How much does Gradient Labs cost?

Outcomes-based pricing with no platform fees, per gradient-labs.ai/pricing; no figures published, quoted through sales. Money-back commitment on delivery.

Does Gradient Labs require human approval before an agent acts?

Full under the Agentic Index verification standard. Gradient Labs publishes public evidence of human approval steps and guardrails, so an agent action can be held for review before it executes. Evidence basis: Guardrails are the shipped mechanism and they are counted: more than 20 financial services guardrails run on every conversation, with 9 million guardrail runs recorded across a single European digital bank deployment, and the agent escalate. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does Gradient Labs have SOC 2, SSO and role based access control?

Partial under the Agentic Index verification standard. Gradient Labs publishes some public evidence of security and identity controls, but narrower than the full set of SOC 2, single sign on (SSO) and role based access control (RBAC). Evidence basis: A Vanta trust center linked from the site footer publishes SOC 2 and GDPR documentation; access controls are not published. More than 20 financial services guardrails run on every conversation, with 9 million guardrail runs recorded at one. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Does Gradient Labs give you tracing and observability into agent runs?

Partial under the Agentic Index verification standard. Gradient Labs publishes some public evidence of logging or observability, but narrower than full tracing and auditability across agent runs. Evidence basis: Quality is measured and published per deployment, with a 98% QA score and guardrail run counts reported at a European digital bank and CSAT tracked against human agent benchmarks across every deployment. These figures report outcomes rather. Graded from public evidence only against the Agentic Index 14 point capability taxonomy. No vendor pays for placement.

Back to vendors

Gradient Labs

Q: What are the best alternatives to Gradient Labs?

Based on documented capability coverage across the Agentic Index 14 point taxonomy, the closest alternatives to Gradient Labs among customer support agents are Worknet, Cendra, Readyly, Siena AI. Each is graded on the same public evidence, and no vendor pays for placement.

Also known as: Otto, Lending Agent, Disputes Agent, Gradient Labs AI

Visit site

Customer support agentindependentVerified 2026-09-05

Entry priceContact for pricingFull pricing detail

Vertical AI agent for regulated financial services whose agent Otto autonomously runs end to end operations, disputes, KYC, lending, fraud, and support, following procedures rather than scripts.

Gradient Labs builds vertical AI agents for regulated financial services, betting that full autonomy, not co-pilots, delivers safer and more compliant outcomes. Founded in London in 2023 by Dimitri Masin, Neal Lathia, and Danai Antoniou, early employees of the challenger bank Monzo, the company grew revenue nine hundred percent in a year and now supports more than thirty two million end users. It raised an extended twenty six million dollar Series A led by Octopus Ventures and CommerzVentures, bringing total funding near forty three million dollars, and counts Wise, Monzo, Zego, Pockit, Current, Stash, and Rho among its customers.

Its flagship agent, Otto, is a procedure following agent trained on a company's specific products and processes, and it resolves complex inquiries end to end rather than deflecting them. Gradient Labs offers a suite of specialist agents for lending, disputes, and know your customer checks, plus a platform that runs the operations in between. On disputes, for example, Otto gathers and verifies the evidence, decides the outcome, escalates to a human for sign off, and files the chargeback with the card scheme. It also automates business verification, claims intake, arrears collection, and fraud interception.

The agents work as interconnected specialists over a platform that runs the operations between them, and Otto resolves issues like freezing a lost card or tracing a missing payment. Coverage spans multiple channels including voice, and the agent works proactively and outbound, contacting a borrower in arrears at the moment they are most likely to respond or intercepting a fraud attempt live. It delivered the first ever live autonomous customer support agent for a large regulated United Kingdom bank.

Compliance is the core of the pitch. Gradient Labs describes itself as the first compliant AI platform for the financial services industry, with financial services specific guardrails and procedures that keep the agent operating safely under regulatory scrutiny, and it escalates high stakes decisions to humans for sign off. It backs its work with a money back commitment, sharing the risk if it does not deliver what it promised. Deployed as an overlay on Intercom, Zendesk, Freshworks, or in house helpdesks through an application programming interface, Otto reports resolving ninety percent of queries with ninety eight percent accuracy.

