Valliance Logo in White
Valliance Logo in White

ISSUE NO.

ISSUE.

011

·

ISSUE NO.

11

·

The Week in AI

Our weekly read on the AI stories that matter to the people building the future enterprise.

ISSUE NO.

ISSUE.

011

·

ISSUE NO.

11

·

The Week in AI

Our weekly read on the AI stories that matter to the people building the future enterprise.

Anthropic has released Claude Fable 5.2, open to everyone, and Mythos 5.2, which has looser cyber-security guardrails and is limited to certified companies. Anthropic reports clear gains in coding, reasoning and science, plus new robotics abilities, and a lower price per token. Its own charts compare Fable at extra-high effort against older models.

Spotted on

Anthropic

Why it matters for you

The change I notice most is that coding quality now holds when you move from research into building, so a hard problem can be quickly iterated in one model. It also writes far more readably than 5.1, though it shows less of its reasoning, so I'd keep a person checking the work that matters. The lower price per token only holds at lower effort settings. Anthropic's own charts show it costing more than the previous model above high effort. Fable is also more capable than most work requires. I’d gate access to the work that needs it, set effort defaults centrally and use cheaper models for routine jobs.

Anthropic has launched Claude Meetings, which joins video calls to draft plans, answer questions and record actions, alongside Claude Whiteboards and Claude Sheets. Together with Claude Docs, they move Claude towards a full office suite. The launch also adds wider data and app integrations.

Spotted on

TechCrunch

Why it matters for you

Put a circle around meetings, documents, spreadsheets and whiteboards and you have most of what people use every day. I read this as Anthropic accepting that the model alone is no longer a defence, so the suite is there to give you more reasons to stay within their ecosystem. Smaller and younger firms without long incumbent contracts will be tempted, but a large enterprise won't leave Microsoft 365 for this - yet. When Anthropic ship their own identity and email services, it may be worth a serious look when your next renewal comes around.

Deloitte surveyed 806 board members and C-suite executives in 53 countries. Only 16% have an established strategy for how technology change affects their workforce, and almost half say boards don't consistently bring people into AI discussions. Productivity dominates board conversations, while employee trust comes up in only 6%.

Why it matters for you

Fewer than one in five organisations have a plan for what AI means for their people, and employee trust comes up in only six in a hundred board conversations. Boards are signing off AI spend without discussing what it means for the workforce. That's a governance gap, and it's one we hear about in client conversations. Closing it has to start at the top, because people copy their leaders, and nobody uses the tools with confidence if senior leaders don't. Our advice is to put the AI programme and the people plan under one owner.

OpenAI has launched a beta that lets verified businesses claim and manage a ChatGPT profile, add product feeds with pricing, and see aggregated mentions and clicks. Businesses can also set up an agent that answers for them and pay to sponsor it through ChatGPT Ads. It starts with US e-commerce.

Spotted on

Dataconomy

Why it matters for you

A business can now have an agent speaking for it inside ChatGPT, and paying to be seen there has arrived. It's US e-commerce only for now, but I'd expect it to spread. Someone has to own that agent and decide what it's allowed to say, so this is a brand and governance question as much as a marketing one. I'd settle it before the profile goes live, the same way you'd sign off a spokesperson. Earn your place through sources others cite, because self-published "best of" lists won't hold up for long.

ISSUE NO.

010

_More of our latest thinking

_More of our latest thinking

_More of our latest thinking

_More of our latest thinking

ISSUE NO.

ISSUE.

011

·

ISSUE NO.

11

·

The Week in AI

Our weekly read on the AI stories that matter to the people building the future enterprise.

ISSUE NO.

ISSUE.

011

·

ISSUE NO.

11

·

The Week in AI

Our weekly read on the AI stories that matter to the people building the future enterprise.

Anthropic has released Claude Fable 5.2, open to everyone, and Mythos 5.2, which has looser cyber-security guardrails and is limited to certified companies. Anthropic reports clear gains in coding, reasoning and science, plus new robotics abilities, and a lower price per token. Its own charts compare Fable at extra-high effort against older models.

Spotted on

Anthropic

Why it matters for you

The change I notice most is that coding quality now holds when you move from research into building, so a hard problem can be quickly iterated in one model. It also writes far more readably than 5.1, though it shows less of its reasoning, so I'd keep a person checking the work that matters. The lower price per token only holds at lower effort settings. Anthropic's own charts show it costing more than the previous model above high effort. Fable is also more capable than most work requires. I’d gate access to the work that needs it, set effort defaults centrally and use cheaper models for routine jobs.

Anthropic has launched Claude Meetings, which joins video calls to draft plans, answer questions and record actions, alongside Claude Whiteboards and Claude Sheets. Together with Claude Docs, they move Claude towards a full office suite. The launch also adds wider data and app integrations.

Spotted on

TechCrunch

Why it matters for you

Put a circle around meetings, documents, spreadsheets and whiteboards and you have most of what people use every day. I read this as Anthropic accepting that the model alone is no longer a defence, so the suite is there to give you more reasons to stay within their ecosystem. Smaller and younger firms without long incumbent contracts will be tempted, but a large enterprise won't leave Microsoft 365 for this - yet. When Anthropic ship their own identity and email services, it may be worth a serious look when your next renewal comes around.

Deloitte surveyed 806 board members and C-suite executives in 53 countries. Only 16% have an established strategy for how technology change affects their workforce, and almost half say boards don't consistently bring people into AI discussions. Productivity dominates board conversations, while employee trust comes up in only 6%.

Why it matters for you

Fewer than one in five organisations have a plan for what AI means for their people, and employee trust comes up in only six in a hundred board conversations. Boards are signing off AI spend without discussing what it means for the workforce. That's a governance gap, and it's one we hear about in client conversations. Closing it has to start at the top, because people copy their leaders, and nobody uses the tools with confidence if senior leaders don't. Our advice is to put the AI programme and the people plan under one owner.

OpenAI has launched a beta that lets verified businesses claim and manage a ChatGPT profile, add product feeds with pricing, and see aggregated mentions and clicks. Businesses can also set up an agent that answers for them and pay to sponsor it through ChatGPT Ads. It starts with US e-commerce.

Spotted on

Dataconomy

Why it matters for you

A business can now have an agent speaking for it inside ChatGPT, and paying to be seen there has arrived. It's US e-commerce only for now, but I'd expect it to spread. Someone has to own that agent and decide what it's allowed to say, so this is a brand and governance question as much as a marketing one. I'd settle it before the profile goes live, the same way you'd sign off a spokesperson. Earn your place through sources others cite, because self-published "best of" lists won't hold up for long.

ISSUE NO.

010

_More of our latest thinking

_More of our latest thinking

_More of our latest thinking

_More of our latest thinking