Epsom-based accountancy firm Love Your Accountants – or LYA – has proudly shunned the naming protocol of so many accountancy firms to better reflect the firm’s culture. “It’s all about client relationships,” explains Managing Director Rob Young. “This isn’t about an individual at the firm. We want clients and staff to feel part of an ecosystem.” 

With Young’s aim to deliver the best possible service to clients, the firm’s use of tech and AI is growing so staff are freed up to spend more time focusing on client service and support. With 90% of clients paying a set monthly fee for services, improving efficiency and accuracy is key, Young says.

The firm is in the early phases of exploring AI’s potential benefits to clients, but it is already reaping rewards, Young says. Up to now, the focus is on using AI to streamline admin and gradually adding workplace AI-driven tools using Google’s Large Language Models (LLMs).

“We reviewed a number of alternatives, discussed with multiple AI specialists who use Microsoft Copilot and Google Gemini. We self-taught ChatGPT and Grok and used the former to help come up with automated internal workstreams, including applications.” 

Because the firm already has contracts with Microsoft and Google, it felt more comfortable integrating Gemini with the newly built applications involving potentially sensitive client data, while ChatGPT is being used to brainstorm the fundamentals of the technological advancement. LYA is also considering taking out a contract for full use of OpenAI.

Use case 1: sorting and allocating snail mail 

LYA has used an external developer to build a portal for staff that uses workflows and Gemini to read scanned in mail, understand what it is, produce summaries, automatically identify which client it relates to and forwards it to the relevant member of staff for action. 

“That’s been a huge use of AI for us. Whether we have 100 letters or 1,000 letters, that automation takes five minutes. All our client managers can see immediately which letters are urgent, and if there’s anything that needs forwarding to clients. When someone’s on holiday, the workflows allow you to reassign to someone else so nothing gets missed.”

Young estimates the system easily saves the firm between 10 and 15 hours a week. Meanwhile, LYA has also used AI technology in Google LLMs for PDF optical character recognition (OCR). “It uses a separate system in the cloud that recognises the data in PDFs and converts it into written data. Search tool Everything allows the firm to find documents on the drive easily and quickly. Although Adobe already allows you to do this, it creates an internal process that can slow down the laptop of the user. By creating an external workflow, the user can continue working without any potential system slowdown.” 

Use case 2: receipt processing

Meanwhile, another tool, Veryfi, allows the firm to upload receipts and it will automatically pull out individual line items and code it in your finance system. “It gets all the data for you instantly and pushes it automatically through to Xero. This is about creating automation around a piece of admin, but in a way that better captures the data.”

Although the firm hasn’t used Veryfi to its full potential, the anticipated time savings and improvements to data accuracy look to significantly surpass that of any receipt-gathering software that allows you to build additional workflows into the same process, Young explains. “We think this will save the company significant time over the coming year, especially with the introduction of Xero’s internal AI called JAX, which works with Veryfi to automate the bookkeeping process.”  

Use case 3: summaries for management accounts

More recently, the firm has been using AI tool Syft, now owned by Xero, to help produce management accounts for clients. Syft pulls information in and summarises the performance of various aspects of a business instantly with the team. 

“When it comes to analysis, we do a lot of the heavy lifting but Syft uses AI to highlight what the biggest changes have been. We use it for anything that isn't opinion-based. For example, it can help us find break evens for clients, who the top creditors are, where their risk is in terms of foreign currencies.”

Managing data security

Concerns about incorrect information being presented means that use of AI has to date been restricted to administrative aspects, and areas where staff can sense check the output. “When using AI we don’t ask for an opinion, we ensure it’s always factual and refers to something specific – for example, HMRC manuals, data governance best practices, CIPD best practice – and we make sure it provides a reference point to show how it came up with the information. We treat it like an advanced Google search,” Young explains. 

Despite the firm’s gusto for AI, it’s important to proceed with caution, Young urges. In addition to his role at LYA, he is finance director of not-for-profit Icebreaker One, which specialises in consolidating data and using the correct governance to create trust frameworks in markets and ecosystems. “A lot of AIs use your inputs to train the models. In theory, those inputs are deleted when you delete the chat. Personally, I don't believe that is true.”

Another concern around the use of AI is data retention. However, this is a risk mitigated by daily backups going back up to five years. IT security has been beefed up using an external IT service provider, Indigo IT. “We use their coaching and their knowledge to help protect us. Whenever we’ve built anything or our process changes, they check the security.”

Staff can only access the system from registered IP addresses and the business is alerted to any attempts to access from other locations. The firm also uses two-factor authentication wherever possible and one-time passwords that generate every 30 seconds. 

For extra security, it is rolling out use of small USBs that require physical touch when attempting to log onto systems to prevent remote hacking. Furthermore, LYA keeps a live risk register that is updated monthly by its Technical Alignment Manager and, from 2026, it will hold quarterly C-level meetings with Indigo’s Virtual Chief Information Officer to ensure its cyber defences remain as up to date as possible. 

How to train yourself – and AI

Social media has been useful for AI training, and the firm has also used AI to devise training. “Courses out there are too expensive in our opinion and discussing AI with some of the people presenting workshops, we knew more about AI than they did!” 

Experimenting with AI tools is time well spent, with management playing with tools to build automated processes to save time. “A lot of time has been spent out of hours to understand it. We used it to free up the biggest time cumbersome tasks first.”

Anyone who thinks that using AI is a quick fix should think again, Young warns. He uses the example of writing blog posts as a case in point. “You have to invest a lot of time in it. It will generate a very AI-driven/computerised form. So you have to train it. We had a feedback loop, so if ChatGPT produced something, it would go to another AI to check it and give feedback. It will do that five instances until they're both happy.”

Young fully envisages further use of AI to help automate workflows built using AI. “We can use the AI to help build an Excel macro in 10 minutes instead of 30. We will use AI as a tool to help automation, not a tool that becomes the automation. Without the fundamental understanding of a task or process you can’t explain it to clients or make informed decisions yourself.”