Intercom opens Berlin AI R&D hub to scale Fin, hiring 100 and pushing toward $100M ARR
AI customer service firm Intercom is opening a new R&D hub in Berlin and hiring 100 people across engineering, AI, data science, product, and design. The investment centers on Fin - Intercom's AI agent that aims to handle end-to-end customer support and commercial flows - which is approaching $100 million in ARR.
Founded in Dublin in 2011, Intercom serves around 28,000 customers with strong bases in North America and Europe. Germany is now the company's third-largest and fastest-growing market among major European countries. While the global HQ is in San Francisco and R&D leadership sits in Dublin, Berlin is the next strategic anchor.
Why Berlin
"Berlin strikes us as a great city, providing a deep concentration of technical talent, universities that feed that, and a culturally rich, creative scene," said Darragh Curran, Intercom's CTO. He added that the city is emerging as a real hub for AI talent and matches Intercom's design-first culture. Proximity to Dublin and San Francisco helps collaboration while giving the hub its own momentum.
The hiring plan
Intercom is building a complete R&D unit in Berlin to accelerate Fin's roadmap and expand its scope beyond traditional support. The company plans to fill 100 roles over the next year.
- AI/ML researchers and engineers
- Data science and platform engineering
- Product management, product design, and UX
- Focus: advancing Fin as a true end-to-end Customer Agent
"We intend to build a holistically complete R&D unit in Berlin - everything from AI and machine learning researchers to product engineers, designers, and managers. A huge focus of our investment is on building Fin, our AI Customer Agent."
Fin's trajectory and what it means for support teams
Fin started in support, but the vision is broader: an agent that guides customers across the full experience - handling upgrades, refunds, and in-product help, not just tickets. Customers are already pushing it in that direction, according to Curran.
Performance has moved from roughly 23% to 66% resolution rate over two years. Only a small portion came from model upgrades; most gains came from Intercom's own architecture and continuous experimentation.
More than 7,000 businesses now use Fin, including Berlin's Ostrom and global names like Anthropic, Vanta, Clay, Whoop, Tado, and Miro. Fin has become a major growth driver as Intercom positions itself among the fastest-growing private software companies.
If you want to see how the product is evolving, check the official page for Fin here: Intercom Fin.
What support leaders should do now
- Define success with hard metrics: resolution rate, deflection, CSAT, AHT, containment by intent, and revenue actions (upgrades, add-ons).
- Map the full experience: account changes, billing, refunds, trials-to-paid, cancellations, and reactivation.
- Set trust boundaries: which intents the agent owns end-to-end, when to escalate, and how to keep context between bot and human.
- Invest in knowledge health: audited articles, process docs, and product FAQs built for retrieval - short, current, and unambiguous.
- Wire up systems: secure connections to billing, subscription, and account services so the agent can take action, not just answer.
- Create an experimentation cadence: weekly shipping, A/B tests by intent, and a backlog ranked by impact and effort.
- Plan governance: data retention, PII handling, audit trails, and a clear incident playbook. Account for EU compliance and localization needs.
- Model the ROI: cost-per-resolution, agent time saved, FCR improvements, and incremental revenue from self-serve upgrades.
Implementation checklist for AI agents (Fin or otherwise)
- Audit top intents by volume, value, and pain - start with the top 10.
- Baseline metrics: resolution, CSAT, AHT, containment, and recontact rate.
- Refactor content: one-intent-per-article, current screenshots, crisp steps.
- Access and actions: set up secure integrations for billing, account, and order systems.
- Guardrails: escalation triggers, compliance filters, and sensitive-topic rules.
- Sandbox: test with synthetic and anonymized real threads; score accuracy and safety.
- Pilot: limited audience, clear fallback to humans, daily review loop.
- Rollout: expand coverage by intent, not all at once; keep shipping improvements weekly.
Impact on roles and careers
Support roles are leaning into higher-skill work: AI operations, conversation design, content engineering, and systems thinking. Frontline agents become subject-matter experts and escalation pros. Managers shift from queue management to performance engineering of bots and teams.
As Curran put it: "If I were an engineer today, I'd want to be at a company doing something meaningful with AI. It's the best way to learn." The same holds for support pros who want to lead the next wave of customer experience.
Quick facts
- New Intercom AI R&D hub in Berlin; 100 hires planned.
- Fin approaching $100M ARR; now a core growth driver.
- Resolution rate improved from ~23% to ~66% over two years.
- 7,000+ businesses using Fin (including Anthropic, Vanta, Clay, Whoop, Tado, Miro, and Ostrom).
- 28,000 Intercom customers globally; Germany is the third-largest and fastest-growing market among major European countries.
- HQ in San Francisco; R&D leadership in Dublin; Berlin chosen for its talent density and creative energy.
Skill up your team
If you're building an AI-ready support org, curate learning paths by role (agent, team lead, ops, product) and track progress. For structured options by job function, explore this resource: AI courses by job.
Your membership also unlocks:
