CRED, the private club for good-credit users in India, is using OpenAI models to transform support and product experiences without losing what defines them: trust, design and security.
How do you scale a concierge-like experience when millions expect fast, accurate answers? CRED's answer was to put AI at the center of customer service and internal processes.
What CRED is doing with AI
CRED built a family of tools powered by models like GPT-4.0, GPT-5 and o3 to serve users, support agents and speed up operations.
The visible face is Cleo, a conversational assistant aiming to move from transactional replies to empathetic, contextual conversations.
Cleo handles three main types of queries:
- Informational: for example, 'What is CRED Cash?'
- Contextual: 'Am I eligible for CRED Cash?'
- Transactional: 'Can I get a refund to my wallet or the original payment method?'
Cleo diagnoses the issue, classifies intent, maps the action to the correct standard operating procedure (SOP) and generates a contextual, accurate response.
Besides Cleo, they designed two internal tools:
- Thea: for support agents, it summarizes conversations in multiple formats (text, voice, Hinglish) and suggests next steps.
- Stark: for operations teams, it lets you create or update SOPs in minutes instead of days.
Measurable results and why they matter
The numbers are early but revealing. CRED reports significant improvements in key metrics since launch:
- 14,000 basis points improvement in CSAT: that's equal to a 140 percentage-point increase, a huge jump that reflects greater satisfaction.
- Cleo reached 98% resolution accuracy within three months.
- 18% more multi-intent conversations successfully resolved.
- Session abandonment rates fell 31%, and average handling times decreased across the board.
What does this mean in practice? Less frustration for you as a user, fewer escalations for agents, and more up-to-date SOPs for operations. In short: a concierge-like experience that can actually scale.
Practical lessons you can apply
- Start from your values: CRED kept trust, security and design front and center when integrating AI. What are you not willing to sacrifice?
- Measure from day one: using an internal evaluation framework (which even uses models) helped build confidence quickly.
- Design for people, not just use cases: Cleo handles contexts and multi-intents, not isolated questions.
- Close the data loop: detect 'data dead-ends' (queries that can't be answered) and feed them back into the knowledge base so SOPs improve in real time.
- Start small, scale fast: piloting with a demanding audience (premium members) forces high standards, which makes it easier to replicate quality later.
Risks, adaptation and culture
There was initial skepticism, as often happens with new tech. The difference was teams saw tangible results and adoption accelerated.
The biggest surprise was the speed of adaptation: when people experience a real unblock, they adopt the tool and look to integrate it into their daily flow.
CRED also aims for AI to 10x each person across functions: engineering, QA, infrastructure, compliance. That isn't just efficiency; it's the ability to make faster, better decisions at scale.
What's next?
The plan is to expand Cleo across all lines of business and keep building tools that detect information gaps and feed organizational knowledge.
The ultimate goal isn't just automation, but amplifying values: speed, accuracy and compound effect.
CRED shows AI isn't magic or a replacement — it's well-designed amplification. If your company wants to offer a premium service without losing quality as it grows, the lesson is clear: integrate AI aligned with your values, measure rigorously and prioritize human experience over automation for its own sake.
