Purple Fabric
Purple Fabric is an enterprise AI platform that banks, insurers and wealth managers use to build, govern and run AI agents on their own policies, documents and records. Intellect Design Arena launched it in 2024 and made it a separate line of business in June 2026, led by Deepak Dastrala. Intellect counts 45+ regulated institutions running it in production and offers an accuracy service level of 95% or better.
The Enterprise Knowledge Garden organizes an institution's knowledge for retrieval. Enterprise Digital Experts are agents that combine models and tools to do a job, such as preparing a credit memorandum or investigating a complaint. The LLM Optimization Hub lets a team pick and benchmark the model for each task, and Enterprise Governance sets access rules, approvals and a trace of the sources behind each output. Ready made suites cover lending (PF Credit), wealth (PF Wealth), trade finance (PF Trade), insurance submission intake, and risk and compliance.
A bank can run the software in its own cloud, on a dedicated tenant on Intellect's cloud, or in a hybrid setup that reaches records kept on premises.
AU Small Finance Bank chose Purple Fabric as its enterprise AI foundation, starting with PF Credit, and Intellect names Norges Bank Investment Management and St. James's Place as references. At an unnamed European financial institution, a complaint investigation workflow cut resolution time by 65% across more than 700 active cases a week.
In India, the Enterprise AI on Tap package costs ₹99,500 a month for up to 50 users.
Capability Axes
Capability grades
15 of 15 axes rated · 12 graded A or B
Enterprise Digital Experts, agents that coordinate language models and tools, carry the work Purple Fabric is sold for: a loan approval memorandum, a complaint investigation, an insurance submission triaged. The Enterprise Knowledge Garden indexes policies and case records so an agent can retrieve them mid task, and Flow Designer connects agents to event triggers and branches the work.
A bank can assign a different model to each step through the LLM Optimization Hub and swap one out when its cost or accuracy changes. With the models taken out, the platform has nothing left to run, since every workflow it sells passes through an agent.
Flow Designer decides when agents run and which step takes their output. A flow can start on a file drop or a webhook, branch, and stop at an approval step a person has to clear before work moves on. Enterprise Governance applies policy filters, toxicity controls and access rules, and Trace Viewer records each decision step with its sources. New agent versions roll out in stages, with canary releases and one click rollback. In lending, the agents prepare the credit memorandum and policy checks, and the underwriter makes the call with the trace of what the agents read and did in front of them.
Benchmarking Studio compares models and prompts on accuracy, latency and cost, so an institution can test combinations for a workflow before choosing one, and each agent release pins its model, prompt and tool behavior. Held out benchmarks run regression and hallucination checks before deployment, and Trace Viewer records every step of an agent run with source citations for a validator to inspect. Intellect backs accuracy with a service level of 95% or better. The validation process is not mapped to SR 11-7 or another banking model risk framework, so the bank still assembles its own validation package.
AU Small Finance Bank selected Purple Fabric as its enterprise AI platform, with PF Credit as the first use in credit assessment, and chief executive Sanjay Agarwal tied the choice to decision speed, risk assessment and personalization. Intellect also names Norges Bank Investment Management and St. James's Place as references.
The measured result comes from an unnamed European financial institution, where 13 agents and six models work more than 700 active complaints a week and resolution time fell 65%. The complaint work won a UK National AI Award in October 2025. None of the named institutions has a published figure attached to it.
Customer documents, prompts and outputs stay inside the institution's private tenancy or virtual private cloud during inference, each tenant has its own encryption keys, and Intellect states zero data retention and no training or improvement of any model on customer data, with prompts and outputs kept from other customers and from the cloud and model providers. A bank that tunes a model on its own material gets exclusive use of it.
The service terms still license Intellect to improve the service with customer data and to compile aggregated statistics from it, and their termination clause covers erasure of personal data only, subject to statutory retention. How long documents and outputs survive the end of a contract is not stated.
Customers keep ownership of their data under the service terms, which license Intellect to run, support and improve the service with it. The same terms say the service is not built to process personal information and ask customers to keep it out of their inputs, a real constraint for a bank whose case files carry customer details.
The privacy policy covers access and correction rights and cross border transfers, with contractual safeguards for data leaving Europe, the UK and Switzerland, and Intellect claims GDPR and India DPDP compliance. Optional sources such as Google Drive and OneDrive connect only when a user authorizes them.
SOC 2 Type I and Type II audits cover Purple Fabric, with an independent audit every year, and Intellect lists an ISO/IEC 42001 certified AI management system. ISO 27001, 27017 and 27018 appear as standards the platform aligns with, which is weaker than a certificate. Data is encrypted with AES 256 at rest and TLS 1.3 in transit, with a separate key per tenant and role based access to each knowledge collection. None of the reports or certificates opens from the site, and their dates and audit scopes are not stated.
Intellect Design Arena supplies Purple Fabric to regulated institutions as software, and the platform holds no financial license, registration or filing role of its own. Banks use its activity trail to show internal audit and supervisors how an agent reached an output, and Intellect's GDPR and India DPDP claims concern how the platform handles data rather than any regulated activity. AU Small Finance Bank's banking license belongs to the bank. Purple Fabric has no supervised standing beyond that, such as a regulator's examination of it as a technology provider or a test inside an approved sandbox.
