EN
Back to the archive

The archive · AI & Models · Strategic decision · 2024–2026

Conduct bets AI agents can run opaque enterprise systems; SAP joins $60M Series A

Ex-Palantir founders map decades of customised ERP code into business logic so agents can act; SAP invested and made it a strategic partner.

Conduct

The betThat decades of customised enterprise code is the bottleneck, and making systems legible and operable for agents is infrastructure worth buying first.Scaling

What the business is

Conduct builds an AI operating system that ingests custom code, configuration, dependencies and integrations across SAP, Salesforce, Oracle, MES and WMS systems and maps them to the business logic they serve, then generates the code and tests needed to change them.

Starting capital$60M Series A co-led by Index Ventures and ICONIQ Growth, with strategic investment from SAP and participation from Creandum, Lucid Capital and Booom.

How it started

Former Palantir engineers Jan Philipp Haas, Philipp Hoefer and Henry Thompson founded Conduct in London in 2024 after watching large enterprises take weeks or months to change systems so customised they had become opaque even to the people running them.

What happened

The platform ingested custom code, configuration, dependencies and integrations, mapping technical components to the business logic they serve. Conduct signed customers including Daimler Truck, Heidelberg Materials, Fraport and DHL, and reported that teams saw 30% or more acceleration in transformation workstreams. SAP named Conduct a strategic AI partner for transformation with SAP Cloud ERP applications, invested in the Series A, and Conduct added partnerships with BCG and NTT DATA Business Solutions.

How it ended up

Still scaling: the June 2026 Series A funds expansion of engineering and go-to-market teams, deepening SAP capabilities and extending work across Salesforce, Oracle, MES and WMS systems.

Background

Conduct, founded in London in 2024 by three former Palantir engineers, bets that the real barrier to enterprise AI is not model capability but system opacity. Decades of customisation have left core systems like SAP so hard to comprehend that humans cannot safely change them and agents cannot act on them at all.

The company maps custom code, configuration, dependencies and integrations to the business logic they serve, letting teams ask what depends on an approval workflow or what breaks if a field changes, then generates the code, tests and implementation work for the change. Customers including Daimler Truck, Heidelberg Materials, Fraport and DHL used it, with teams reporting 30% or more acceleration in transformation workstreams.

In June 2026 Conduct raised a $60M Series A co-led by Index Ventures and ICONIQ, with SAP investing strategically and naming Conduct an AI partner for SAP Cloud ERP transformation. The round validates the bet that making legacy systems legible is infrastructure, not a feature: SAP's own investment signals that even the ERP incumbent sees the mapping layer as necessary.

What has to be true

  • The bet is contrarian and concrete: enterprises should not rip out or replace legacy ERP but make it legible enough for AI agents to operate.
  • A dated arc with sources: founded 2024, named customers by 2026, SAP partnership and $60M Index-ICONIQ Series A in June 2026.
  • SAP investing in and naming Conduct a strategic partner is unusually strong validation for a two-year-old startup.
  • Specific reported outcomes are citable: 30%+ acceleration in transformation workstreams and named Fortune 500-class customers.

What can be applied

The unglamorous bottleneck is the market: whoever makes a decade of customised enterprise code legible becomes the layer that both AI agents and transformation consultants depend on.

Aftermath

As of 2026-09-02 Conduct is live with enterprise customers and scaling: the Series A is funding engineering and go-to-market growth, deeper SAP Cloud ERP capabilities, and expansion across Salesforce, Oracle, MES and WMS systems. The test ahead is whether the mapping layer generalises beyond SAP-heavy workloads and becomes the default interface agents use to operate enterprise systems, rather than a niche transformation tool.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases