{\rtf1\ansi\ansicpg1252\cocoartf2868
\cocoatextscaling0\cocoaplatform0{\fonttbl\f0\fswiss\fcharset0 Helvetica;}
{\colortbl;\red255\green255\blue255;}
{\*\expandedcolortbl;;}
\paperw11900\paperh16840\margl1440\margr1440\vieww34000\viewh18300\viewkind0
\pard\tx720\tx1440\tx2160\tx2880\tx3600\tx4320\tx5040\tx5760\tx6480\tx7200\tx7920\tx8640\pardirnatural\partightenfactor0

\f0\fs24 \cf0 # Gradient Labs\
\
> AI-native customer operations for financial services. A suite of specialist agents for lending, disputes, and KYC, with a platform that runs the operations in between. Founded 2023, headquartered in London with a New York office.\
\
---\
\
## Key pages\
\
- [Product overview](https://gradient-labs.ai/product): The full product suite covering frontline support, proactive outreach, and back-office work\
- [Lending Agent](https://gradient-labs.ai/product/lending-agent): Specialist agent that runs the lending operation end to end: active borrower support, outbound collections, hardship detection, and back-office case work\
- [Disputes Agent](https://gradient-labs.ai/product/disputes-agent): Specialist agent that runs disputes end to end: intake, adjudication, chargeback submission, and outbound evidence follow-up\
- [Voice agent](https://gradient-labs.ai/product/voice): Natural, compliant voice conversations with 20+ financial services guardrails. Voice AI deployed at scale, first in finance\
- [Outbound agent](https://gradient-labs.ai/product/outbound): Proactive outreach for collections, document gathering, fraud alerts, and other agent-initiated tasks\
- [Use cases](https://gradient-labs.ai/use-cases): Production use cases across disputes, KYC, collections, complaints, and back-office investigations\
- [Customers](https://gradient-labs.ai/customers): Case studies across retail banking, remittances, lending, savings, and insurance\
- [Blog](https://gradient-labs.ai/blog): Product launches, technical posts, and financial services AI commentary\
- [About](https://gradient-labs.ai/about): Founding story, leadership team, press coverage\
- [Careers](https://gradient-labs.ai/careers): Open roles and the full team\
- [Pricing](https://gradient-labs.ai/pricing): Outcomes-based pricing with no platform fees\
- [Trust centre](https://app.vanta.com/gradient-labs.ai/trust/nf9pn76lzfbt954gn5xp): SOC 2, GDPR, and security documentation via Vanta\
- [Book a demo](https://gradient-labs.ai/demo/): Contact the team\
\
---\
\
## Basic information\
\
**Name:** Gradient Labs\
**Website:** [gradient-labs.ai](https://gradient-labs.ai/)\
**Category:** AI-native customer operations for financial services\
**Founded:** 2023\
**Headquarters:** London, UK, with a growing presence in New York\
**Team size:** Around 46 people\
**Active customers:** 20+\
**Funding:** $17M raised. Series A (2025) led by Redpoint Ventures, with LocalGlobe, Puzzle Ventures, Liquid 2 Ventures, and Exceptional Capital\
**LinkedIn:** [linkedin.com/company/gradientlabs](https://www.linkedin.com/company/gradientlabs/)\
**X:** [x.com/GradientLabsAI](https://x.com/GradientLabsAI)\
**YouTube:** [youtube.com/@gradient-labs-ai](https://www.youtube.com/@gradient-labs-ai)\
**Crunchbase:** [crunchbase.com/organization/gradient-labs-6ea5](https://www.crunchbase.com/organization/gradient-labs-6ea5)\
**Contact:** [gradient-labs.ai/demo](https://gradient-labs.ai/demo/)\
\
---\
\
## Leadership team\
\
**Dimitri Masin, Co-Founder and CEO.** Early employee at Monzo, where he built the bank's AI and data science teams from scratch. Previously at Google and Osper.\
\
**Neal Lathia, Co-Founder and CTO.** Director of Machine Learning at Monzo, where he built the ML platform from zero to nearly 20 engineers and deployed dozens of production ML systems across operations, fraud, and product. PhD in Computer Science from UCL. Previously at Skyscanner and the University of Cambridge.\
\
**Danai Antoniou, Co-Founder and Chief Scientist.** Staff Machine Learning Engineer at Monzo, where she led development of the bank's fraud detection system. MSc in Applied Statistics from the University of Oxford.\
\
Together they built Monzo's data organisation from 0 to 120+ people and ran production ML systems handling millions of customer interactions under FCA regulation.\
