AutoRek and the money that refuses to match

The money moved. The records disagree. Somewhere between a bank statement, a payment file and an internal ledger, a transaction has become a small detective story. Multiply that problem across a financial institution and the back office acquires an awkward job: establish what happened, explain the difference and leave evidence that somebody checked.

AutoRek sells software for that job. The Glasgow company combines financial data management, automated reconciliation and controls for banks, insurers, asset managers and payments businesses. Its territory is the gap between systems that each tell part of the truth. A matched record closes one question. An unexplained difference opens several more.

That distinction now drives its AI ambitions. Matching has long been automated. Investigating the leftovers still consumes attention and specialist knowledge. AutoRek wants its ARIA technology to help with that work, while keeping decisions visible to the people responsible for them. For a business built around financial evidence, an impressive answer needs a convincing explanation.

The matching engine was already there

ClearBank offers a useful starting point because it already had reconciliation software. According to AutoRek’s customer account, the bank’s previous tool handled transactional matching but lacked workflow functionality, cash reconciliation and balance substantiation. Excel and manual work filled the gaps. As payment volumes increased, the arrangement became harder to manage.

AutoRek brought transaction and cash reconciliations into one platform across payment schemes and the bank’s core Bank of England accounts. Its case study describes real-time exception handling, configurable reconciliations and integration with Microsoft Azure. The intended economic benefit was specific: let the existing reconciliation team keep up with growth without immediately increasing headcount.

The revealing part is what required replacement. ClearBank’s problem extended beyond pairing two lines of data. It needed to organise the investigation, connect cash and transaction views, and make the result reviewable. AutoRek competes for precisely that larger assignment. A matching engine is useful; a finance operation also needs to know what remains unresolved.

Why 76% can be a useful number

Insurance intermediary Howden faced a related headache: spreadsheet-based insurance broking accounting and large quantities of data arriving in different formats. AutoRek implemented insurer statement and bank reconciliation, using automated matching and analysis of breaks, the industry’s term for records that do not agree.

The published case study reports a 76% automatic match rate and reconciliation processing time reduced by more than half. Those two numbers deserve to stay together. A system can leave exceptions for people to handle and still free substantial capacity. The useful measure is the work removed from the whole process, including reporting and investigation.

Funding Circle’s case points to a different improvement. Its previous approach aggregated reconciliation data at a high level. When balances diverged, tracing the cause was difficult. Multiple investors could participate in a single loan, and new products complicated the flows. More detail became a practical requirement.

Its AutoRek deployment replaced that aggregation with more granular reconciliation. The customer account reports work completed in around half the previous time, less dependence on knowledge held by particular individuals and easier cross-training. The lesson is appealingly ordinary: give people enough detail to locate the problem, then make the process repeatable by somebody else.

A Scottish software business, before the AI pitch

AutoRek’s origins predate the current vocabulary by decades. Gordon McHarg, George Urquhart and Steven Morosi founded Oasys Technology in 1994. It became API Software Limited the following year. The company’s history dates its core reconciliation solution to 2001 and the AutoRek product brand to 2002. A deal with then HBOS followed in 2003.

Scottish Equity Partners announced a growth equity investment in March 2021 to support product development and international expansion. Chris Livesey joined as chief executive in October 2023. In an interview with the investor, he described a company whose relatively modest size contrasted with the scale of the financial institutions it served. McHarg remained available as a non-executive director.

That history helps explain AutoRek’s expertise. Its product has developed around financial institutions with complicated records and reporting obligations. The company lists Bank of England, Lloyds Banking Group, Worldpay and Beazley among its customers. Its August 2026 acquisition announcement puts its reach at more than 100 clients in 15 countries.

There are lighter details behind the controls vocabulary. AutoRek’s published community material describes staff mentoring young people through Career Ready and MCR Pathways, birthday leave and support for professional development. It also records sponsorship of a Scottish beaver called Benjamin. Even a reconciliation company gets to have an animal anecdote.

The data has to arrive before it can agree

The core platform ingests financial data from sources including APIs, SWIFT and ISO messages, and uploaded files. Configurable rules validate, enrich and transform that information before matching and reporting. Audit trails preserve the history of actions. Dashboards let teams move from summary views into the underlying records that need attention.

