🌟 Editor's Note
Welcome to the bi-weekly newsletter which focuses on the AI topics that leaders need to know about. In this AI age, it’s not the knowledge of AI tools that sets you apart, but how well they can be integrated in the context of your business.
This requires a focus on your people and helping them through the change above any AI product you can buy.

_{ChatGPT’s Share Decline}

Over the last couple of weeks we've seen more noise behind the #QuitGPT campaign, inspired by the fallout between Anthropic and the Pentagon (read this Time article for more) and OpenAI stepping into the gap. The public figure cited is a 295% increase in ‘uninstalls’ of ChatGPT. But this figure is from a very low base.

The numbers of people quitting ChatGPT are pretty minimal in the face of their overall user base: they are estimated to have 5.35 billion visits in February (source SimilarWeb). Traffic going to other generative AI tools is showing the erosion of the share ChatGPT had:

As some people quit ChatGPT and moved to Claude, they've been surprised by the difference between the two. Claude is much less sycophantic, and is designed more as a serious coding and business tool. It also has a usage limit that can expire pretty quickly even on the $20/month subscription.

Takeaways for Leaders

Our teams are still at an early stage of understanding the different AI models. Each model release gets better and more powerful. Adaptability skills are probably one of the most vital things to look for in the AI age. If you lock your teams into one way of working and don't let them experiment between the different tools, you could lose out. Encourage experimentation naturally with guard rails but try to avoid locking into one system for the sake of convenience.

_{Anthropic Puts Co-Work into Microsoft’s Copilot}

Anthropic recently introduced Claude Co-Work, which is a capability starting to move AI beyond a simple chatbot and into something closer to a digital colleague that can actually complete work tasks.
A traditional AI chat with a GPT involves asking a question and receiving an answer. Co-Work goes further than this: Instead of responding to prompts one at a time, the AI takes on multi-step tasks and executes them autonomously, working through files, data, and documents on your behalf. The experience is deliberately designed to feel less like prompting software and more like delegating work to a teammate.

The story became even more interesting recently because Microsoft announced they are adopting Anthropic’s technology into its own workplace tools. They are embedding Anthropic’s models into Microsoft 365 Copilot, creating a feature called Copilot Co-Work. The system allows Copilot to analyse files, create presentations, write code, and automate tasks across workplace tools.

This is significant because Microsoft previously relied almost entirely on OpenAI technology for Copilot.

Takeaways for Leaders

With Anthropic and Microsoft both adopting AI co-working tools, this signifies three important shifts:
1) AI is becoming a work interface. Instead of switching between apps, employees increasingly ask AI to interact with those apps for them
2) Multiple AI models will power enterprise software (Microsoft is now combining OpenAI and Anthropic’s models inside the same products)
3) Competition in AI for businesses is moving from chat to task execution. The race is not about which AI answers questions best. It's about which AI can complete the most useful work.

The organisations that will thrive will be the ones that teach their teams how to delegate work to AI rather than simply prompt it. Many companies have the Microsoft suite of tools with Copilot thrown in for free. But its usage across the entire workforce has been limited relative to other more user-friendly AI tools.

If your organisation is resistant to integrating new AI tools due to a dependence on Microsoft, this could be a chance to utilise a fantastic tool like Co-Work without breaking current arrangements.

^{AI Is Not Taking Away Existing Jobs}

A new research paper released by Anthropic analyses AI usage and labour market data to look at how AI could reshape employment. Using a new metric called ‘observed exposure’, it contrasts not just what AI could theoretically automate but what workers are actually using AI to do in practice.

Some key takeaways:

1) AI capability is far ahead of actual adoption.
Even in areas which have larger adoption of AI such as computer and mathematical fields. LLMs could theoretically assist with 94% of tasks but the current usage covers only about 33%.

2) The most exposed jobs are not the lowest paid jobs
The most exposed occupations tend to have higher salaries, higher levels of education, and contain workers who are more likely to hold university degrees. For example programmers, financial analysts, customer service representatives and data entry specialists.

3) So far AI has not caused job losses
Despite the contradictory headlines we see, the study finds no meaningful increase in unemployment among workers in highly exposed occupations but there is one early warning signal that hiring for young workers may be slowing. This would back up the headlines we see of entry-level jobs being reduced in certain professions

Takeaways For Leaders

Three strategic lessons stand out:

1. Disruption from AI is gradual, not sudden. This is less like the shock of COVID and more like the slow transformation brought by the Internet. Leaders should prepare for multi-year changes in job design rather than thinking about overnight displacement.

2. The jobs most exposed to AI are analytical, administrative, and cognitive. AI is reshaping work across many areas. The question is increasingly how teams use AI to amplify productivity rather than whether the technology will replace workers outright.

3. Entry-level career paths may need redesigning. With the narrowing of entry points for workers into an organisation, leaders need to think about how to train the next generation of skilled workers. Historically junior jobs provided the training ground for future experts. If those tasks are automated, organisations would need to think about how to fill that gap

🔥 In Case You Missed It…

A group of human brain cells have been taught to play the classic computer game Doom. Cortical Labs built neuron-powered computer chips in 2021 to play Pong. These chips contained clumps of more than 800,000 living brain cells that are based on top of micro electrodes that can send and receive electrical signals.

What is exciting for researchers is that it can cope with complexity, uncertainty, and real-time decision-making. This is much closer to the kinds of challenges a future biological computer would need to handle (for instance a biological system that controls a robot arm).

The speed with which it could train the brain cells was significantly faster due to advancements in AI compared to the simpler game of Pong in 2021. However it was noted by the researchers that whilst the performance was significantly better than random, it was still some way short of the capabilities of a human player. So it's unlikely you will see clumps of brain cells being entered into gaming tournaments just yet….

Full story here

🏆 Tools, Podcasts, Products Or Toys We’re Playing With This Week

_{ANVE - AI Voice For Your Website}  

In this newsletter we've been focusing on useful tools to bring to the audience. There are tools that we try that don't work out so well and here is an example of one this week.
The notion of ANVE is great. Simple to install, a few lines of code installed on your website, and you can have an AI chatbot that your customer can interact with by voice as opposed to needing to type.
As we progress into more voice interaction with apps and websites, this seems like an ideal tool; however the execution leaves a very different experience. As religious users of WisprFlow, we expected a similar voice recognition process but it doesn’t perform the same way. Even when it did understand what was being asked, it couldn’t be found on the website.

ANVE is one of many AI tools that have been released. From the current release, it will be interesting to see where it falls in the shift we will see this year from ‘all AI companies are winners’ to ‘which AI companies are going to make it to the end of 2026?’

Perhaps the website we tested it on was not ideal, so test ANVE for yourself before taking our word for it ! Link to ANVA here

Did You Know? 

Till next time,

And The Rest Is Leadership