Sacra estimates that Invisible hit $134M in annualized revenue at the end of 2024, up 123% from $60M in 2023. This represents a moderation from the company's hypergrowth phase when it achieved 300% annual growth in both 2021-2022 and 2022-2023, scaling from $3.75M in 2021 to $15M in 2022 and then to $60M in 2023.

Invisible's recent growth has been propelled by significant contracts in the AI space, particularly for Reinforcement Learning with Human Feedback (RLHF) services. A pivotal moment came in 2022 when OpenAI engaged Invisible to help fine-tune their models, leading to additional contracts with Amazon, Microsoft, Cohere, and other AI labs.

Prior to this AI-driven expansion, Invisible had a breakthrough with DoorDash during the COVID-19 pandemic in 2020, helping them rapidly onboard restaurant menu data when traditional outsourcing firms were disrupted by lockdowns.

With profitability of approximately $15M EBITDA (11% margin) on its 2024 revenue, Invisible has established itself as a financially sustainable player in the growing AI services sector.

Invisible was valued at approximately $500 million as of 2023, representing a 3.7x multiple on its $134M ARR.

The company has taken an unconventional funding path, raising approximately $23 million in equity funding through 2020. Early investors included Backed Ventures (leading the $2.6M seed round in 2018), Day One Ventures, Greycroft, Horizons Alpha, and Loup Ventures, along with notable angels like Mark Pincus.

Invisible combines human workers with AI to handle virtually any repetitive or complex task that companies need to outsource. Think of it like having a team of assistants working in the background that you can delegate any tedious work to—whether it's research, data entry, content moderation, or AI model training.

When a client needs something done, they simply email a special address or use Invisible's dashboard. Invisible's platform automatically breaks that request into subtasks. For scheduling a meeting, one agent might find calendar openings while another compiles a briefing document. The platform assigns each piece to trained human workers (called "agents") and uses software bots for repetitive parts.

What makes this different from traditional outsourcing is Invisible's digital assembly line approach. They break processes into micro-tasks handled by different specialists, then stitch everything back together. The client receives a single, seamless result as if one person did everything, when in reality a coordinated team augmented by AI tools completed the work. Clients can track the status and cost of each job through a dashboard—for instance, seeing that 15 expense reports were filed at around $1.83 each.

For AI companies, Invisible's workers provide human feedback on model outputs, rating responses for accuracy, helpfulness, and safety—crucial for improving AI systems through reinforcement learning. An AI researcher might send thousands of model outputs to Invisible, whose trained specialists evaluate each one according to specific criteria, helping the AI learn from human preferences.

Invisible operates as a B2B tech-enabled service company bridging the gap between pure software automation and traditional outsourcing. Its value delivery mechanism functions as "Operations-as-a-Service"—clients hand off cumbersome work processes, and Invisible executes them by combining human labor with automation technology.

The go-to-market approach targets companies experiencing scaling challenges in their operations or AI initiatives. While Invisible initially served some individual professionals, its core focus shifted to businesses needing help with labor-intensive processes or AI training data. Its client base spans from tech startups to large tech firms and traditional industries seeking back-office automation.

Monetization follows a usage-based model, with clients typically paying for outcomes rather than hours. For individual executive support, Invisible set a minimum spend of $2,000/month, positioning itself as a premium service. For corporate clients, deals are structured around defined processes—for example, a rate per 1,000 annotations for AI training, or a monthly retainer for handling customer onboarding tasks.

The operational structure relies on a globally distributed workforce of 3,000+ agents in 35+ countries, organized by specialization and seniority. The proprietary "Workplace" platform acts as the work distribution engine, coordinating assignments and monitoring quality. This creates a high variable cost component, with gross margins determined by the markup over wages plus efficiencies gained through automation. The business achieved 11% EBITDA profit margins in 2024.

In the AI training sphere, Invisible competes directly with companies like Scale AI and Surge AI. Scale AI started with a focus on providing training data for self-driving cars and has grown into a heavyweight platform for AI dataset generation and model alignment.

Scale offers a highly automated platform where clients can upload data or connect via API, and Scale handles the labeling with ML assistance and human reviewers.

Surge AI is a newer startup focused specifically on NLP data labeling and RLHF. It has made a name by emphasizing elite labelers and modern tooling, reportedly displacing Scale in some high-profile cases like Anthropic's RLHF work.