Vendor details

Canonical URL

https://gradient-labs.ai

Company status

independent

Use cases & customers

In practice

A customer disputes a card charge. Otto gathers and verifies the evidence, decides the outcome, escalates to a human for sign off, and files the chargeback with the card scheme, running the dispute end to end.

A borrower falls into arrears. Otto contacts them at the time they are most likely to respond, personalizes the message from their account history, and secures a promise to pay, all without a human agent placing the call.

A fintech onboarding a business needs verification. Otto checks each document against internal policy, requests anything missing from the customer, verifies the business, and routes it to onboarding, automating a slow manual compliance step.

Agentic Index coverage score

7.0 / 14 capabilities · 50%

Integrations & Tool Calling	Full
Authenticated action in outside systems is the whole product, and it is unusually consequential action: the Disputes Agent handles the full chargeback lifecycle including filing with the card scheme, the Lending Agent verifies identity and secures promises to pay across a collections book, and agents freeze cards and trace payments in a bank's own systems. The platform deploys as an overlay on Intercom, Zendesk, Freshworks or an in-house helpdesk. Sourcegradient-labs.ai/product/disputes-agentread 2026-09-05
Workflow Orchestration	Full
Regulated workflows that a person would otherwise run over days are executed in multiple steps across multiple agents. The Disputes Agent handles the full chargeback lifecycle, gathering and verifying evidence, deciding the outcome, escalating for sign-off and filing with the card scheme; the Lending Agent runs the collections lifecycle, verifying identity and securing promises to pay; and specialist agents for lending, disputes and KYC sit over a platform that runs the operations between them. Sourcegradient-labs.ai/product/disputes-agentread 2026-09-05
Knowledge Grounding & RAG	Partial
Otto is a procedure-following agent trained on a company's own products, processes and internal policies, so answers and decisions are grounded in the institution's documented procedures rather than open-ended generation, which is what lets it run regulated workflows end to end. No index, embeddings layer, retrieval architecture, refresh behavior or source-level citation is documented, so a maintained retrieval structure is asserted rather than shown. Sourcegradient-labs.ai/productread 2026-09-05
Human Oversight & Guardrails	Full
Guardrails are the shipped mechanism and they are counted: more than 20 financial services guardrails run on every conversation, with 9 million guardrail runs recorded across a single European digital bank deployment, and the agent escalates high-stakes decisions to a person for sign-off before acting, including dispute outcomes and lending conversations that need a human specialist. That is a vendor-owned gate before the action rather than routing after it. Sourcegradient-labs.ai/product/voiceread 2026-09-05
Security, Identity & Governance	Partial
A Vanta trust center linked from the site footer publishes SOC 2 and GDPR documentation; access controls are not published. More than 20 financial services guardrails run on every conversation, with 9 million guardrail runs recorded at one European digital bank. The terms of service commit contractually that neither Gradient Labs nor any subprocessor may store or use customer personal data for any purpose beyond performing the service, and they name training AI models as forbidden. No SSO or other identity integration, named role model or permission scheme appears anywhere, including in the legal documents, and the subprocessor list is available on request rather than published. Sourcegradient-labs.ai/terms-of-service/latestread 2026-09-05
Observability & Auditability	Partial
Quality is measured and published per deployment, with a 98% QA score and guardrail run counts reported at a European digital bank and CSAT tracked against human agent benchmarks across every deployment. These figures report outcomes rather than the run itself; no step level decision trace, procedure execution record or exportable audit trail is documented on the vendor's own pages. Sourcegradient-labs.ai/productread 2026-09-05
Memory & State Persistence	Partial
The architecture implies carried state: specialist agents for lending, disputes and KYC run over a platform that handles the operations between them, so context travels across agents and across a case rather than dying with a conversation. No memory layer with a stated scope or lifetime is documented on the product surface, and a buyer cannot say where memory lives or delete one customer's memory without touching the record. Sourcegradient-labs.ai/productread 2026-09-05
Deployment & Data Residency	Not documented
No hosting region, residency commitment, VPC, single tenant or self hosted option is published on the site. The privacy policy covers site visitors, cookies and embedded widgets rather than the product, and it places processing only at the owner's operating offices and wherever the processing parties are, naming no region. The subprocessor list is available on request rather than published, and the terms of service bind subprocessors contractually without naming them or their locations. A trust center is linked from the footer, but for a regulated bank the residency detail sits behind a login or a request. Sourcegradient-labs.ai/terms-of-service/latestread 2026-09-05
Prebuilt Agents, Templates & Packs	Full
A named set of prebuilt specialist agents the buyer adopts as units, each with its own product page: Lending Agent for the lending lifecycle, Disputes Agent for chargebacks and the full dispute lifecycle, Voice agent, Outbound agent for collections, document gathering and fraud alerts, and Collaborate, over a platform that runs the operations between them. Each is stated as live in production at a named customer. Sourcegradient-labs.ai/productread 2026-09-05
Triggers & Channel Coverage	Full
Work reaches the agent across voice, text and email, and the agent also starts the conversation: a dedicated Outbound agent handles collections, document gathering, fraud alerts and other agent-initiated tasks, with more than 100,000 outbound voice calls a month across lending customers alongside inbound calls answered end to end. Back-office investigations such as disputes and KYC arrive as cases rather than messages. Both directions and multiple modalities are covered. Sourcegradient-labs.ai/product/outboundread 2026-09-05
Model Flexibility & Routing	Not documented
The customer gets no model selection, routing or bring your own model control anywhere on the product surface, and the vendor's own pages name no provider, describing the platform as purpose built for finance without saying what sits underneath. A case study on a model provider's own site says the platform is built on that provider's models. Sourcegradient-labs.ai/productread 2026-09-05
APIs, SDKs & MCP Extensibility	Not documented
Gradient Labs publishes no API reference, SDK, developer documentation or MCP server for its own platform, and the site has no developer section. The API deployment it describes runs the other way, connecting the platform outward into a customer's helpdesk, whether Intercom, Zendesk, Freshworks or an in house system. Sourcegradient-labs.ai/productread 2026-09-05
Testing, Debugging & Optimization	Not documented
Before deployment, Gradient Labs replays real customer conversations against expected procedure and generates synthetic conversations to test edge cases, according to a case study on a model provider's site. That is the vendor testing its own system. Customers get no documented testing, simulation or evaluation surface of their own, and no way to run those replays against their own changes is published. Sourcegradient-labs.ai/productread 2026-09-05
Browser & Computer Use	Not documented
The agents act through integrations into a customer's helpdesk and financial systems and through the card scheme itself, a programmatic route by the vendor's own description. No browser control, hosted session, virtual desktop or computer use appears on the vendor's own pages. Sourcegradient-labs.ai/productread 2026-09-05

