The Future of Digital Workers

Company Philosophy & Definition of Agents

• Level Next prioritizes customer outcomes over technology; the company will avoid using agents unless they provide superior results.
• Prabhav defines a "true agent" as a system capable of planning, reasoning, reflecting, and learning over time, specifically for problems where even humans lack a clear answer.
• The company distinguishes between "agentic" tasks (e.g., deep research, content writing) and "orchestrated" tasks (e.g., sending emails), applying the most effective technology for each specific step.

Product Portfolio: Alice and Mike

• Alice: An AI SDR product designed for outbound revenue teams that leverages first-party CRM data and third-party data to generate qualified meetings and pipeline across multiple channels.
• Mike: A voice-based agent handling inbound and consented outbound use cases, operating 24/7 across multiple languages with deep CRM and calendaring integrations.
• Strategic Pivot: The company recently underwent a total re-architecture of both products to transition from basic prompting to fully agentic frameworks, a decision driven by the realization that the new architecture would deliver significantly more customer value.
• Re-architecture Execution: During the transition, the majority of the team built the new product while a small "life support" team maintained the legacy product, successfully migrating customers with minimal disruption.
• Go-to-Market Strategy: Unlike competitors, Level Next provides free access to the new product during the sales process to validate performance before requiring a commitment.

Technical Architecture & Innovation

• Voice Technology Evolution: The team shifted from assembling separate text-to-speech and speech-to-text models to adopting end-to-end "voice-to-voice" models to reduce latency and improve naturalness.
• Vendor Agnosticism: The platform is built to rapidly swap between model providers (including recent DeepSeek releases) based on task-specific performance metrics like tool-calling or instruction following.
• Evaluation Strategy: To address subjective quality metrics, the company employs a "human-in-the-loop" process, leveraging historical data on open rates, replies, and meetings to empirically determine what performs best for specific customer segments.
• Infrastructure: The system consolidates 20-30+ disparate sales tools into a unified stack, managing complex integrations for web crawling, hosting, and data warehousing to present a seamless experience to the end user.

Operational Decisions & Market Positioning

• Multi-Product Strategy: Level Next launched two distinct products early to support anti-fragile revenue motions (inbound, outbound, voice, email) rather than adhering to a single-solution path.
• Team Structure: Products are developed by small, zero-to-one "pods" tasked with rapidly achieving Product-Market Fit (PMF) before the platform is standardized for scale.
• Customer Education: The company actively educates clients that their subjective perception of a "good" lead or email may not correlate with actual performance data, guiding them to trust the system's data-driven optimizations.
• Scope Control: While listening to customer pain points, the company rejects custom feature requests that would shift the product from a "digital worker" model to a traditional SaaS customization model.

Reflections on Past Decisions (Retroactive Analysis)

• Design Integration: The founder notes that hiring a lead designer earlier would have accelerated development, as user interface design significantly impacts the user's perception of the agent's invisible work.
• Geographic Relocation: Moving the company from London to San Francisco earlier would have accelerated growth by tapping into the local "mecca" of AI agent development and collective intelligence.
• Product Opinionation: The company would have moved faster to provide specific, opinionated playbooks for success rather than offering a blank slate, helping to set clearer customer expectations for outcomes.

Hiring Criteria & Talent Strategy

• Velocity: The primary hiring filter is the ability to move quickly in response to rapidly shifting market conditions and model releases.
• Ownership: Candidates must demonstrate a "die for the customer" mindset, treating customer success as the sole determinant of the company's success.
• Chaos Tolerance: The team seeks individuals who can handle constant pivots, such as scrapping weeks of work overnight if a new model unlocks better outcomes.
• Leadership Profile: Leaders are expected to balance high-level vision with hands-on execution, a trait common among the many former founders on the current leadership team.

Future Outlook

• Trust Mechanisms: Future product updates will focus on "taking the user along for the journey," visualizing the step-by-step reasoning and data gathering of agents to build transparency.
• Continuous Iteration: The company plans to maintain a weekly rhythm of alternating investments between new feature velocity and system scaling reliability to manage the dual pressures of growth and technical debt.
• Adaptive Learning: The ecosystem relies on a distributed network of vendors and beta testers to continuously experiment with new tools, ensuring the platform remains on the cutting edge of AI infrastructure.