Clinically Intelligent is written by Juwon Akinyande in a personal capacity and is not affiliated with, endorsed by, or representative of Barts Health NHS Trust or any other NHS organisation. Content is for general educational purposes only. It does not constitute clinical, legal, or information governance advice. Before applying any guidance to your own practice, consult your Trust information governance lead, your Caldicott Guardian, your line manager, and your professional body.

I have been avoiding Copilot.

I remember it as Bing. I was not impressed. Claude and ChatGPT worked better for how I write, how I think and what I do in my week. There was no reason to pick up a third tool that would do the same jobs less well.

Then a pattern started showing up. Every time I had a conversation about AI at work, Copilot came up. Management. The AHP Informatics Officer at my sister site. The AHP Informatics Board. The education academy. Colleagues. Some had used it. Others were about to. The ones using it were not choosing it over Claude or ChatGPT. They were using it because it was the only AI tool the NHS had approved.

Then the NHS England announcement about the 505,000 user rollout landed in Issue 10.

And I noticed something about my own position. I write a newsletter about AI in healthcare for the people who work inside it. I have been recommending governance frameworks, testing tools, running analyses. The whole time I have been relying on consumer AI. The tool the NHS has actually approved and is actively rolling out, I have never opened.

That is what I am fixing this week. This issue is the first honest look at Copilot from someone who has been sceptical of it. Not a review. A first exploration.

What Barts has given me access to.

The version I have is labelled Copilot Chat (basic). That is the free tier of Copilot. You sign in at copilot.microsoft.com with your NHS account and you have a chat tool comparable to ChatGPT or Claude, running on Microsoft infrastructure under NHS enterprise terms.

That distinction matters because Copilot is not one thing. Issue 10 walked through it. The free Copilot Chat is a tab you visit. The paid Microsoft 365 Copilot sits inside Word, Outlook, Teams, Excel, and PowerPoint. Different products. Different work. When people at your Trust say “Copilot,” ask which one they mean.

Barts is on the free tier. The paid version is what was announced in the 505,000 user rollout. It is coming to more staff over the next year. Right now, at my Trust, the tool available to me is the version that has been there all along.

What happened when I actually tried it.

I ran a small test. I had done a functional and cognitive assessment on a composite scenario. Nothing that touched a real patient. I gave Copilot a plain prompt. Summarise my findings as if you were an OT.

The output surprised me. It was clean. It was structured the way I would structure a summary. It picked out the clinical themes I had prioritised. It read like something an OT colleague would write, not like generic AI output pretending to be clinical.

That is one test. Not a workflow. Not a habit. One prompt and one summary. But it was enough to tell me the tool can do a specific job for the specific work I do.

The IG discipline for that test was the same as every other AI tool this publication has covered. Composite scenario. No patient identifiable information. Safety Check applied before I opened the tab. The tool being a Microsoft product under an NHS enterprise account does not change the discipline. It just changes the paperwork behind the tool. The clinician is still responsible for what they put in.

The reason I want you to notice Copilot is not just the chat. It is the wider Microsoft ecosystem it sits inside.

Two apps caught my eye when I opened the environment properly for the first time.

OneNote. Described as a digital notebook you can share with your team. I have not used it. The first page tells me you can build visual notebooks, share them with collaborators, add voice transcription and organise information the way you think rather than the way a document forces you to. For someone who prefers visual thinking, that is potentially useful. For team collaboration on complex clinical topics, it is worth a proper test.

Power Apps. Described as a way to build mobile and web apps using data your organisation already uses. I have not used it. The learning is built into the app through video tutorials. What interests me is what it makes possible. A clinician with a specific service problem could build a small tool to solve it, without needing to commission a full IT project. That is a different category of thing from a chat tool.

Both of these are worth exploring properly and I will report back on whichever earns a full issue.

One small note on Power Apps at Barts specifically. The build with Copilot template picker is turned off at my organisation. That means if you want to build an app, you build it from scratch. Not from a template. That is something to keep in mind if you are planning to try it. My organisation has enabled some Copilot features and disabled others. Yours will be doing the same, and the specifics will be different. Worth asking your digital lead which features are on and which are off before you start.

Information governance.

Microsoft has stated publicly that prompts and responses processed under Microsoft 365 Copilot enterprise licensing are not used to train their foundation AI models. Whether the enterprise deployment has been assessed against DCB0129 and DCB0160 clinical safety standards is a live question that varies by Trust and by use case. Ask your digital lead where your Trust has got to on that.

That does not remove your responsibility. It changes where the lines sit.

For consumer AI, the rule stays what it has been. Composite scenarios only. No patient identifiable information. The Safety Check from Issue 07 applies as it always has.

For enterprise Copilot in your Trust account, patient information handling follows your Trust’s usual clinical governance rules. Your Trust’s Copilot guidance is the rule that applies in your Trust.

If your Trust has published guidance on Copilot use, read it before you paste anything into the tool. If it has not, ask your digital lead when it will be available. If nobody knows, that itself is worth knowing.

Where I am with this.

I have not switched tools. Claude and ChatGPT still work better for how I write, how I think and the specific jobs I do most weeks. Copilot may not overtake them for personal use.

But that is not the whole factor anymore. The tool your Trust is rolling out. The tool colleagues are being trained on. The tool that will be integrated into shared workflows. The tool that has IG approval attached to it. That tool is Copilot. There is value in learning it even if you personally prefer something else because your colleagues will be using it and your professional work will increasingly touch it.

The next few weeks will tell me whether Copilot earns a place in my workflow or whether it stays a tool I use because the system asks me to. I will report back honestly either way.

What you can do this week.

One action. If Copilot is arriving in your Trust, or has already arrived, do not wait for training. Sign in. Open it. Pick one small task. Summarise a meeting note. Draft an email. Ask it to reformat a document you have already written. See what happens.

The first test tells you more than any policy document will.

That is all for Issue 14. Next week, back to work. If someone you know would find this useful, pass it on.

Clinically Intelligent drops every Wednesday. If you are not yet subscribed you can join free at clinicallyintelligent.com.

The AI tools discussed in Clinically Intelligent are consumer products. They have not been independently assessed by the author against DCB0129 or DCB0160 clinical risk management standards, and they may not be approved for clinical use by your employer. Before using any tool described in this newsletter in connection with your clinical practice, you must satisfy yourself that its use is permitted under your Trust information governance policy, your DSP Toolkit obligations, your professional registration requirements, and any applicable contractual terms with your employer. The author accepts no liability for use of any tool or workflow described in this publication. Patient identifiable information must not be entered into any consumer AI tool under any circumstances, irrespective of any guidance contained in this newsletter.