Stitch, a banking operating system provider active in Saudi Arabia and the UAE, has launched Stitch AI, which puts AI agents directly inside the operating system banks and lenders use to run accounts, cards and loans. Instead of chatbots sitting on top of copied data, the agents act on the core system itself, within each bank’s permissions, approval rules and audit trails. The product has three parts: a Copilot for staff, ready-made Stitch Agents and an Agent Harness for building custom agents.
Quick Take
- The launch: Stitch AI, announced November 17 and 18 in Riyadh and Dubai.
- Copilot: Staff ask in plain language to fetch data, run a task or change configuration, including setting up products with limits, pricing, fees and eligibility. Every change waits for approval.
- Stitch Agents: Ready-made agents for onboarding, credit decisioning and collections.
- Agent Harness: A framework for banks to build, test and deploy their own agents with permissions, guardrails and audit trails.
- Data protection: ID numbers, account numbers, phone numbers and addresses are masked before any model sees them. Banks choose the model and can run it in-country or in a private cloud.
In plain English
Many banks have tried AI, but most projects stop at a demo because the AI cannot safely touch the systems that hold real money and customer data. Stitch builds the core banking software itself, so it can let AI agents work directly inside it, with each agent having its own identity, limits and a full record of what it did and who approved it. Sensitive customer details are hidden from the AI model before it sees any data.
What happened
According to Stitch, every agent has its own identity, scoped permissions, hard limits, maker-checker controls and a full audit trail. Every question, action and approval is logged with the reasoning behind it, so risk and compliance teams can see what happened and who signed it off. Agents act through the same services the bank’s products already run on, with no copies of data and no separate connectors. Banks can switch models without rebuilding anything.
In the company’s own words, all three parts operate within the limits the bank sets: “Above them, they ask.” The launch comes with a campaign film called “Anyone can be a hero”, starring actor Patrick J. Adams.
Background
The launch lands in a busy week for agentic banking: yesterday we covered Savana’s Acta, which carries out 139 banking tasks, and Model ML raised $250 million to put personal AI agents inside banks. Our Banks Using AI tracker collects how institutions deploy AI.
Why it matters
Regulators and bank risk teams care less about how clever an AI is and more about who approved what. Building governance into the core system, rather than bolting AI onto the side, is the approach most likely to get agents past compliance reviews. Core banking vendors that offer this natively could speed up AI adoption at mid-sized banks that cannot build it themselves.
What’s still open
- Which banks are using Stitch AI in production. The company did not name customers.
- Which models are available by default and how pricing works.
- How regulators in Stitch’s markets will treat agent-made credit decisions.
About the analysis
Researched and written with AI assistance under the editorial responsibility of the i10X News team. The title image is an AI-generated image. Read our AI Use Policy. Spotted an error? Request a correction.
This independent i10x analysis is based on Stitch’s press release and its Stitch AI product page. Capability descriptions are company statements.
Sources (plain text, no links):
- Stitch press release: “Stitch Launches Stitch AI, Bringing Agentic AI Directly Into Banking Operations” (November 18, 2026)
- Stitch product page: Stitch AI
i10x Perspective
Banking AI is moving from chat to action. The vendors that control the ledger have an edge, because they can enforce rules where the money actually moves. More on our Finance page.