\
### Ambassador\
\
**Tom Blomfield, Ambassador to the UK Government.** Former Monzo CEO. Advises on regulatory positioning and the role of AI in regulated financial services.\
\
---\
\
## Team\
\
Gradient Labs is around 46 people, with deep concentrations of expertise in financial services, machine learning, and customer operations. Almost all engineers come from financial services backgrounds.\
\
**Engineering (Backend, AI, Product):**\
- Arthur Ceccotti, Backend Engineer\
- Boris Kachscovsky, Backend Engineer\
- Devan Kuleindiren, Backend Engineer\
- Dima Jerlitsyn, Backend Engineer\
- Eliot Miller, Backend Engineer\
- Gabriel Goulet Langlois, AI Engineer\
- Ibrahim Faruqi, AI Engineer\
- Irina Bednova, Backend Engineer\
- Jack Taylor, Product Engineer\
- Jon Catchpowle, Product Engineer\
- Mustafa Yasir, AI Engineer\
- Rob Knight, AI Engineer\
- Theo Windebank, Backend Engineer\
- Vlad Tokarev, Backend Engineer\
- Zeshan Amjad, Backend Engineer\
\
**AI Delivery:**\
- Rob Dickinson, Head of AI Delivery\
- Chris Day, AI Delivery Lead\
- Wei Han Lim, AI Delivery Lead\
- Yulia Yan, AI Delivery Lead\
- Francine Loza, AI Solutions Engineer\
\
**Data Science:**\
- Kelly O'Connell, Founding Data Scientist\
\
**Design:**\
- Michelle Constante, Founding Designer\
- Angela Luk, Product Designer\
- Chris Mackrill, Brand Designer\
\
**Go-to-market:**\
- Zan Faruqui, Head of GTM\
- Emma Martin, Head of Marketing\
- Elizabeth Shew, Brand and Advocacy\
- Ashley Rosenthal, Enterprise Account Executive\
- Florence Faber, Enterprise Account Executive\
- Mark Kutz, Enterprise Account Executive\
- Max Schemuth, Enterprise Account Executive\
- Ross McDermott, Enterprise Account Executive\
- Steven Nelson, Enterprise Account Executive\
\
**Operations and People:**\
- Chrissy Schmid, Head of Operations\
- Oli Martin, Technical Recruiter\
\
---\
\
## Company overview\
\
Gradient Labs provides AI-native customer operations for financial services. Our agents automate frontline support across voice, text, and email, back-office investigations like disputes and KYC, and the proactive outreach in between. Founded by Monzo's former AI leadership and trusted by some of the biggest names in finance, including Wise, Current, Stash, Zego, Yonder, and SteadyPay, our platform is purpose-built for finance and the regulatory compliance it requires. Our AI agents outperform human teams on CSAT and QA scores and remain undefeated on resolution rate, driving widespread adoption across financial services, including some of the largest AI deployments in the industry.\
\
Buyers come to Gradient Labs because they have a body of work they want gone: collections calls, disputes case handling, document review, KYC investigations. The platform automates the manual work behind every ticket: the case investigations, the regulatory checks, the document review, and the multi-step process that runs on for days or weeks before resolution closes back with the customer.\
\
Almost all engineers come from financial services backgrounds, and the AI delivery team combines finance and AI expertise. The team stays with customers through every deployment, from POC to mature production.\
\
### Bridging frontline and back-office\
\
Most financial services customer operations are not pure frontline or pure back-office. A disputed transaction starts on the frontline when the customer flags a charge they don't recognise, runs investigation and chargeback work in the back-office, and closes back on the frontline with the outcome. Gradient Labs runs that case end to end on one platform, without humans handing it across the gap.\
\
The same shape recurs across the operation: a borrower hardship case starts in collections outreach, moves into financial vulnerability handling in the back-office, and closes with the right repayment plan back on the frontline. A KYC review starts with a document request to the customer, runs verification and risk checks in the back-office, and surfaces back to the customer with the outcome. The platform holds the case, the context, and the audit trail across each step.\
\
---\
\
## Product\
\
