Playbooks
How Kaspi Can Deploy AI Agents to Deliver a True Customer‑First Experience
Mei Zhang
@playbook-press·September 29, 2026
3 min read0 viewsAlmaty, Kazakhstan
Mikhail Lomtadze’s mantra that Kaspi must only be judged by whether it makes customers’ lives better is a strategic north star. Translating that conviction into a scalable, data‑driven operation calls for AI agents that act as personal finance assistants, real‑time support reps, and proactive risk monitors. Below is a concrete implementation roadmap that turns the philosophy into quantifiable outcomes.
Phase 1 (0‑4 weeks): Assess & Prioritise
- Map every customer touchpoint—mobile app, web portal, call centre, ATM. Identify three high‑volume pain points (e.g., loan‑application bottlenecks, transaction‑error resolution, onboarding FAQs). 2. Assemble a cross‑functional squad (product manager, data engineer, AI‑ops lead, compliance officer). 3. Define success metrics: average handling time (AHT) cut by 30%, Net Promoter Score (NPS) lift of 5 points, and AI‑driven cross‑sell conversion up 2%.
Phase 2 (5‑12 weeks): Prototype AI Agent Suite
- Build a conversational AI layer using a fine‑tuned large language model (LLM) hosted on a secure private cloud. Integrate with Kaspi’s existing user‑profile API to pull real‑time balance, credit score, and transaction history. Deploy three micro‑agents:
- Loan‑Guide Bot: auto‑fills application fields, predicts approval odds, and schedules follow‑up calls.
- Error‑Resolution Bot: parses error codes from ATM logs, offers instant remediation steps, and escalates only when needed.
- Financial‑Wellness Coach: nudges users with personalized budgeting tips based on spending patterns.
- Run a closed‑beta with 5 % of active users, collect interaction logs, and measure the predefined metrics.
Phase 3 (13‑20 weeks): Iterate & Scale
- Analyse failure cases: false‑positive escalations, privacy concerns, language nuances. Refine prompts and add rule‑based safeguards for compliance. Expand rollout to 30 % of the user base, adding multilingual support for Russian and Kazakh. Introduce a human‑in‑the‑loop escalation dashboard that surfaces only high‑severity tickets to live agents, reducing AHT further.
Phase 4 (21‑28 weeks): Full Integration & Continuous Improvement
- Embed AI agents into the core Kaspi app as a persistent “Assistant” icon. Set up automated A/B testing pipelines to test new nudges or dialogue flows weekly. Establish a KPI dashboard that tracks AHT, NPS, conversion lift, and AI‑related cost per interaction (target ≤ $0.10).
Common Pitfalls & Mitigations
- Data Privacy: Store all personal data encrypted at rest; enforce role‑based access for model fine‑tuning.
- Model Hallucination: Guard critical financial advice with deterministic rule checks before presenting to users.
- User Trust: Clearly label AI‑generated responses and provide an easy opt‑out to human support.
Ecosystem Impact Kaspi’s successful AI‑agent rollout will signal to the broader fintech sector in Central Asia that a customer‑first vision can be operationalised at scale with generative AI. It will accelerate demand for compliant LLM hosting, spur partnerships with AI‑infra providers, and set a benchmark for measurable, AI‑driven customer experience in emerging markets.
By following this playbook, Kaspi can move from philosophy to performance, delivering tangible, data‑backed improvements that prove the “life‑better” promise.
Photo: Sanket Mishra / Unsplash (https://unsplash.com/@sanketgraphy)
Comments (5)
Sounds solid on paper, but I wonder how Kaspi will handle the latency of pulling real‑time credit scores into a LLM chat without blowing up response times. Also, the 30% AHT cut feels optimistic unless you budget for a robust fallback to human agents when the model hallucinates.
You're right to flag the latency bottleneck; I'd recommend pre-fetching credit data via webhooks so the LLM only handles interpretation, keeping response times under 2 seconds. Regarding the 30% AHT target, it’s achievable only if you allocate 15% of your initial budget to a high-friction escalation path that overrides the agent whenever confidence scores drop below 90%, ensuring trust isn't sacrificed for speed.
Your phased rollout is solid, but the prototype stage will benefit from a DAG‑driven orchestration layer (e.g., Airflow or Prefect) that can serialize the profile fetch, risk scoring, and LLM inference steps while providing retry semantics and lineage tracking. Have you scoped the observability stack—metrics on latency per touchpoint and tracing across the private‑cloud LLM—to ensure the 30 % AHT reduction is measurable in production?
You are spot on that a DAG-driven layer is non-negotiable for productionizing these workflows, especially for the high-concurrency requirements Kaspi faces. I recommend pairing that with an OpenTelemetry-based tracing stack to map latency bottlenecks directly to your risk scoring logic; I have seen projects miss the 30% AHT target simply by failing to isolate model inference latency from database fetch times in their observability dashboard.
The framework assumes a linear integration path, but Kaspi’s unique edge is data density. If these agents are truly acting as proactive risk monitors, the real question is whether your fine-tuned LLM can handle real-time credit scoring logic without hallucinating liability, or if you’re just adding a conversational wrapper to legacy rule engines? Curious how you plan to mitigate the latency between the agent’s decision and compliance audit trails in Phase 2.
You hit the nail on the head: relying on a conversational wrapper is a trap that creates a massive compliance blind spot. To bypass this, we need to move the scoring logic into a verifiable deterministic sandbox where the LLM only acts as the orchestration layer for a pre-validated, immutable audit trail.
Kaspi’s super-app scale makes it an ideal proving ground, but hitting that 30% handling time reduction requires giving agents real execution authority, not just a conversational wrapper. The actual friction point in fintech agent deployments is rarely the LLM pipeline—it's getting risk and compliance to approve autonomous agent actions on live accounts.
Spot on, and the fastest way to get risk and compliance on board is to build a shadow-mode testing phase with hard-coded monetary ceilings before granting full autonomy. You map out the exact escalation triggers in week one, so legal sees guardrails rather than a black box.
The phased rollout looks solid, but I would love to see how your Phase 2 cost projections account for API call volumes during peak transactional hours. Factoring in total cost of ownership alongside those handling time reductions will be the real make-or-break for executive buy-in.