Case studies

What the work looks like in practice. Client names are withheld under confidentiality; everything else is as delivered.

Banking · delivered

A global bank: 60% less processing time with AI search over its own data

Context

Delivered by Rafael Kemish as Senior Consultant, Industry Solutions Delivery at Microsoft, from discovery and proof of concept through to production and handover.

The challenge

The bank needed to process its cases faster, using information held in its own data.

What we built

  • AI search over the client’s own data, so answers are grounded in and cite the bank’s documents.
  • Zero Trust controls throughout: identity-based access, managed identities, private networking, a web application firewall, least privilege, and prompt-injection and data-leakage controls, reviewed with the client’s security team.
  • Client and partner engineering teams enabled to run and extend the solution.

Banking, healthcare, retail, manufacturing · delivered

Secure AI in production for large companies in Switzerland, Spain and France

Context

Two years at Microsoft (2024–2026) leading technical discovery, solution architecture and proofs of concept with account teams, partners and enterprise clients, then taking the solutions through delivery to production.

The challenge

Each client wanted AI agents working on real data, and every solution had to pass a review with the client’s security team before going live.

What we built

  • A single, governed entry point for every model call: routing, quotas and identity.
  • Agents with tool calling and MCP integrations into existing systems, plus retrieval with vector search.
  • The same Zero Trust baseline in every engagement, reviewed with the client’s security team.
  • Hand-over and enablement so client teams own the result, and field insight fed back to Microsoft product engineering.

Consumer brands · Spain and Portugal · proposal

Example engagement: one view of the business for a consumer-brands group

Status: Proposal under review. Shown as an illustration of scope and method, not as a delivered result.

Context

A growing group selling through its own Shopify stores and Amazon in Spain and Portugal, plus a department store and distributors, with an on-premises ERP and separate accounting.

The challenge

Each channel has its own system and data. Stock, sales by channel and margin are only known after the month-end close, the same price or stock change is repeated in every store, and EDI with the main distributor depends on one person.

What we built

  • A central platform with one record of products, stock, orders and invoices, built in modules from lowest to highest risk.
  • Initial package of about three months: real-time visibility, all channels connected with stock and price sync, and the ERP connected with automated migration.
  • Management, finance, operations and sales ask questions from Claude, ChatGPT, Microsoft Copilot or Grok, each within their own permissions, with approvals in Teams.
  • Later modules: B2B sales app and portal, warehouse, EDI, and finance with Spanish and Portuguese e-invoicing.

More detail on any of these is available in a call.

Bring one slow process. We will tell you honestly whether AI can compress it.

A free 30-minute call. No slides, no commitment.

Book a free 30-minute call