Company
One human. A handful of AI agents. The same architecture we sell.
Corial is built by one human and a handful of AI agents working under his direction. It's the same approach we sell to our customers, applied to ourselves. If you want to see how a one-person company can ship like a ten-person company, this is the proof.
Why Corial exists
Ingredient sales cycles run 12 to 24 months across procurement, R&D, formulation, regulatory, production, and marketing. The context that decides whether a deal moves forward (who tested what, who pushed back on which spec, which competitor showed up at which stage) lives almost entirely in the heads of the people doing the selling.
Traditional CRMs are built for 30-day transactional cycles. They force this complexity into linear pipeline stages and depend on manual data entry that collapses within weeks. We've watched this happen at multinationals and at startups, with two different enterprise CRMs and several lightweight ones. The pattern is always the same: the data isn't there because nobody had time to type it in.
Corial exists to make that pattern stop. Voice notes, emails, and meeting transcripts go in. Structured pipeline data comes out. Nothing falls through the cracks across an 18-month qualification, because the AI agent is doing the bookkeeping a human would never have time for.
The team
Matthias Foerster
Founder · 20+ years in chemical raw material commercialization
20+ years in B2B chemical raw material commercialization. Grew from account manager through distributor accounts and global key account director into EMEA Business Director, running a 15+ person organization across sales managers, business development managers, and global key account directors. Later moved into biotech startups in the same industry: same customers, completely different speed and resources.
Built Corial because he lived the problem from both sides. As a rep, every enterprise CRM he used felt like a tax on selling. As a manager trying to make strategic decisions from those same systems, the issue was never reporting. The information was never in there to begin with, because nobody had time to type. The startup move made the gap worse: smaller teams, higher speed, bigger stakes, even less time for hygiene.
Every company in that arc has been a raw material supplier. A CRM that understands the raw material sales cycle (sampling vs. pilot, INCI naming, multi-department dynamics, regulatory gates, 18-month qualifications) has to be built by someone who has lived inside it. Not by a tech founder guessing at the problem.
Ada
Marketing & content · AI agent
Drafts blog posts, landing copy, and outreach in Matthias's voice. Runs the analytics dashboards, tracks competitor moves, and surfaces what's worth writing about. Trained on the ingredient industry context the rest of the product runs on.
Operates inside Claude Code with the same agentic architecture we ship to customers. Every output is reviewed before publish. Ada exists because the marketing of an AI sales agent should be done by an AI agent.
Hopper
Engineering & operations · AI agent
Writes code, runs tests, deploys to production, watches the logs, debugs incidents. Reviews its own pull requests against the CLAUDE.md spec the rest of the team works from. Code-reviewed by Matthias on every change.
Hopper never has autonomous access to customer data and never holds production credentials it does not need. The same security architecture documented in Corial's security spec applies internally: human-in-the-loop on anything that touches a customer.
Turing
Data security & infrastructure · AI agent
Watches the perimeter. Reviews every change that touches authentication, tenant isolation, secrets handling, or data egress. Runs continuous checks against the 6-layer security spec: prompt injection defense, multi-tenant isolation, AI safety, infrastructure hardening, audit trail integrity.
Customer data lives on European servers, isolated per tenant at the database layer. Turing makes sure it stays that way. Every customer-facing change requires Matthias's explicit approval before it ships.
Norman
User experience · AI agent
Tests every flow the way a sales rep would actually use it: in transit, with one hand, after a long day, while juggling six accounts. Catches the friction points before they become tickets. Pushes back on anything that adds steps a real seller would never do.
The principle is simple: a CRM that requires more than two minutes of attention after a meeting will not get used. Norman holds that line on every screen we ship.
What we believe
Domain depth beats generic AI
A CRM that does not understand the difference between sampling and pilot, or what an INCI name is, will lose to one that does.
Dogfood the product
If we would not bet our own company on an AI agent, we have no business asking yours to.
Honest about what is human and what is not
No fake team photos, no invented executives. The agents are agents and the human is one human.
Get in touch
Questions, feedback, or curious how a one-person company built this? Write to hello@corial.app. Matthias reads every message himself.
Ready to stop losing what your team knows?
Join the founding member program and lock in launch pricing.
Secure your founding spot