🚀 PR Futurist #103 🏴‍☠️ Why people smuggle AI into work 💻 Copilot proves its worth 📗 New AI ethics charter for IC  17 February 2025

Dumb and Dumber - law firm restricts AI because it finds 'significant' staff use

Okay, maybe the law firm isn't really called Dumb and Dumber. But if I was a client of international commercial law firm Hill Dickinson I'd be more than a little concerned. The chief technology officer of Hill Dickinson has just warned staff about the use of AI tools. So far, so good.

What's alarming is it's a commercial law firm and it has taken until now to notice that hundreds of its staff are using unauthorised AI tools thousands of times a day! This apparently includes DeepSeek, the new Chinese AI that stores data on servers in China and has T&Cs that clearly state you don't have have any privacy protection.

"The firm said much of the usage was not in line with its AI policy, and going forward the firm would only allow staff to access the tools via a request process."

Seriously? What sort of half-baked AI policy did it have that staff already couldn't access AI tools via a request process?

Our research shows that just 40% of PR professionals have an AI policy in their workplace. This highlights that even those that have might not have an effective one.

Last year Sir Geoffrey Vos, Master of the Rolls and the head of civil justice in England and Wales said it was unethical for lawyers NOT to use AI. Using Sir Geoffrey Vos's rationale, the same applies to PR and communications professionals.

bbc.co.uk  

How a Copilot fail trial actually proves how effective Copilot is

Where to start with this article about a Microsoft 365 Copilot trial for the Australian government Treasury department? The Register, in the typical snarky style we love it for, implies the trial was a failure as "government staff rated Microsoft's AI less useful than expected."

Now let's break it down and see what it really means.

It was only a 14-week trial, so not long enough for Copilot to become a daily habit and be used extensively.

The trial concluded Copilot was useful, just "less useful than they hoped it would be, as it was applicable to fewer workloads than they hoped it would be." Well, as the trial was with volunteers, that's not surprising. It is likely to have been early adopters who've succumbed to the AI hype or have been using personal AI and expect the same at work.

The article makes no reference to whether Copilot Studio was used to tackle specific "workloads" where people had hoped to use it. Creating custom AI agents for specific 'workloads' can make a significant difference to how useful AI is.

The most astounding sentence is "Treasury thinks it probably set unrealistically high expectations before the trial, and noted that participants often suggested extra training would be valuable."

No **** Sherlock! Give people an advanced tool without a framework for how it can be applied, and then be surprised they can't use it without adequate training.

After all this, "the report finds that if Copilot saves 13 minutes a week for mid-level workers, it will pay for itself." Most studies show two to four hours saved per week. But if Copilot pays for itself in just 13 minutes, and the real time saving is even higher, the business case becomes even more powerful.

It sounds like a trial that was set up to fail. Especially when you have statements like an the Treasury says an unanticipated benefit that Copilot displayed was “to contribute to accessibility and inclusion for neurodivergent and part-time staff, or those experiencing medical conditions that require time off work."

This is a known benefit of Copilot and AI assistants. How can it be a proper trial if a known benefit is unknown to the people running the trial?

It goes on to say: "Treasury’s learnings from the pilot include more careful selection of staff who use Copilot, the need for more consideration of necessary training on how to use AI."

No **** Sherlock.

The big takeaway from the Australian government Treasury report is that despite it being a badly run trial, it still found Copilot provided some benefits and paid for itself.

Just imagine what a well-run trial would have resulted in.

Wait, you don't have to imagine, as there are hundreds of case studies of exactly that. Commercial companies and public sector bodies that trialled it (properly) with 300 people and are now up to 2,000 because the results were so positive.

Hat-tip to Alan Morrison for spotting this and asking me what I thought.

Stuart Bruce

theregister.com  

How the BBC is using generative AI

The BBC's use of AI is guided by three principles:

• always act in the best interests of the public
• always prioritise talent and creativity
• always be open and transparent with audiences when we use AI to support content-making

Like many principles they are very motherhood and apple pie but what do they mean in practice.

Some of the uses the BBC has successfully trialed include:

• Adding subtitles to some programmes
• Generating short-form animated sequences (instead of static images) to promote programmes on BBC Sounds
• Translation to publish more quickly across the world

The most interesting is for BBC Sport. BBC local radio does hundreds of live commentaries of English Football League matches. It is using AI tools to quickly create transcripts of the commentaries and highlight key match moments, like goals or red cards. After being checked by journalists, they are published as live text commentaries on the BBC Sport app.

It's also continuing to develop its own Large Language Models – trained on the BBC’s own content and data, so outputs reflect its values and tone.

AI is working at the BBC because it's investing heavily in training in AI tools like Microsoft Copilot, Adobe Firefly and GitHub Copilot.

Stuart Bruce

bbc.co.uk  

Channel 1 is a new news channel where stories are scripted, edited and presented by AI

While the BBC is taking a cautious approach to AI Channel 1 is being bolder. It is a new rolling news channel with a difference - its stories are scripted, edited, and presented by AI. The Guardian visited its creators in Los Angeles to learn more... and audition for a role.

Stuart Bruce

theguardian.com  

A TV ad with no Actors, locations or sets

Take a look at this video that was created using AI using Google DeepMind's Veo 2.

Wow or what?

Stuart Bruce

forbes.com