Every patient discharge in a European hospital triggers a paperwork cascade. Clinical information must be converted into standardized ICD codes that determine what the hospital gets paid. A trained medical information specialist spends most of their working day on exactly this — navigating legacy software, pulling records, selecting codes, entering them correctly.

Parallel, a Paris-based startup, is building AI agents to do it instead. And they just raised €20 million Series A led by Index Ventures to do it across Europe.

The Approach That Changes Everything

Parallel’s technical bet is counterintuitive: don’t integrate with hospital software at all.

Traditional healthcare IT integration takes 12-24 months and frequently fails. Hospital systems are old, fragmented, and built on proprietary stacks that resist API-based automation. Most healthtech startups crash against this wall.

Parallel’s agents operate at the UI layer — the same way a human would:

  • Reading screens to understand what’s displayed
  • Clicking through interfaces to navigate workflows
  • Entering data into existing systems
  • Learning the navigation patterns of each hospital’s specific software

The result: deployment in approximately one week instead of 12-24 months.

This is the computer-use agent pattern applied to one of the hardest integration environments on earth. While GPT-5.4 demonstrated 75% on OSWorld benchmarks for general-purpose computer use, Parallel is proving that domain-specific UI agents can be commercially viable now — in a regulated, mission-critical environment.

Why Healthcare Admin Is the Perfect Target

25-30% of healthcare spending goes to administration. That’s not a rounding error — it’s a structural problem that’s been resistant to automation precisely because the software stack won’t cooperate.

Parallel chose to start with medical coding in French public hospitals — specifically the PMSI (Programme de Médicalisation des Systèmes d’Information) framework. This is high-value, high-complexity work:

  • High stakes — coding errors directly affect hospital revenue
  • High volume — every discharge triggers the workflow
  • High skill — requires trained specialists who are scarce
  • High tedium — mostly manual data entry across legacy systems

Starting in the hardest vertical and working outward is a stronger signal than building a general-purpose agent and hoping it works in healthcare.

The Funding and Team

DetailValue
Round€20M Series A
LeadIndex Ventures
ParticipantsFrst, Y Combinator, Hexa
Angel investorsArthur Mensch (CEO, Mistral AI), Felix Blossier & Quentin de Metz (Pennylane)
Seed$3.5M (April 2025)
Total raised~$24M

The team combines healthcare domain expertise with operational scale experience:

  • CEO Paul Lafforgue — École Polytechnique, HEC, ex-Meta, ex-McKinsey
  • CTO Christopher Rydahl — co-founded Hublo (healthcare staffing, 2,800+ facilities, €22M raised)

Previously known as Kiosk Medical, Parallel entered Y Combinator in 2024 and has been deploying in dozens of French public and private hospitals.

The Expansion Plan

The €20M funds three priorities:

  1. Scale existing coding agents across more French hospitals
  2. Expand internationally — Netherlands, Belgium, and other European markets with similar hospital administrative structures
  3. Build new agent types — billing, admissions, and other administrative workflows

The European healthcare market is Parallel’s structural advantage. Public healthcare systems running on older infrastructure with fewer integration points are exactly where UI-layer agents shine — you don’t need APIs when you can operate the software directly.

Where This Fits in the Healthcare Agent Landscape

The healthcare AI agent market is heating up fast. At HIMSS 2026, we saw:

  • Epic Agent Factory — 50+ pre-built clinical agents, EHR-native
  • Oracle Health — 30 specialty-specific agents, clinical documentation
  • Nuance Dragon Copilot — ambient AI for clinical workflows
  • AWS Connect Health — first HIPAA-eligible agent platform

Most of these target clinical workflows and require deep EHR integration. Parallel targets administrative workflows and requires zero integration. Different problem, different approach, different deployment timeline.

The closest comparison is Axiory’s MCP trading layer — taking the agent concept into a specific regulated vertical where speed-to-production matters more than feature breadth.

The Computer-Use Agent Pattern

Parallel’s success validates a broader thesis: UI-layer agents may be more commercially viable than API-integrated agents in environments with legacy software.

The tradeoffs are real:

FactorUI-Layer AgentAPI-Integrated Agent
Deployment speed~1 week12-24 months
ReliabilityDepends on UI stabilityDepends on API stability
Depth of integrationSurface-levelDeep system access
MaintenanceUI changes break agentsAPI changes break agents
CostLower upfrontHigher upfront, lower ongoing

For healthcare, where the integration timeline kills most projects, the UI-layer approach isn’t a compromise — it’s the only practical path.

What This Signals

  1. Vertical AI agents are where the money flows — €20M for hospital admin automation, not another general-purpose agent platform
  2. Computer-use agents are production-ready — not just benchmarks, but live in dozens of hospitals
  3. Europe is building its own AI agent ecosystem — Parallel (France), Mistral (France backing), Index Ventures (Europe) — this isn’t Silicon Valley
  4. Legacy software isn’t a blocker anymore — UI-layer agents turn “impossible to integrate” into “one-week deployment”

The next test: whether the one-week deployment and near-term accuracy claims hold up as Parallel scales across countries with different healthcare systems, different coding frameworks, and different legacy stacks. Index Ventures is betting €20M that they will.