Enterprise Governance lets teams set policy filters, toxicity controls and human approvals, and Intellect lists an AI management system certified to ISO/IEC 42001. PF Credit puts agents inside credit work, where they assess risk and flag deviations while a bank processes applications. For that use there is no fair lending method, no result by applicant group and no adverse action process for a decision the agents inform. The governance tooling is general purpose, so measuring disparate impact in credit falls to the bank.
Intellect's service terms disclaim any warranty on the accuracy, completeness or reliability of generated output, warn that it may contain errors or hallucinations, and leave the risk of using it with the customer. General damages are capped at the fees paid in the three months before the event. Against that, Trace Viewer records the sources and steps behind each output, so a bank can reconstruct and challenge what an agent did in a given case. The institution carries the loss from a wrong output, with a trail to contest it but no compensation behind it.
The LLM Optimization Hub serves OpenAI models through Microsoft Azure, Anthropic Claude through Amazon Bedrock and Gemini through Google Cloud, and a bank chooses the model for each agent or use case. The Azure route offers a private link, and an institution can plug in a private model under its own credentials. MongoDB runs the knowledge store, while the model services generate responses from the context an agent passes them. Intellect names its inference providers but keeps no public subprocessor list covering the whole service.
PF Credit can run on Intellect's own eMACH.ai Lending or connect to a bank's existing origination, servicing and collection systems through APIs that feed structured data into an agent flow. The Enterprise Knowledge Garden indexes documents and files, with Google Drive and OneDrive as optional sources and MongoDB Atlas behind vector storage and search. Every named lending integration is to Intellect's own platform, and a production deployment on another vendor's core banking or loan system is not named.
Banks can run Purple Fabric in a virtual private cloud they control on AWS or Azure, take a dedicated tenant on Intellect's own cloud, or use a hybrid setup in which the platform runs in the cloud and reaches records that stay on the bank's premises. In the customer cloud option, the knowledge layer, agent orchestration and governance components sit inside the institution's own network. The hosted option isolates each tenant and uses private links to internal systems. Agents run in portable runtime cells, so the same expert can move between environments and serve web, mobile, Microsoft Teams and API channels.
In India, Enterprise AI on Tap costs ₹99,500 a month for up to 50 users. A bank that runs the platform in its own cloud pays a $0 platform license plus an annual support subscription charged per core, with a 60 core minimum. That subscription buys third and fourth level platform support; other support is priced separately, and the bank contracts directly for language models, MongoDB and other services. The per core rate itself is quoted, so a buyer outside the India package knows the billing unit and the minimum footprint before talking to sales, but not the annual cost.
Commercial and retail banks use the platform for lending, trade finance and service work, and PF Credit runs from origination and underwriting through loan operations and collections. Insurers get a separate suite for submission intake and triage, built for carriers, managing general agents and brokers. Wealth and asset managers use PF Wealth for complaint investigation and investment research. Suites for risk and compliance and for finance and accounting sit beside these. Each group works from its own knowledge collections and workflows on the same agent design and governance tools, and Intellect counts 45+ regulated institutions in production.
Alternatives to Purple Fabric
Other products that do the same kind of work as Purple Fabric, and what each one does.
Intellect Design Arena, the Chennai software company demerged from Polaris in 2014, sells eMACH.ai, a composable banking platform for retail, commercial, central and Islamic banks and credit unions.
Zafin AIOS is an agent orchestration platform for banks and other regulated institutions, sold separately from Zafin's Banking Platform and Zafin IO.
Docsumo reads and checks documents for lenders, financial services companies and insurers. The New York company, founded in 2019, lets a lender send in an application with bank statements, pay stubs, tax returns and supporting files.
Titan sells AI built for the daily work of banks. Staff use Titan Foundry, a secure chat assistant, to draft documents, run research and answer regulatory questions with citations.
Finzly sells BankOS, an operating layer that banks, credit unions and fintechs connect to their existing core.
Zafin, a Toronto software company, sells the Zafin Banking Platform, which takes a bank's product, pricing, offer, billing, loyalty, relationship and deal rules out of its core systems and runs them in one place.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
₹99,500 per month for up to 50 users in India.
|
Enterprise AI on Tap is a monthly subscription for up to 50 users. A deployment in the customer’s cloud has a $0 platform license fee and an annual support subscription charged per core, with a 60 core minimum. | The platform supports deployment in the institution’s own virtual private cloud. Under the standard service terms, the customer keeps ownership of its data, and personal information is excluded from service inputs. | The annual support subscription covers third and fourth level platform support. Other support is priced separately. Professional services cover model integration, agent development and deployment, with no fixed implementation fee stated. | Vendor Published |
The ₹99,500 monthly package was introduced in India. For customer cloud deployments, language models, MongoDB and other services are separate subscriptions contracted directly with their providers. Per core support rates are quoted separately.