Inside Legora's Tech Stack: Why Token Maxing is Failing Enterprise Startups | Legora CTO

Financial & Operational Milestones:

• Legora reached $100 million ARR in 18 months.
• Jacob Loretz, CTO, predicts Year-End revenue will exceed $272 million.
• The engineering headcount has grown from an underestimated 20 to approximately 80, with a forecast of 270 engineers by December 2027.
• Cloud Code and Cursor are currently the only AI coding tools used by Legora's engineering team, accounting for over 50% of all code generation between them.

Engineering Strategy & AI Tooling Economics:

• Loretz views AI tooling spend as an investment in opportunity cost rather than a direct cost, stating the cost of not adopting AI outweighs token expenses.
• The primary engineering bottleneck has shifted from code writing to product definition and code review efficiency.
• AI code review is in a nascent phase; current tools lack the capability to assess systemic architectural impact and security boundaries, necessitating human oversight for high-stakes PRs.
• The future engineer role is shifting from typing code to defining system architecture and acting as "meta-engineers" who design and optimize agent loops.
• Legora is building internal "guardrails" and custom rules to mechanically enforce system behavior for autonomous agents, moving away from unrestricted agent operation.

Process Evolution & Organizational Structure:

• Product Management (PM) roles are being redefined; PMs should focus on customer synthesis and strategic prioritization rather than coding, as the opportunity cost of them coding is too high.
• PMs utilize "vibe coding" to create high-fidelity prototypes to reduce handover friction, but the core product work remains their primary value.
• Post-mortems are now automated via SRE and incident agents that analyze logs and telemetry to generate reports, drastically reducing response time.
• A dedicated Developer Experience (DX) team of three engineers exists to build custom linting, onboarding tools, and concurrent agent setups, though Loretz admits this team should have been staffed earlier.
• The company rejects remote work for core engineering teams to eliminate "handover costs" and maintain the speed of in-person collaboration.

Security & Risk Management:

• Loretz acknowledges a significant threat of AI-generated code vulnerabilities, citing recent vendor security incidents.
• Despite high AI adoption, all human Pull Requests are still reviewed to ensure security, a practice Loretz deems necessary until AI review tools mature.
• The company is experimenting with "risk scoring" systems to automate security checks and reduce manual review bottlenecks.
• Loretz expresses concern over a potential global monopoly on AI models, advocating for strong European and American open-source alternatives for sovereignty.

Product Philosophy & Market Position:

• "Taste" (opinionated design and stance) is identified as the primary differentiator against AI-generated "slop" and copycat competitors.
• Legora focuses on the "90% of cases" (edge cases, audit logs, RBAC, complex workflows) that are difficult for AI to replicate, rather than the easy 10% of functionality.
• The company builds custom internal HR, payroll, and talent acquisition systems via "vibe coding" when customization costs outweigh off-the-shelf pricing, specifically for shallow, highly customized systems.
• Loretz advises against "token maxing" in corporate incentives, suggesting rewards should be tied to efficiency and output rather than raw token consumption.

Hiring & Talent Management:

• The most difficult hiring challenge is finding senior technical management who can also manage; many technical leaders are no longer willing to manage people.
• Legora prioritizes "A-players" with low ego over filling headcount targets, using rapid feedback loops (strong feedback within two weeks) to manage underperformers.
• Loretz admits underestimating the need for hiring speed, shifting from a "300 Spartans" mentality to a more aggressive scaling approach.
• The company values equity over titles during hiring, though it faces challenges educating European candidates on the value of venture equity compared to US candidates.
• Acqui-hiring small, high-performing teams (5-8 people) is preferred over individual hiring to accelerate growth, with integration relying on the low-ego nature of the hires.

Future Outlook & Speculation:

• Loretz predicts the IDE will evolve away from line-by-line coding toward graphical system architecture interfaces for planning and review.
• He foresees a new enterprise role focused entirely on building internal AI systems, transforming IT from infrastructure support to efficiency engineering.
• Open-source models are expected to play a massive role in on-device processing and sovereignty, potentially challenging the dominance of closed models.
• Loretz believes lawyers will eventually operate one layer above contract language, focusing on negotiation stance and risk assessment rather than drafting text.
• The biggest threat to Legora is not direct competitors like Harvey, but its own inability to constantly reinvent itself and adapt to the rapidly shifting AI landscape.