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Recent platform changes

2026-08-19·Observability / auditabilityVerified

Gradient Labs released Collaborate, a new workspace that allows operators, engineers, and AI agents to work together as peers. The update introduces built-in version control, evaluations, and continual learning capabilities for agent management.

Bears on: Observability / auditability

View source

View all 1 change for Gradient Labs →Tracked since Aug 2026 · Verified from public vendor sources

Pricing

Contact for pricing

Agentic Index verified 2026-09-05

Official pricing page

Related vendors

Alternatives to Gradient Labs

The closest documented capability profiles to Gradient Labs among customer support agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.

Worknet7.0 / 14Fuller documented coverage on Knowledge Grounding & RAG
Cendra7.5 / 14Adds documented Testing, Debugging & Optimization
Readyly7.5 / 14Adds documented Testing, Debugging & Optimization
Siena AI7.5 / 14Adds documented Testing, Debugging & Optimization
BIK8.0 / 14Adds documented APIs, SDKs & MCP Extensibility and Testing, Debugging & Optimization
Crescendo8.0 / 14Adds documented Testing, Debugging & Optimization

Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded

Compare Gradient Labs with other vendors →Compare all customer support agents →See where Gradient Labs ranks →

Data confidence: medium

Head to head

Gradient Labs vs Lorikeet