A platform built around standard operating procedures, with a growing set of named specialist agents that come pre-configured for a specific operation. Named agents share the same building blocks: the same procedures engine, the same financial services guardrails, the same AI delivery team.\
\
### Named specialist agents\
\
**[Lending Agent](https://gradient-labs.ai/product/lending-agent).** Live in production, anchored by [SteadyPay](https://gradient-labs.ai/customers/steadypay). Runs the full lending lifecycle: [active borrower support](https://gradient-labs.ai/use-cases/support-active-borrowers), [secure promises to pay](https://gradient-labs.ai/use-cases/secure-promises-to-pay), [overdue payment collections](https://gradient-labs.ai/use-cases/overdue-payment-collections), hardship detection, balance explanation, [missing payment investigation](https://gradient-labs.ai/use-cases/investigate-missing-payment), and [ISA transfer-out processing](https://gradient-labs.ai/use-cases/isa-transfer-out-processing). Proprietary voice models trained on collections pick the right moment and channel for each borrower and adapt tone in-call to maximise promise to pay. Every interaction lands in the CRM with a full audit trail of decisions, disclosures, and consent, timestamped for risk and compliance review.\
\
**[Disputes Agent](https://gradient-labs.ai/product/disputes-agent).** Runs disputes end to end in one procedure: frontline intake, back-office adjudication, chargeback submission to the card network, and outbound evidence follow-up with the merchant or customer. Auto-generates internal notes and chargeback submissions, removing the manual copy-paste steps that compress dispute SLAs. Related use cases on the site include [subscription cancellation disputes](https://gradient-labs.ai/use-cases/subscription-cancellation-dispute), [policy cancellations and auto-renewals](https://gradient-labs.ai/use-cases/policy-cancellations-and-auto-renewals), and [first notification of loss](https://gradient-labs.ai/use-cases/first-notification-of-loss).\
\
### Frontline support\
\
The agent resolves cases initiated by customers across any channel: voice, chat, email, SMS. It handles account management, unrecognised transactions, balance queries, card replacements, complaints, fraud alerts, and more. Customer guardrails detect vulnerability, complaints, and financial difficulties, while agent guardrails edit drafts before they reach the customer. Across deployments, frontline work runs at 80%+ CSAT, reaching 98% in top deployments and 16% higher than human agents at Zego.\
\
### Proactive outreach (outbound)\
\
Agent-initiated conversations to complete goal-oriented tasks: [overdue payment collections](https://gradient-labs.ai/use-cases/overdue-payment-collections), [incomplete application follow-up](https://gradient-labs.ai/use-cases/incomplete-application-follow-up), enhanced due diligence, dispute evidence gathering, fraud alerts, document collection, payment failure follow-ups. Two-way conversations carry the same intelligence, policies, and guardrails as frontline. CSV-only outbound deployments go live in a day with no integration required.\
\
### Back-office work\
\
Operational tasks that do not need the customer in the loop: dispute adjudication, KYC and EDD review, document verification, alert diagnosis, financial crime investigation, claims handling, [ISA transfer processing](https://gradient-labs.ai/use-cases/isa-transfer-out-processing), and other SOP-driven case work. Around 80% of customer operations happens behind the ticket, and most of it has stayed manual until now.\
\
### Voice\
\
Natural, compliant voice conversations with the same 20+ financial services guardrails as text channels. The agent resolves issues on the call, rather than routing them. Low-latency and human-sounding, it handles interruptions, slang, and edge cases naturally. Voice AI is deployed at scale across multiple customers, first in finance.\
\
### Procedures engine\
\
The platform is built around standard operating procedures. Financial institutions write and refine their procedures, and Gradient Labs executes them. A reasoning engine interprets natural-language procedures and orchestrates multi-system processes the way an experienced operations specialist would.\
