AI Data Preparation for Accountants: Get Data Ready

When accountants discuss AI, the conversation often revolves around whether machines will replace them, with AI 'taking over.' However, before considering the replacement of a human's contribution—built on years of experience and specialist skills—it's important to understand how data feeds and creates AI.

This article explores the relationship between AI and data, and the many beneficial ways AI can be used to remove mundane tasks and improve accuracy. Once understood, it's clear how AI can help accounting firms grow their business. Sometimes, this is by creating capacity to support new business, but more often, it's about helping firms complete work to a higher quality using fewer resources, while maintaining profit margins.

What Would a "Magic System" Look Like for Accountants?

Accountants in the UK often seek two main outcomes:

1. 1.Linking all the data everywhere – so they don't have to input or update numbers and descriptions, but instead have data that's always automatically available and up to date.
2. 2.Ensuring accuracy and context – concern that AI won't get it right everywhere because every client file is different and not all context is available to the system. Who verifies and provides judgment if you automate this?

The answers to these observations are not straightforward, but there are promising possibilities.

Linking Data: The Role of Cloud Tools and AI

Linking data requires cloud tools and open APIs. Sharing data between tech providers has become standard practice, but this doesn't mean the data is ready to be pulled into automated workflows. Often, data is messy, incomplete, and in a custom format. Clients may use different bookkeeping software, custom general ledgers, different reporting periods, and specific naming conventions. If data flows in automatically but differently for every file, how does that help?

The solution is to use two key tools: cloud syncs and AI.

• ​Cloud syncs ensure that, regardless of the bookkeeping or practice management software, your data is transformed into the same structured data model. However, naming conventions and formats are often client-specific, requiring some interpretation.
• ​Artificial intelligence can make a best guess for these, using what is called data enrichment. By adding meaning to all imported data (such as general ledgers), the system knows how and where to use this in later steps. Because AI is probabilistic and can make mistakes (though unlikely), users are asked to confirm this enrichment once. This process takes only a few minutes but saves hours and days in the future. More than half a million general ledgers have been enriched this way in the UK alone last year.

Once data is always available and standardised (through AI enrichment), Silverfin only needs to standardise the accounting processes once, and the system will know how to use all process steps. This AI is zero-config, meaning it works on whatever data comes from any data source.

Addressing Objections: Linking Data and Adapting to Client Differences

The same two objections arise: can we link this data between each step of the process, and how can this work if every client is different and requires different reports, reconciliations, filings, etc.? Again, the answer is to link the data between process steps and use AI to enrich.

In Silverfin, you only have to fill in data once. For example, if you add a reconciliation in a working paper, it automatically flows through to accounts production and corporate tax. This should be the case with any connected system.

You can also adapt the process for each client so you only fill in what is absolutely necessary. This consistency not only makes the process smoother, but also improves review and handovers between colleagues, enabling much-needed work flexibility.

Finally, all this enriched data allows you to be future-proof, supporting comparative analysis, advanced analytics, and even a powerful AI Assistant. However, that is a topic for another time. This article focuses on the first stepping stone of AI, with more advanced steps to come.