Fractal Analytics is an enterprise AI company that helps large organisations make better decisions - thousands of them every day - inside the messy, fragmented systems where data lives and small mistakes are expensive. Founded in 2000 as a private analytics firm and now publicly listed, it has grown into a pure-play AI partner trusted by over 100 Fortune 500-sized enterprises, with a vision to power every human decision in the enterprise. The recurring approach that explains everything else is this: Fractal builds a common platform layer of AI agents that can plan, coordinate and execute work, then uses that platform to deliver services, products and workforce transformation - turning a service business into something closer to a software business where work compounds.
Business segments
Three go-to-market pillars sit on one shared AI platform, with a small but fast-growing set of IP-led businesses alongside - a services-and-software mix where the platform makes every engagement feed the next.
1. AI-led Transformation: redesigning how enterprises work
This is the process dimension - Fractal reimagines entire business workflows with AI to drive measurable outcomes, anchored in deep industry expertise across consumer goods, telecom, banking and healthcare.
- Revenue growth, not just cost-cutting - the practice helps companies understand demand, optimise pricing and promotions, and grow categories using AI, targeting business and functional CXOs rather than just the IT department.
- Supply chains that think - one engagement built an agentic AI engineering platform for a global consumer goods major that cut data pipeline build time by up to 75%, from a full day to roughly two hours, used by hundreds of data engineers.
- Seconds instead of days - for a pharmaceutical major, Fractal built an agentic platform that accelerated accurate clinical response from 3-4 days to seconds, enabling real-time engagement with healthcare professionals.
- Machines that self-optimise - in work with PepsiCo, AI was brought into manufacturing for smart packaging, where the system adjusts packs per minute automatically without human intervention.
2. AI Foundations: the technical backbone for AI at scale
This is the layer underneath everything - the data, agents and controls that enterprises need before AI can work reliably, targeting CIOs, CTOs and chief AI officers.
- Enterprise data as a moat - Fractal builds and trains models inside clients' own data systems, creating ontological layers of knowledge on top of frontier models, which the company considers a defensible competitive advantage because it is hard for model providers to replicate.
- Smaller, specialised models - LLM Studio helps organisations build and run domain-specific language models using open-source models, with one module for creating compact, specialised models and another for managing their full life cycle after creation.
- Governance built in - the pillar includes responsible AI, tech stack migration and golden datasets, so that AI at scale does not become AI out of control.
3. AI for Work and Workforce Transformation: the people dimension
This is about hiring, training, evaluating and coaching in a world where people and AI agents work side by side, targeting CHROs and business leaders.
- Eat your own cooking - the company uses its own AI-powered hiring assessment platform, Iqigai.ai, and runs AI coding agents and talent allocation software internally, so the workforce practice is tested on Fractal's own 5,868 employees first.
- Government-scale training - Fractal's course "AI for Karmayogis" on the Government of India's iGOT platform crossed 25 lakh completions.
4. Fractal Alpha: IP-led businesses with recurring revenue
A small but fast-growing segment of platform-like businesses with annualised recurring revenue models, distinct from the core services pillars - losses have narrowed sharply as they scale.
- Asper.ai: an operating system for consumer goods commercial planning - this AI platform for revenue growth management serves about 15 CPG and FMCG clients, with annualised recurring revenue that has grown roughly 77% year-over-year for three years running.
- Analytics Vidhya: community meets enterprise - a data science community and enterprise training business that grew 49% in the last fiscal year, serving over 15 clients including healthcare companies and banks similar to Fractal's largest customers.
- AutoGrader: AI that evaluates learners - a multi-agentic system that assesses code and project outcomes from learner submissions, delivering real-time feedback; it has evaluated over 18,000 submissions and is deployed with several clients.
Cogentiq - the shared platform layer
Every pillar and product runs on Cogentiq, Fractal's agentic AI platform that allows AI agents to plan tasks, coordinate with one another, access enterprise systems and execute work reliably - turning each client engagement into reusable platform capability.
- One platform, one Fractal - Cogentiq underpins all three go-to-market pillars, so the client sees a single company rather than fragmented offerings, and every engagement builds the platform further.
- Always-on e-commerce brain - Cogentiq e-commerce, launched in mid-2026, monitors over 70 profit-driving marketplace signals daily for each product SKU and recommends corrective actions across customer, media and supply chain functions within minutes, going live in as little as two days using only a marketplace API.
- Agents that actually do the work - Flyfish.ai, built on the platform, deploys 35 coordinated AI agents across the B2B sales lifecycle to research accounts, draft outreach and manage the pipeline; unlike tools that suggest what a salesperson should do, Flyfish does it, helping early adopters close deals up to 30% faster and improve sales productivity by 42%.
AI Research and proprietary models
A dedicated research team spending 6-8% of revenue on foundational AI advances - producing healthcare models, reasoning systems and agentic frameworks that feed the platform and products.
- Vaidya 2.0: a healthcare model that reasons clinically - the world's first model to cross 50 on OpenAI's HealthBench Hard, one of the toughest clinical reasoning benchmarks, available on all major mobile platforms with features including report interpretation from PDFs and images, a symptom checker and medication reminders.
- PiEvolve: AI that learns from failure - an evolutionary agentic framework that transforms one-shot machine learning into continuous optimisation, iterating until it exhausts its computational budget; it ranks among the top performing agents on OpenAI's MLE-Bench Benchmark.
- Open-sourced reasoning - Fathom-R1-14B is an open-sourced Large Reasoning Model from the research team, alongside launches including MarshallGoldsmith.ai and Kalaido.ai.
- 68 patents and counting - the company held 68 patents granted and applied for as of March 2026, with R&D investment running at INR 212 crore in the latest year, 48% higher than the year before.
Group structure and partners
A publicly listed Indian company with global operations across the Americas, Europe and APAC, a simplified subsidiary structure, and deep partnerships with hyperscalers and frontier AI labs.
- Global footprint, one listed entity - Fractal Analytics Limited employs over 5,000 professionals across the United States, Canada, the UK, the Netherlands, Ukraine, India, Singapore, South Africa, the UAE and Australia.
- Structure simplified - in March 2026, the company reduced the equity share capital of subsidiary Senseforth AI Research, extinguishing shares as part of reducing the number of legal entities in the group.
- Spin-out success - Fractal spun out Qure.ai, a global healthcare AI leader whose lung nodule malignancy score was evaluated across 23 healthcare centres in five countries and which was named to the TIME100 Most Influential Companies 2025.
- Hyperscaler partnerships - the company partners with frontier AI labs, hyperscalers, data platforms and enterprise software providers, won Microsoft Partner of the Year 2025 in Retail and Consumer Goods, and secured preferred supplier status with two of the "magnificent seven" technology companies.
9:46
13:31
8:21
8:12
12:40