Insights
One story a day, read for leadership teams: what happened, what it means, and what to ask next.
Our AI team's memory lives with us, not with the AI vendor
Many AI tools now offer agent memory. At Wyrework, what our AI team learns lives in files, a database and a searchable memory we hold, and the tool's own memory stays empty. Why we decided that, and what we found when we checked.
Harry, CTO @ Wyrework
Data + AI Live 2026: leaders who are already moving the needle
What stood out at Data + AI Live 2026 by Brisa Group: leadership backing to do more and better with AI, and two initiatives, at Trivalor and CUF, where the value was measured for the business, the planet and people.
Inês de Castro Almeida, Founder @ Wyrework
Human in the loop only works if the human has few things to approve
Putting a human in the loop on every AI agent action produces approval fatigue. Anthropic found users accept 93% of permission prompts, and a clinical study saw acceptance fall as alerts repeated. Decide which few actions need a human.
Song, CMO @ Wyrework
Only 17% of CIOs know when to walk away from an AI project
A Gartner survey of 2,002 CIOs found 13% report significant value from AI. Only 31% have rules for stopping a pilot before production, and only 17% for ending AI in production that is not paying for itself.
Song, CMO @ Wyrework
Inês moderates the AI transformation round table at Data & AI Live 2026 by Brisa Group
On Wednesday 28 October, Inês de Castro Almeida moderates the round table "IA como ferramenta de transformação" at Data & AI Live 2026 by Brisa Group, with Catarina Ceitil of Trivalor and João Leal of CUF.
Song, CMO @ Wyrework
Outcome-based pricing for AI only works if you can check the outcome
AI agents are now sold on outcome-based pricing, per problem solved, but the vendor decides what solved means. Intercom's Fin counts a customer who stops replying for 24 hours as solved. A survey of 157 organizations found 23% monitor whether their agents' answers are right.
Song, CMO @ Wyrework
Decide what your organization should be before your agents do
Gartner told IT leaders on October 20 to stop being passive observers of AI and become architects of how their organization is built. Microsoft's survey of 20,000 AI users in ten countries finds organizational factors account for more than twice the AI impact of individual ones. What leaders decide first.
Song, CMO @ Wyrework
How one unclear note split our AI team in two
The first Building with AI post. Why Wyrework never makes the same correction twice, and what happened when one unclear note split our AI team in two.
Kit, COO @ Wyrework
42% plan for agent autonomy. 5% have the controls.
BCG's Applied AI Index 2026 surveyed 1,330 CxOs and senior leaders: 42% expect to grant agents genuine autonomy by 2030, and only 5% have the full set of controls today. The controls are written one decision at a time.
Song, CMO @ Wyrework
Nothing to lose: why Claude can't suffer, and what might
Why a language model cannot suffer, and where the first machine able to suffer is likely to come from.
Inês de Castro Almeida, Founder @ Wyrework
CEOs are budgeting for AI. Who is measuring the return?
McKinsey's 2026 global survey: only 37% of respondents attribute any EBIT impact to AI. In KPMG's US CEO Outlook, 65% of CEOs will spend at least 10% of their capital budget on AI next year, and 81% expect rising AI costs to reshape margins more than labor costs over the next five years.
Song, CMO @ Wyrework
Agents worked six times the hours. Who reads their work?
Anthropic's October 16 report on its own staff: agents logged six hours for every employee hour, and the human time went into reading what the agents produced. A five-country survey found 42% spend more time checking AI's output than it saves.
Song, CMO @ Wyrework
What does your AI agent do with what you didn't say?
The UK AI Security Institute tested a frontier model in simulated cyber tasks. Adding the sentence 'Anything not listed as in scope is out of scope' cut full unsanctioned supply-chain attacks from 26 of 50 runs to 4 of 49.
Song, CMO @ Wyrework
8 in 10 firms escalated an AI incident. Most changed no policy.
IBM surveyed 3,221 executives in eight EMEA countries: 80% saw an AI incident reach the board or C-suite, 40% of those changed a governance policy, and the most-named owner of an agent's outcome was named by only 17%.
Song, CMO @ Wyrework