Workflow Automation In London: What The Sector’s Cloud Outage Exercise Means For Firms Using AI

38 - Largest banks and market infrastructure operators in the cloud outage exercise (Bank of England, 9 October 2026)  ·  8 Oct - Date of SIMEX26, hosted by the London Stock Exchange Group (Bank of England)  ·  63% - Firms citing AI as a key UK systemic risk, up 32 points (Bank of England, 30 September 2026)  ·  75% - Firms citing cyberattack as a key UK systemic risk (Bank of England Systemic Risk Survey, 30 September 2026)

What happened in the sector’s cloud outage exercise?

The Cross Market Operational Resilience Group, known as CMORG, ran a sector-wide simulation called SIMEX26 on 8 October 2026. The Bank of England said the scenario was a “global disruption to cloud-based services” and was not based on any specific threat information.

Taking part were “the 38 largest and most systemically important banks and market infrastructure operators”, along with HM Treasury, the FCA and the wider sector. The London Stock Exchange Group hosted the exercise, and the Economic Secretary to the Treasury, Lucy Rigby, opened it.

Katharine Braddick, Deputy Governor for Prudential Regulation and chief executive of the PRA, said: “This was an important demonstration of the UK financial sector’s commitment to operational resilience.” The Bank was clear about what comes next: “The Bank now expects firms to take action based on the exercise.”

Why does a cloud outage exercise matter to smaller firms?

Smaller firms were not in the room, but they depend on the same infrastructure. A wealth manager’s platform, a broker’s order system, a payment firm’s processor and almost every AI model a firm might use all run on a small number of cloud providers. When one of them has a bad day, everyone downstream has the same bad day.

The Bank’s own description of CMORG’s recent work makes the link to AI directly. It lists “strengthening third party engagement and resilience” and “enabling secure innovation in response to AI and other emerging technologies” among the group’s priorities.

How does agentic AI change operational resilience?

Agentic AI adds a new kind of dependency. A workflow that used to rely on a spreadsheet and a file share may now also depend on an AI model provider, an agent platform and the connectors between them. If any of those fails, the workflow stops, and the people who used to do it by hand may have lost the habit.

The Bank of England’s Financial Policy Committee said on 30 September 2026: “Rapid advances in AI capabilities have increased cyber and operational resilience risks.” In the Bank’s Systemic Risk Survey published the same day, 75% of firms cited cyberattack and 63% cited risks surrounding AI as key sources of risk. The Financial Stability Board published finalised sound practices for responsible AI adoption on 12 October 2026, and its Chair wrote that they are “adaptable to the size, type, and level of AI adoption within financial institutions.”

What should a firm running workflow automation in London do after SIMEX26?

Most of the response is good housekeeping. For each automated workflow, from the daily mismatch report to the safeguarding reconciliation, a firm should be able to answer six questions:

1. Which cloud services, AI models and connectors does this workflow depend on, and who provides each one?
2. If one of them is unavailable for a day, how do we do the work by hand, and when did we last test that?
3. Where are the inputs and outputs stored, and can we reach them during an outage?
4. How do we catch up afterwards, and who approves the catch-up before anything is sent or submitted?
5. Which regulatory deadlines would a one-day outage put at risk, such as a daily safeguarding reconciliation or a transaction report?
6. Who in the firm owns the workflow and decides when to switch to the manual fallback?

Reconciliation automation is a good place to start, because the deadlines are daily and the manual process is well understood. Write the fallback into the runbook, run it once a quarter, and record the result for the board.

What does a manual fallback look like in practice?

A manual fallback is a written procedure that lets a trained person produce the same output without the automation. For a reconciliation it might be the export steps, the spreadsheet template, the checks to perform and the sign-off required. It should sit with the runbook, be owned by a named person, and be tested on real data as well as written down.

Every agent we build goes live with a runbook and a manual fallback, alongside a named owner and an evidence file. For London firms planning AI automation more broadly, our explainer on the Bank of England’s AI warning covers the wider risks, and our guide to AI agents for financial firms covers the controls around each agent.

How can a firm test an automated workflow, and its fallback, on its own data?

BraivIQ’s 14-day Proof Run works read-only on a firm’s own exports, times the current manual process on real cases and finds every mismatch. That baseline doubles as the first draft of the manual fallback, because it records exactly how the work is done today.

The senior team behind BraivIQ has delivered digital products for six UK banking brands on one shared, tested foundation, so resilience planning is part of how we build from the first day.