Felicis Leads $18M Series A for Magentic’s Industrial AI Agents
Fundraise Insider tracks newly funded startups each week and delivers verified sales leads of C-suite contacts straight to your inbox, so you reach the right people while the funding is still fresh. See pricing.
Magentic, which builds AI agents that do operational work inside large industrial companies, has raised $18 million. Felicis led the Series A, with existing investors Sequoia Capital and The Westly Group taking part again. The round lands about a year after the company launched with a $5.5 million seed, and total reported funding now sits at $23.5 million.
The company is based in London and New York City. CEO Robin Van Aeken came from McKinsey, and CTO Odhran O’Donoghue is an OpenAI alumnus. Their bet is that enterprise AI is moving past assistants that answer questions or draft content, toward agents that carry out work across the systems manufacturers already run.
The target is procurement and supply chain operations, which get messy fast inside global manufacturers. Purchasing data tends to be spread across ERP platforms, supplier agreements, spreadsheets, invoices, email, and databases inherited through years of acquisitions. Pressure on those teams is rising too. The Hackett Group projects procurement workloads will grow 8% in 2026 while headcount and budgets shrink. Meanwhile Goldman Sachs puts cumulative AI-related capital expenditure at roughly $7.6 trillion between 2026 and 2031, which translates into a huge volume of equipment, materials, construction, and supplier contracts to manage.
Magentic calls its agents digital workers, and the distinction that matters is autonomy. Instead of another dashboard or chat window, the platform sits on top of existing enterprise infrastructure. Its architecture covers data ingestion, a harmonized data layer, an agentic engine, procurement-specific knowledge, real-time monitoring, feedback loops, and end-to-end execution. It pulls fragmented information from ERPs, spreadsheets, PDFs, and contracts into a structure agents can use, and it can pick different underlying AI models depending on the task.
In practice an agent might notice the same component being bought at different prices across factories, find overlapping suppliers to consolidate, catch purchases made outside negotiated contracts, or surface rebates that were never collected. The agents then work the resulting process: reviewing spend, examining contracts, calculating savings, drafting supplier communications, reconciling orders against invoices, and assembling the evidence needed to recover overpayments. They operate across spend, sourcing, contracts, and invoices rather than one isolated task.
Rather than cleaning and standardizing data before deployment, as traditional transformation projects do, Magentic lets its agents work across the fragmented environment as it stands and organize the information a task requires as they go. The agents communicate through tools employees already use, including Microsoft Teams and email.
The company says one customer is now processing more than 1.2 million orders through its digital workers, and another deployment identified $4 million in savings. Across its Global 500 customer base, it reports savings of roughly 2% to 5% and an average 60% improvement in data quality. These are company-reported figures, not independently audited ones.
Because a wrong procurement decision has immediate financial consequences, human review points are built into the workflows. An agent can scan supplier contracts and flag unusual payment terms, but a procurement executive still decides whether to open a negotiation. Agents can also check one another’s work, and logs and evidence trails show how decisions were reached. Magentic is SOC 2 Type II and ISO 27001 certified, GDPR compliant, and aligned with the EU AI Act. Customer data is not used to train shared or public models.
The Series A money will widen the set of procurement and supply chain workflows the agents can handle and fund further AI research. One technical challenge is context: industrial optimization problems can span years of purchasing data, thousands of contracts, pricing histories, inventory records, invoices, and technical documents. Magentic says it is building agents that can diagnose problems, plan fixes, take action, and follow the work through across terabytes of multimodal data.
Software companies selling into enterprise operations teams can get verified contacts at newly funded B2B software startups every week through B2B SaaS Leads.