Invisible also faces competition from Business Process Outsourcing (BPO) firms like Accenture, TaskUs, and Teleperformance. These large outsourcing companies handle everything from call centers to back-office processing for enterprises and could potentially take on workflows that Invisible targets.

The difference is that these BPOs often lack specialized AI tooling and may not handle highly dynamic tasks as effectively. They typically operate on multi-year contracts with teams dedicated to specific clients. Invisible differentiates by offering fractional, on-demand teams that can scale up or down rapidly, and by using its software to integrate with client systems in a more fine-grained way.

A third competitive category includes veteran data annotation firms like Sama (formerly Samasource) and newer specialized players. Sama built large annotation centers in East Africa to provide training data to Silicon Valley companies, with an emphasis on ethical treatment of workers. They were notably involved in providing content moderation for OpenAI's GPT-3 model.

Other specialized players include Appen, Lionbridge AI (now part of TELUS International), iMerit, Toloka, and Prolific. Each has carved out a niche—for example, iMerit focuses on image and geospatial data labeling, while Prolific provides participants for surveys and research studies.

With the explosion of interest in custom large language models, Invisible has significant growth potential as a partner for companies wanting to train or fine-tune AI models but lacking in-house resources. This market is expanding rapidly as organizations across sectors explore customizing LLMs for their needs.

Invisible has already moved beyond basic data labeling to offer end-to-end support for model development, including data sourcing, fine-tuning with human feedback, model evaluation, and deployment monitoring. By becoming an "ML Ops plus humans" platform, Invisible can embed itself in the AI development lifecycle for companies without large machine learning teams.

This significantly expands Invisible's addressable market beyond data labeling into model performance management—essentially providing "Model Alignment as a Service." If enterprises increasingly train or customize their own models, each will need ongoing human feedback loops for alignment, safety, and domain adaptation, creating a massive opportunity for Invisible's managed service approach.

While Invisible has taken a horizontal approach serving many functions across industries, there are substantial opportunities to develop industry-specific solutions. For instance, in healthcare, there's growing demand for human curation and verification of training data for medical AI applications that must comply with regulations like HIPAA.

By building vertical-specific workflows and training agents with specialized backgrounds, Invisible could enter markets like healthcare AI operations, financial services (for tasks like KYC checks and fraud monitoring), legal document processing, and other regulated industries. The company could develop "vertical templates"—packaged solutions for specific industries like e-commerce catalog management or legal document summarization.

Each vertical successfully entered adds to Invisible's total addressable market. The $100+ billion global BPO industry represents the potential TAM for replacing traditional outsourcing if Invisible's technology-enhanced model gains widespread adoption across business functions like HR, finance, IT operations, and customer service.

As large language models become more capable, Invisible can expand by orchestrating both AI agents and human agents working in tandem. The company is already exploring "chain of thought & agentic AI training," suggesting research in how AI can handle multi-step reasoning tasks with human guidance.

This presents an opportunity to offer AI-augmented project teams for complex knowledge work. For example, a client could request market research, and Invisible would deploy an LLM agent to draft an outline and gather information, have humans verify sources and fill gaps, then use AI to format the final output—all managed through its platform.

By creating standardized offerings that combine AI automation with human oversight, Invisible could charge premium rates for these complex, partially-automated projects. This approach would capitalize on the latest AI capabilities to reduce human effort while maintaining humans for quality control and creative judgment.

This expansion would deepen monetization by allowing Invisible to charge not just for human hours but for its proprietary workflows and AI tools. The more Invisible's technology contributes to delivering results, the more it can justify software-like margins on that portion of the business.

Quality control challenges: As Invisible's workforce grows beyond 3,000 agents to meet increased demand, maintaining consistent quality becomes increasingly difficult. Any serious quality issues in high-profile projects like AI model training could damage Invisible's reputation and lead to lost contracts, especially since clients like OpenAI or Microsoft would quickly drop a vendor if data quality became questionable.

AI disintermediation threat: As AI models improve, they might require less human feedback or simpler input, potentially reducing demand for Invisible's human-centered services. If techniques like simulated feedback or AI-to-AI reinforcement learning advance, companies like Invisible could see their cutting-edge services become less critical, forcing continual adaptation to find areas where human judgment still adds measurable value.

Labor model scrutiny: Invisible's business benefits from global wage differences, paying workers in emerging markets approximately $5-10 per hour. As scrutiny of digital labor practices increases, Invisible could face regulatory challenges in worker classification, minimum wage requirements, or public perception issues if viewed as exploitative rather than empowering for its global workforce.