comprehensive.
Frequently Asked Questions
Q: What are Computer Agents in enterprise finance?
A: Computer Agents are AI systems that log into enterprise software like ERPs and accounting platforms using standard credentials, then navigate screens, enter data, and complete workflows the same way a human employee would. Unlike traditional automation that requires API integrations, they learn from screen recordings of existing processes.
Q: When do EU AI Act high-risk system requirements become enforceable?
A: The majority of EU AI Act rules, including requirements for high-risk AI systems listed in Annex III and transparency obligations under Article 50, become enforceable on 2 August 2026. However, the Digital Omnibus proposal could extend certain deadlines by up to 16 months.
Q: What percentage of European venture capital now goes to deeptech?
A: According to the 2026 European Deeptech Report from Dealroom, Walden Catalyst, and Lakestar, deeptech now captures 32% of all European venture capital – more than double its share a decade ago. Total sector value has reached $690 billion.
Q: How is Spotify addressing AI-generated music misattribution?
A: Spotify is piloting Artist Profile Protection, an opt-in feature that lets artists review and approve releases before they appear on their profiles. Artists receive notifications when music is delivered and can approve or decline it. A unique "artist key" can be shared with trusted distributors for automatic approval.
Q: What is the European deeptech funding gap?
A: European deeptech companies face an estimated $4 to $24 billion annual funding shortfall, particularly at the growth stage. As a result, 70% of late-stage capital comes from non-European investors, which often shifts strategic control and economic value capture outside Europe.
Q: What sectors are driving European deeptech investment in 2026?
A: Defence, security, and resilience now account for 43% of all European deeptech funding, up from 20% in 2022. The Future of Compute segment saw funding more than double to $2.5 billion in 2025, while Novel AI attracted $3.4 billion.