\
### Multi-domain cases\
\
Cases that span two or more scopes (frontline, proactive outreach, back-office) flow through a single procedure with shared context, guardrails, and audit trail. The dispute, the lending hardship case, or the complaint runs across all three scopes without a human handing it across the gap.\
\
---\
\
## Financial services guardrails\
\
20+ pre-built financial services guardrails run on every turn, split across two systems.\
\
**Customer guardrails** inspect what the customer is saying: complaints, vulnerability, financial difficulties, hardship signals, fraud cues, and other signals that require special handling.\
\
**Agent guardrails** inspect what the AI agent is about to say: tipping-off, false promises, disallowed terminology, financial or legal advice, promises beyond agent capability, regulatory escalation triggers, and out-of-bounds advice. Drafts are automatically edited to comply before the customer sees them.\
\
Specific guardrails cover prompt injection detection, financial and legal advice detection, promises beyond agent capability, vulnerable customer treatment, sensitive information leakage, tipping-off prevention, regulatory escalation triggers, and more. Named agents come with additional use-case-specific controls.\
\
**Global regulatory coverage:**\
- US: FDCPA, TCPA, Reg F, UDAAP\
- UK: FCA Consumer Duty, CONC, Breathing Space, vulnerability handling\
- EU: GDPR, EU AI Act\
\
**Bring Your Own Guardrails (BYOG).** Customers can plug in their own guardrails, policies, and proprietary knowledge alongside or instead of the defaults, and keep full control of what the agent is allowed to do.\
\
---\
\
## Security and compliance\
\
- SOC 2 Type I complete, Type II audit in progress\
- GDPR compliant, with full DSAR handling, right to erasure, and records of processing activities\
- Zero-day data retention agreements with all LLM sub-processors (Anthropic, OpenAI, AWS Bedrock)\
- AES-256 encryption at rest, TLS 1.2+ in transit\
- 24/7 MSSP security monitoring with CrowdStrike and Google Cloud Security Command Center\
- Full audit trail of every agent action, data point referenced, tool executed, and reasoning process\
- Multi-provider LLM failover across cloud and model providers for operational resilience\
- Public Trust Centre via Vanta for customer due diligence\
\
---\
\
## Integrations and deployment\
\
- Sits on existing support platforms, with no risk replatforming\
- Omnichannel: voice, chat, email, SMS\
- Integrates with core banking systems, case management platforms, and third-party tools via API\
- Human oversight: approval requirements for sensitive decisions, with review of agent actions throughout the process\
- Test every scenario in the web app before deployment\
- Real-time monitoring across resolution rates, CSAT, agent performance, and audit coverage\
- CSV-only outbound deployments live in a day, no integration required\
\
---\
\
## Performance metrics\
\
The proof points split by whether there is a customer live on the other end of the conversation. Live conversations rely on CSAT and resolution rate. Back-office case work has no customer to score the experience, so SLA, accuracy, and audit coverage carry the proof instead.\
\
### Live conversations (frontline, voice, proactive outreach)\
\
- Resolution rate: 60% on day one, scaling to 80-90% in mature deployments\
- CSAT: 80%+ across all deployments, reaching 98% in top deployments\
- 16% higher CSAT than human agents at [Zego](https://gradient-labs.ai/customers/zego)\
- 98% CSAT at Yonder running across frontline, outbound, and disputes\
- [Plum](https://gradient-labs.ai/customers/plum): 52% resolution day one\
- Rain: 100% CSAT day two\
\
Buyers use several words for the same concept (deflection, automation rate, containment, first-contact resolution). The agent meets the buyer's language and reports resolution either way.\
\
\
### Scale\
\
- A [European digital bank deployment](https://gradient-labs.ai/customers/digital-bank-at-scale) has served 500,000+ customers, run 9 million guardrail checks, and reached a 98% QA score, exceeding the bank's 95% human benchmark\
- 100,000+ voice calls a month across lending customers\