The same foundation supports distinct operational jobs. An asset manager may need custody asset or client-money reconciliation. An insurer may need cash allocation or agreement between general and subsidiary ledgers. A payment provider must connect records across banks and schemes. The common requirement is a controlled explanation of how one set of records relates to another.

GoCardless illustrates the payments version. Its AutoRek case study describes daily reconciliation of millions of transactions across several banking schemes, sponsor banks and currencies. The platform supports the company’s safeguarding controls and changing funds flows. That is a continuing operating task: each additional bank or scheme gives the reconciliation process something new to accommodate.

Two named products extend the range. MION addresses crypto asset reconciliation, supporting precision of up to 18 decimal places and connections between blockchain data and existing financial systems. RegToolKit organises regulatory obligations, risks, controls, ownership and evidence. Its electronic breach register tracks remediation. Both put more structure around information that can otherwise become scattered.

ARIA has to show its working

Launched in September 2025, ARIA brings AI into matching and exception work. A July 2026 update added automated rule generation, broader pattern recognition and improved explainability. AutoRek reported a 95% reduction in time spent evaluating manual matches, with routine reconciliation configuration taking under 30 minutes. Those are company-reported outcomes for specific tasks, rather than a measurement of every customer’s total operating costs.

The appeal is easy to understand. Rules can process familiar patterns efficiently. An exception may demand context, investigation and a decision about who should act next. ARIA’s proposition is to help users interpret those situations and configure the system more quickly. Its practical test is whether the resulting work remains understandable and controllable.

“nearly right, is wrong in this world.”

Jim Sadler, AutoRek Chief Transformation Officer
August 2026 webinar recap

Sadler’s warning captures the constraint better than a string of AI adjectives. In the company’s webinar discussion with CPP Investments, governance and human accountability remained central. The speakers advocated improving data quality, standardising processes and beginning with meaningful operational problems. Faster decisions are valuable when the institution can still explain who authorised them and why.

Three ways to buy the next chapter

AutoRek’s August 2026 purchase of Grath widens the proposition. The combined portfolio spans AutoRek’s enterprise private-cloud platform, Grath’s multi-tenant SaaS and Topa, infrastructure that lets firms embed matching and agentic capabilities into their own systems. The acquisition also extends customer reach in the United States and the Middle East.

The business earns recurring software revenue and fees for implementation and configuration. Buyers are therefore purchasing both a system and the work required to fit it to their operations. Microsoft provides a marketplace route and preferred Azure infrastructure. Capgemini contributes implementation and advisory services. These relationships address the practical business of getting enterprise software purchased, configured and used.

Duco is one direct alternative, also advertising cloud reconciliation, AI-assisted setup and automated exception management. AutoRek’s case rests on its financial-services workflows, established customer relationships and range of controls. AI alone provides little separation when competitors also offer it. The comparison has to reach the awkward data, operating requirements and deployment choices of the institution buying the software.

Bring the awkward records to the demo

A practical reading of the customer stories suggests a better evaluation than watching a clean file match perfectly. Bring representative records, including differences that have previously consumed staff time. Trace one exception from ingestion to resolution. Ask another team member to repeat the process. Funding Circle’s experience makes team coverage as interesting as raw speed.

That trial should also make the handoffs visible. A reviewer needs to see the original information, the matching logic and the action taken on a difference. A useful automation project makes those steps easier to follow as well as quicker to complete. Treat this as an evaluation principle drawn from the customer accounts: speed has more value when the next person can understand the result.

The cost comparison should include the subscription, configuration, integration, training and continuing ownership of match rules. Put those alongside the time currently spent investigating and reporting. Howden’s result supplies a useful question: how much work disappears across the process? It does not supply a universal savings percentage or a budget for somebody else’s deployment.

The case is weaker where transaction volumes are small, existing accounting tools already provide adequate control or data ownership remains unresolved. Automation still needs usable inputs and clear responsibility for decisions. AutoRek’s customers show the opportunity when complexity has outgrown the old arrangement. The deciding demonstration is a record that will not match, followed by a clear account of what happens next.