Let’s start with the vendors

We still have a mix of the traditional, do-it-all vendors and the newer, cloud-based, API-swinging providers of specialist products

A cynical person may suggest simply renaming a product to include the term “AI” is not actually a solution if the product retains its 1990’s paradigm

(We’re being polite here, as some of them are still “modernising” the 1980’s stuff)

Before believing any of their assertions, we suggest you recall what they were saying about the cloud very few years ago – “you don’t need the cloud”

When doing your due diligence, look for phrases like

• “AI-powered” this that and the other” 
• “fleet of AI agents” 
• “the power of applied AI”

…and run away!

Accounting firms

We’re being bombarded by conflicting information and promises

Many of the trade conferences have been talking about AI for well over a year and the end doesn’t appear to be in sight

Have you decided whether AI is the solution to our problems or just an undercooked, overhyped solution looking for a problem? 

Cynare

AI – nothing new

It’s been around for over 70 years

Alan Turing and his mates worked on it in the 1940’s and 1950’s and it was adopted as an academic field in 1956

Cynare’s CIO has a 33-year-old AI qualification from his degree days

The big difference is we now have the computing power required for mainstream adoption (helped by the hype from the vendors who spot something new to sell)

 

AI – many flavours

You need to understand the differences between the following

• Machine learning 
• Large language models 
• Neural networks 

You’ll hear enthusiasts talking about

• Generic chatbots (ChatGPT, public data etc)
• Agentic bots (AI bots acting autonomously)  
• AI endpoints (plugging AI into workflows)
• Product Copilots (Chatbots with context of a single product/data source)
• Enterprise chatbots (eg ChatGPT plus private data)

AI – business and reputational risks

We suggest AI is not yet mature enough to form the core of operations and now is the time for targeted change rather than total transformation

The various failures of legal and accounting firms in their use of AI are well-documented, so we won’t say any more here

 

AI – financial sense or a bottomless pit

Accounting firm members aren’t known for their comfort with bleeding edge technology

Failure is common in AI projects 

• MIT suggests 95% have no cost savings or enhanced profits 
• Capgemini agree, saying 88% of AI projects fail to make it to production 
• Forbes has a simple view – 85% of AI models fail  
• S&P says 42% of pilots abandoned 

The challenge is the same as any technology project – people and process 

 

AI – needs proper policies and training

Please, please ensure you have the policies in place regarding your firm’s use of AI and be open with your Clients, insurers and commentators about this

Wouldn’t it be funny if firms used AI to write their AI policies…?

We then need to learn how to talk to large language models (massive databases of documents)

We need to understand the difference between information derived from searches within our own Firm’s domain and information gathered from anywhere in the world

 

AI – regulation

Don’t get me started!

However, I note the HMRC’s “Guidelines for using generative artificial intelligence if you’re a software developer” only came out at the end of January 2026

 

AI – Cynare’s products

We allow natural language search across our document management system (DMS) and a firm’s typical production and practice management systems

Yes – even the “traditional” ones!

We provide targeted training to decision-makers and Users