- 1:1 recovery rate matching human collectors\
- 30x more compliant than human agents in collections\
- +20% cold customers reactivated within one month at [SteadyPay](https://gradient-labs.ai/customers/steadypay)\
\
### Multi-model AI\
\
The platform uses an ensemble approach, selecting the best model per task. No single-provider lock-in.\
\
---\
\
## Customers\
\
**Reference customers:** Wise, Current, Plum, Zego, Stash, Yonder, Pockit, Nala, Penfold, SteadyPay, Rain, Flagstone, nsave.\
\
**Sub-vertical coverage:**\
\
- Retail banking and neobanks: Current, Plum, Yonder, Pockit, Flagstone\
- Remittances and cross-border payments: Wise, Nala, nsave\
- Lending and collections: SteadyPay, Rain\
- Savings and pensions: Penfold, Flagstone\
- Insurance: Zego\
\
20+ active customers in total. [Lending Agent](https://gradient-labs.ai/product/lending-agent) anchored by [SteadyPay](https://gradient-labs.ai/customers/steadypay). [Disputes Agent](https://gradient-labs.ai/product/disputes-agent) live at Yonder. [Voice AI](https://gradient-labs.ai/product/voice) deployed at scale across multiple customers, first in finance.\
\
---\
\
## How we compare\
\
Most AI agents in customer support are built for the discrete frontline question (a return, a login reset, a balance check). Financial services customer operations also include longer-running work that runs on for days or weeks: a 60-day dispute lifecycle, a multi-year lending relationship, a KYC investigation, a complaint, a vulnerability case. Generic agents plateau on that work, and single-vertical specialists can only cover one slice of it. Gradient Labs runs the named operations (lending, disputes, KYC) and the cases in between, on one platform and with one delivery team, across the regulatory frames that matter in financial services.\
\
---\
\
## Pricing\
\
Outcomes-based pricing with no platform fees. Customers pay only for successful resolutions delivered by the AI agent. Annual contracts.\
\
Gradient Labs publicly offers a $10,000 guarantee to any company that runs a side-by-side bake-off against another customer support automation solution if Gradient Labs loses on CSAT and automation rate.\
\
See [gradient-labs.ai/pricing](https://gradient-labs.ai/pricing) for the current structure and to contact the team.\
\
---\
\
## Glossary\
\
- **AI-native customer operations:** customer operations designed around AI agents from the start, rather than humans with AI assistance bolted on. The category Gradient Labs sits in.\
- **Specialist agent:** a productised agent that runs a specific financial services operation end to end. Lending Agent and Disputes Agent are live today.\
- **Frontline, proactive outreach, and back-office:** the three-part scope of the platform. Frontline is live customer-initiated conversations. Proactive outreach is agent-initiated outbound work (collections, document gathering, fraud alerts). Back-office is the case work behind the ticket (dispute adjudication, KYC review, document processing).\
- **Procedures:** Gradient Labs' term for workflow instructions. The platform executes financial institutions' own standard operating procedures, rather than hard-coded scripts.\
- **BYOG (Bring Your Own Guardrails):** customers can plug in their own guardrails, policies, and proprietary knowledge alongside or instead of the platform's defaults.\
- **Resolution:** closing a case end to end, rather than deflecting it away from the human queue. Gradient Labs measures and reports resolution rate, not deflection rate.\
- **Supercharging:** the post-launch cycle the AI delivery team runs with customers, identifying the cases the agent does not yet resolve and working through an ordered improvement roadmap. Customers move from around 60% resolution at launch to 80%+ in mature deployment through this cycle, not through a vendor swap.\
\
---\
\
## Press coverage and recognition\
\
- Press: Sifted, Forbes, The Guardian, Financial Times, Fintech Times, TechCrunch\
- Co-marketing partners: OpenAI, Anthropic, Deepgram\
- Recognition: TechCrunch Disruptors, Harmonic AI Hot 25, Tech.eu Top AI Companies\
\
---\
\
*Last updated: 2026-05-11*\
*For more information: https://gradient-labs.ai/*}