Reconciliation Automation Moves From Flagging Breaks To Fixing Them: What UK Firms Should Take From Sibos 2026

97% - Cut in investigation time per exception, 14 minutes to 30 seconds (Smartstream pilots, 2 June 2026, vendor-reported)  ·  13% - UK payment firms doing daily safeguarding reconciliations (Kani Payments, 6 May 2026)  ·  64% - Firms relying on spreadsheets for monthly safeguarding returns (Kani Payments, 6 May 2026)  ·  21% - Firms able to produce a complete safeguarding evidence pack in real time (Kani Payments)

What is reconciliation automation?

Reconciliation compares two or more records of the same thing, such as a bank statement and a ledger, or the firm's positions and the custodian's, and explains every difference. Reconciliation automation uses software to do the matching, so people only see the items that do not agree, often called breaks or exceptions.

Matching has been automated for years. The slow part is what comes after: working out why each break happened and what fixes it. That is the part AI agents are now aimed at.

What did Smartstream launch around Sibos 2026?

Sibos, the annual financial services conference run by Swift, took place at the Miami Beach Convention Center from 28 September to 1 October 2026. Five days before it opened, on 23 September 2026, Smartstream announced a new version of Smart Reconciliations Premium.

In it, Smartstream says, “Smart Agents now investigate, resolve and progress exceptions autonomously.” The company describes this as moving from AI that flags breaks for staff to AI that acts on them. It also claims “approximately 25% lower total cost of ownership and 50% more throughput on hardware clients already own.” Those are the vendor's own figures.

The results behind the launch come from earlier pilots. On 2 June 2026 Smartstream reported that “Pilot results demonstrate a 97% reduction in investigation time per exception, from 14 minutes manually to 30 seconds.” It described these as Tier 1 pilot deployments and did not name the institutions.

Can AI do reconciliations safely?

It can do much of the work safely if the controls are designed in. Smartstream's own release stresses them: “Every agent action is logged, explainable and subject to human accountability.” Its June announcement says “configurable human-in-the-loop approvals ensure analysts remain in command of higher-risk decisions”, and that its agents run in assistive and autonomous modes.

Timing pressure is part of the pitch. Smartstream's release says: “Settlement cycles compressed to T+1 leave no time to chase breaks across systems and inboxes.” The UK's move to next-day settlement is set for 11 October 2027, according to the Accelerated Settlement Taskforce, which puts T+1 settlement automation on the agenda for brokers and asset managers.

The choice between assistive and autonomous is the one that matters most for a smaller firm. In assistive mode the agent investigates, explains and proposes, and a person decides. In autonomous mode the agent resolves breaks itself within set rules. For a regulated firm with a small operations team, assistive mode is easier to evidence to an auditor and easier to switch off without losing control.

Why are UK safeguarding reconciliations under pressure?

Since 7 May 2026 the FCA has required payment and e-money firms to carry out daily checks that the right amount of customer money is safeguarded, alongside annual safeguarding audits for most firms and monthly reporting (FCA, rules announced 7 August 2025).

Many firms started from a long way back. Kani Payments' research, published on 6 May 2026 and based on 75 compliance and finance professionals at FCA-regulated payment and e-money institutions, found that “just 13% are performing daily safeguarding reconciliations.” The majority worked weekly (53%) or monthly (28%), and 64% still relied on spreadsheets for monthly safeguarding returns. Only 21% could produce a complete safeguarding evidence pack in real time.

The FCA expects audits to find problems. Its Payments Regulatory Priorities report of March 2026 says that as standards have been strengthened, “we may see an increase in ‘adverse’ audit opinions in the short term.”

How do you automate reconciliation without losing control?

1. Measure the manual process first. Count the breaks per day and time how long each one takes to investigate, so any gain can be proved.
2. Let the agent prepare, not decide. It matches records, investigates each break and proposes a likely cause and fix, and a person approves the fix.
3. Keep the evidence. Every match, explanation and approval is logged, which is what an auditor or the FCA will ask to see.
4. Keep a manual fallback. If the agent is switched off, the team must still be able to run the reconciliation that day.

That is how we build safeguarding reconciliation automation and the daily mismatch report. Our agent prepares the daily safeguarding reconciliation and writes up every difference, and a person approves it before anything is recorded. Our agents never move money.

Where should a UK firm with 10 to 250 staff start?

With one reconciliation and real numbers. BraivIQ offers AI automation in London and across the UK for financial firms, and every engagement starts with a 14-day Proof Run: a read-only run on up to three of the firm's own exports that finds every mismatch with its likely cause and times the manual work.

If the numbers are there, the agent goes live in 60 days after a parallel run on live cases. Our senior team learned how to build exception queues that compliance teams trust on real-time trade surveillance for a global investment bank. For firms where agent-initiated payments will add volume, see our guide to agentic commerce.