Copilots are here, and very much here to stay. Enterprise-level artificial intelligence (AI) solutions, designed to support a huge breadth of activities, are now appearing in all manner of platforms. Only last month, Engine B launched a Copilot tool to assist with audit testing, initially focusing on lease accounting. 

“Copilots allow firms and professionals to maximise the opportunities provided by generative AI, such as those around reducing resource burdens and improving quality through automation,” Franki Hackett, Head of Audit and Ethics at Engine B, says. “Your new team member, the Copilot, provides you with intelligent insights that can then be considered in line with your experience and training, to provide the best possible service to clients.” 

When it comes to AI Copilots, none has garnered more attention in the business world than Microsoft 365 Copilot.

Microsoft 365 Copilot was launched on 1 November to much excitement, but also much disappointment – initially, the solution will only be available to users on certain enterprise licence types, with a 300-licence purchase requirement – a minimum cost of $9,000 (£7,159) per month. This puts it out of reach for the significant majority of smaller businesses. 

While such restrictions won’t last forever (some movement is expected in 2024), it has left many companies twiddling their thumbs and wondering what to do in the meantime. The good news is that there are very real, tangible actions that all organisations can and should be taking, to be ready for the Copilot era.

Data preparedness

At the heart of good AI is good data quality and governance. This is the case with traditional analytics and AI tools, and is no less the case with Copilots. Many AI solutions rely on the information that is used to feed the tool residing in appropriate, cloud-based solutions. To fully utilise Microsoft 365 Copilot, for example, organisations need to be using SharePoint and OneDrive, and not shared drives or other storage solutions. 

Having your data and broader information (such as policy documentation, templates, branding, reports and so on) in the right place, with correct tagging and categorisation, will allow the AI engine to be better trained on the context of the organisations in which they are deployed. 

M365 Copilot will be unaware of any files that exist outside the standard Microsoft Azure cloud computing ecosystem. Even if it is stored in SharePoint, if the AI engine cannot understand the context of the documents, it is likely to disregard them or, even worse, utilise them incorrectly. The exercise to tidy, tag and migrate documents currently residing on network drives will be no small feat for many organisations, and does not need to wait until the AI tools are available to be deployed.

Getting employees on the AI train

One of the key elements to the successful roll-out of any type of AI is people. Embracing technology that many legitimately fear will steal their jobs can be daunting, so it is worth taking the time to introduce the technology progressively so that it does not overwhelm. 

Training will play a key part in the adoption of AI, but so will exposure to the capabilities that AI Copilots can bring. As Hackett observes: “AI technology is arriving now, so we need to act now to ensure that all of us feel confident using it ethically and to deliver good quality.” 

Many organisations are reluctant to encourage their staff to experiment with freely available AI tools such as ChatGPT. They typically advise against providing them with any personal or business confidential information (which is in line with ICAEW’s guidance on Generative AI).

Even if M365 Enterprise users cannot get their hands on Copilot yet, there are AI tools available now that can be used safely. Bing Chat Enterprise (BCE) is a special, enterprise-safe version of the ‘Bing Chat’ solution, utilising the same engine as ChatGPT. 

Similarly, Microsoft Teams Premium is already available and delivers some AI capabilities directly in meetings. For many organisations, these are potentially powerful ways to start getting users familiar with interacting with generative AI tools at low or no additional cost. The risk is also minimal due to the security in place around the tools. 

BCE in particular mitigates many of the concerns that organisations have about encouraging the use of ChatGPT by protecting personal and organisational information that is loaded into it, not saving data after the chat session, and not using company data to train models. 

While it isn’t quite as far reaching or as powerful as M365 Copilot, it can carry out some of the same tasks, such as summarising internal documents or helping to analyse confidential data. It also brings much of the most appealing capability of ChatGPT around content and framework generation.

Apart from the Microsoft suite, there are plenty of other tools available that infuse AI. Canva – a design and productivity tool recently covered by the ICAEW Excel Community – delivers a number of AI capabilities for free, while AccountingWEB has produced an Accountant’s AI Toolbox listing a number of AI-enabled products. 

Is your data really ready?

The need for good data governance cannot be overstated. Proper access controls should be put in place to ensure that staff can only access the files that they need for their roles. This is because Copilots will leverage the same governance and controls and will deliver information to users based on what they have access to. 

According to Daniel Teacher, Managing Director of T-Tech, a specialist Microsoft implementation partner: “AI is hugely exciting for all. However, we are now giving a tool access to the entire network and company data set. Firms need to plan and think carefully how to manage this game-changing transition.”

Minimising the number of different versions of files will ensure that AI engines pick up only relevant information from within your organisation. In the age of the Copilot, emails and chat messages also start to come into play, so keeping your mailbox in order has never been more important.

While many businesses may want to use tools such as M365 Copilot now, this enforced delay is a perfect opportunity to really get their houses in order. A clean, tidy data and information repository, alongside staff who are comfortable interacting with AI, may take many months to put in place. 

Anything done now to prepare for M365 Copilot before it is released will be time well spent to maximise the return on investment when licensing opens up more. As Teacher notes: “Every department should be thinking now about how best to take advantage of this opportunity when it comes.”

Further reading: ICAEW Conference: AI, automated ESG and copilots