AI for insolvency professionals: how to get started safely, ethically and effectively.
16 September 2025
By James Grant, Co-founder at iwantmore.ai.
There's a lot of noise around AI in the world of professional services and insolvency is no exception. Lots of firms, especially those at the larger end of the spectrum, are shouting from the rooftops about how much they are spending on implementing AI.
You can see why. There is a lot of scope for using AI within professional services. From document analysis to forecasting outcomes, AI is seen as both a disruptor and a potential enabler. But for many licensed insolvency practitioners (IPs), particularly those operating in tightly regulated environments, the question is how to start using it safely, ethically, and effectively.
At the recent R3 Annual Conference, I spoke on a panel with Charlotte Hill from Penningtons Manches Cooper and Tony Walker of Armalytix, where we focused on what was practical and achievable without needing to upload vast quantities of sensitive data or trying to replace complex human judgement. There is a tendency for many businesses to immediately focus on how they can use AI to solve their biggest, riskiest challenges. But there is so much you can do now that will have a real impact on personal productivity without exposing your business to risk.
Here's how firms and individuals can begin to explore AI safely.
Start small
The first rule of using AI in a sensitive profession is: you don't have to give the model all of your sensitive data.
For example, if you are using Microsoft technology, then Microsoft Copilot will likely be your first port of call when it comes to enterprise AI. This is already integrated into Microsoft 365 (Word, Excel, Outlook) and operates within your M365 environment. This means that no data is moved outside your Microsoft cloud tenant and your data is not used to train the underlying models. The data you can access via Copilot is also the same data you can already access, and it honours your existing security permissions.
Using a tool like Copilot is a great place to start and would allow you to do things like:
- Summarise email chains, meeting notes or case documents
- Generate first-draft emails or letters from a few bullet points
- Reformat and analyse spreadsheets or generate formulas from plain English
- Carry out deep research on different businesses or industries
- Create presentations and proposals in a fraction of the time
Because it's built into the Microsoft ecosystem, there's no need to export data to third-party tools. That significantly reduces regulatory concerns and tends to make IT, risk and regulatory teams more comfortable.
A number of IPs I have spoken to use other tools like ChatGPT or Claude. These can still have a part to play in your day to day, you just need to be clear about what version you are using (enterprise vs free), what their data controls are and be conscious about what you are using them for.
Define what each tool can be used for (or not)
Every firm should understand what tools can be used and for what. Map out your use cases and then clearly define the areas that are safe for implementing AI without risking confidential or sensitive data.
Here are just a few example areas:
- Internal knowledge support
- Build agents over your own firm's staff handbook, process manuals, templates, or FAQs and use them to surface relevant information faster.
- Client / creditor communications
- Use AI to help rewrite or reformat standard comms: emails, letters, status updates.
- Learning and training
- AI makes an excellent tutor. It can quiz you on new legislation, summarise guidance notes, or explain complex insolvency terms in plain English.
- Marketing
- Use AI to help write blog and social media posts and design infographics.
These just scratch the surface of the possible use cases and the key here is that none of these use sensitive client data, yet all offer real productivity gains.
Keep your ethics and regulators in mind
Of course, the risks are real. IPs are dealing with confidential financial data, distressed companies, and vulnerable creditors. Professional judgement is not optional.
So how do you stay on the right side of your ethical and legal obligations?
1. Never upload confidential data into free tools.
ChatGPT, Claude, Gemini, these tools are useful for certain things, but unless you're using paid enterprise versions with signed agreements and secure storage, they are not compliant environments for client data.
2. Document your usage policy.
If your firm doesn't yet have an internal AI usage policy, now's the time to create one. Set out what tools are allowed, for what purposes, and what kinds of data must not be used. This protects both the firm and its people.
3. Keep humans in the loop.
Use AI to augment your work, not automate decisions. Everything that goes to a client, a court or a regulator should still go through human review.
4. Be transparent.
If you're using AI to generate or help draft something, don't try to hide it. Be clear in your engagement letters about the use of AI. The ICAEW and IPA have made it clear: disclosure, professionalism and judgement still apply in an AI-assisted world.
Don't wait for perfect
Many IPs are still waiting for the ‘perfect' AI product tailored to the insolvency industry. But you may be waiting a long time.
The truth is: most AI tools today are general-purpose and work across sectors. There is unlikely to be a magic insolvency-specific tool that lands next month. And if/when it does arrive, it will likely be built on the same foundations, GPT-4, Claude, Gemini, etc.
So start now. Use what exists. Learn how it works. That way, when a tailored solution appears, your team won't be starting from scratch.
Buy the same way as before
It's so easy for anyone to sign up to any of the well known AI tools. So easy in fact that nearly one in eight of the world's population have used ChatGPT, according to Sam Altman in an interview at TED 2025 back in April. But don't think just because these tools can be bought for $20 per month online that the level of IT security and data privacy scrutiny should be any less. Treat them with the same levels of rigour that you would when buying any other software. Complete your data protection impact assessments, review privacy policies etc. and, if in doubt, seek legal advice before you enter into any such contract.
And knowing that these tools are cheap and easy to buy, if not free, do you know what your team are already using?
Be clear that many, if not most of your team will already be using AI tools to help them in their day-to-day work and these may not be tools sanctioned by your business. One of our clients thought that none of their team were using AI. It turned out they all were. All 120 of them. And amongst them they were using 25 different tools. All paid for on credit cards, none sanctioned by AI. Why? Because it made them more productive. Have a clear plan for how you will address this.
Focus on AI literacy
AI policies are essential and should be part of every business' governance arsenal, but there is no replacement for training. If your teams are AI literate, they are more likely to understand the risks and therefore less likely to put yours or your client's sensitive data at risk.
AI literacy should be the core part of a wider change management programme that also includes clear communications. Don't leave anything to chance when it comes to your data security. Communicate to your teams regularly and clearly about what they can and can't do. Most issues arise from lack of knowledge rather than anything more sinister.
What next?
If you want to start using AI in your firm but aren't sure how, here are five practical steps:
1. Identify and prioritise your use cases for AI or automation
2. Choose a tool to start using e.g. Microsoft Copilot
3. Draft an AI policy and communicate it to your team
4. Train your team on AI - all of them, not just a select few
5. Build an AI working group to lead AI advancement in your firm
The best use cases for AI adoption aren't necessarily the biggest, shiniest, most expensive projects. It's the smaller, easier yet still impactful productivity gains that everyone can make. These can be made safely, ethically and effectively without exposing your business to undue risk.
Final word
AI isn't magic, and it shouldn't be seen as a threat to your human judgement or professional rigour. It's a tool, just a very smart one that IPs can and should learn to use. With the right guardrails, you can benefit from its huge power without compromising your professionalism or your duty of care.
