AI in action at Irwin Mitchell: from prompt engineers to productivity pioneers
Originally published in Briefing October
Graham Thomson, chief information security and innovation officer at Irwin Mitchell, shares how his firm is shaping its approach to understanding and embracing the potential of generative artificial intelligence in legal through close collaboration with Microsoft and Peppermint Technology
Irwin Mitchell chief information security and innovation officer Graham Thomson has long looked to artificial intelligence to model predictions, with a focus on critical security oversight. Now, with the rapid rise of generative AI (genAI), he’s turning attention to securing a future for the firm that liberates people across departments from a multitude of the more mundane parts of processes — embedding scalable efficiency gains, and powering personal productivity, with appropriate management of the risk that can accompany this technology.
“The firm needed secure and private AI tools because of the sensitivity of the data we handle,” he says. This has meant an initial “build before buy” strategy. “We’ve put together a very simple chatbot using Microsoft’s OpenAI models. As the models get better, the capabilities of the tool really grow. We also have Microsoft Copilot — the corporate version that’s secure and private, so ideal for a professional services firm.”
At the same time, a cross-functional AI working group was formed to filter the range of possible use cases and prioritise attention — resulting in these five key areas for focus:
Task automation — managing the firm’s more repetitive tasks, allowing people to focus on higher-value work.
Text summarisation — parsing long documents quickly and accurately, providing more significant time savings.
Text and image creation and modification — “text modification, in particular, is a strong use case in drafting and refining documents”, says Thomson.
Data and text analysis — AI can now analyse smaller datasets in tables or spreadsheets — “everyday data analysis that saves time”.
Code generation and analysis: “AI can help to generate code, whether Excel formulas, Power Apps or KQL. It simplifies tasks for people for people who aren’t coders but need to use these technologies in their work,” he says.
The firm also built up a centralised prompt library to guide people around the identified use cases with around 50 pre-set prompts — but Thomson says people are finding them less necessary as AI evolves: “You don’t need to prompt-engineer. The models are evolving, and the tools now do a good job with plain English commands.”
Partnership with Peppermint on the case
Alongside Microsoft, the next phase of Irwin Mitchell’s AI journey is closely connected to that of its case management system (CMS) from Peppermint Technology — the trio have taken part in workshops to understand possible ways forward.
“The real endgame for us is where AI becomes fully integrated into the CMS,” explains Thomson. “This gives us a much richer set of use cases, where it doesn’t just manipulate text — it can see all the data; emails, case files, and SharePoint documents. That’s when it becomes extremely powerful.”
Future AI won’t just wait for you to input prompts — it will also suggest things as you work. While drafting an email, for example, it might suggest information you didn’t know you had. You open an email, and it suggests relevant information about the client and who else has worked with them, or flags a document you may want to review. That kind of intelligent help can ultimately enrich client relationships
Take document analysis, for example: “There’s a case where we need to analyse 100,000 documents for insights. With AI we can extract valuable data that traditional methods couldn’t handle.
“It can also recognise that a lawyer is writing a clause, and although it may be very specific to the client situation, suggest similar clauses from past cases. That’s a huge potential time-saver in legal drafting.”
Prompts for content in context
As AI becomes ever-more proactive, another area for efficient improvement is a firm’s maze of communications. Thomson continues: “Future AI won’t just wait for you to input prompts — it will also suggest things as you work. While drafting an email, for example, it might suggest information you didn’t know you had. You open an email, and it suggests relevant information about the client and who else has worked with them, or flags a document you may want to review.
“That kind of intelligent help can ultimately enrich client relationships. Peppermint is moving towards becoming more of an intelligent assistant.”
There’s even scope for prompts from AI to improve the data in the client relationship management (CRM) system itself. “It could proactively suggest entering a phone number, or correct details from recent emails, updating the CRM with a click,” says Thomson.
A collaborative ecosystem for AI exploration
Thomson says the benefit of access to the Microsoft and Peppermint relationship and joint roadmap can’t be overstated. Through the workshops, the firm has played a genuinely active role in shaping how tools will be developed. “We’re not just buyers,” he stresses. “We have a seat at the table, providing feedback and influencing development of new features. It’s a two-way street with a major leader at turning things into usable products within a low-code environment.
“We’re currently exploring how Peppermint can integrate more tightly with Copilot, such that users don’t have to think about which tool to use. AI just becomes part of their workflow, seamlessly embedded in the system.
“Copilot works on data across systems — your OneDrive, SharePoint, Outlook — and that’s when the value really ramps up, the more data it sees the better. It can reveal insights you couldn’t find manually without spending hours sifting emails or files.”
Regarding Microsoft 365 Copilot, that tool comes at a price. “It’s a fantastic tool, but giving everyone access would be too expensive, so we’re rolling it out team by team, prioritising groups such as executive assistants, who can use it to summarise video calls and transcripts and identify goals or actions quickly,” he says.
“You can mine richer insights by accessing all your data but need to make a solid business case for deploying it widely across the firm as not everyone will use it.”
Accelerating ‘non-chargeable’ administrative tasks like this are another clear win in the near term, but at the same time Thomson has his security hat firmly on as the collaboration — and work of his working group — unfolds.
“There’s always a risk that sensitive business data becomes too accessible — people should only have access to the data they’re supposed to. AI makes it easier to search,” he acknowledges.
On the other hand, he sees certain potential for AI to continue contributing positively to firms’ ongoing risk-management effort — in client onboarding, for example, with “conflict-checking insights that go beyond what traditional tools offer”.
“If you onboard a client but don’t input all details, it can also prompt you to fill in missing data or might auto-fill from emails and ask you to confirm the changes.”
The bottom line is that groups right across this forward-looking firm are set to benefit as the opportunities continue to surface — from clear timesaving on certain marketing, HR or EA tasks, to lawyers leveraging the firm’s extensive knowledge management assets within their daily workflows more productively. Exploring alongside the deep technical expertise of Microsoft and Peppermint helps Irwin Mitchell remain on the front foot in a rapidly evolving innovation ecosystem.