A San Francisco startup founded by a Nigerian-born engineer just secured $3.8 million to solve a problem haunting enterprise AI adoption: legacy desktop software that refuses to talk to modern systems. Minicor, co-founded by Saheed Akinbile, announced the seed round on October 14, 2026.
The funding arrives as frontier models grow increasingly capable at reading screens and manipulating interfaces. Yet hospitals, freight brokers, and manufacturers still run critical operations through Windows applications built decades ago, programs that expose nothing but a graphical interface. No API means no easy handshake with an AI agent.
Brad Gillespie of IA Ventures led the round. Other participants include Y Combinator, Paul Graham, Pioneer Fund, Robinhood Ventures, Garage Capital, Rebel Fund, K5 Global, and 20VC. The company operates from San Francisco, not Nigeria, though African tech outlets have amplified the story given Akinbile’s roots.
Minicor’s core bet: converting desktop workflows into callable API endpoints. Developers record or define a task inside an old application, and Minicor exposes that sequence as a standard interface. An AI agent then calls the endpoint, receives structured outputs, and never touches the screen directly.
This approach differs from traditional robotic process automation. RPA scripts replay clicks and keystrokes, breaking whenever screens change. Minicor claims its automations self-heal, using model-driven intelligence to detect interface shifts and adjust accordingly. That claim remains unverified from outside.
The Y Combinator narrative drove much of the coverage. Akinbile and co-founder Faizaan Chishtie gained admission to the Spring 2026 batch after seven applications and three interview invitations. Some headlines framed this as seven rejections. The reporting confirms the application count but does not state every attempt ended in refusal.
A usage figure of 2.5 million automation executions has circulated online. One source flags that number as independently unverified. No customer names, revenue figures, or valuation details appear in the confirmed reporting.
The market context matters. Investors continue pouring money into infrastructure layers around AI agents: security, identity, and now access. Infisical’s $40 million Series B for agent security charted one slice of that trend. Minicor targets another: the unglamorous plumbing connecting agents to software companies actually run.
Enterprise buyers face real questions. Where do credentials live? Who can trigger an endpoint? How do actions get logged? Regulated industries will demand audit trails before allowing agents near patient or account records. The announcement does not detail these controls.
Meanwhile, incumbent RPA vendors hold large customer bases and can bolt language models onto existing tools. Model labs can improve computer-use accuracy until middleware becomes unnecessary. Minicor must prove the middle layer pays for itself.
The next signal to watch: named enterprise deployments, not another investor on the cap table.
