david decemme

I'm increasingly convinced Claude Code is an early glimpse of what general purpose AI with agency feels like. I'm experimenting with making it more usable for non-programmers (with an initial focus on policy/SNAP use cases) — here's an example: - I created a custom output style (which overrides the default, programming-oriented system prompt of Claude Code) called "non-technical user" - That custom style enforces that Claude Code will describe to the user — in non-technical terms — what it's doing, why, and any risks the person should be aware of - This screenshot shows a bit of this custom output style in action: working on a task that a SNAP policy expert might want to do: "Can we make a tool to chat with the policy manual for CT located here? https://lnkd.in/gpuQu2-s" - Because we're using a custom output style, it's explaining in lay terms what's happening and any risks before the user says "Yes" to running code This is very much a proof of concept! But I *really* want more people to be able to experience the more "agentic" capabilities of AI. It's really an entirely different experience relative to the standalone chat apps. Anthropic #ai #snap

280

21 commentaires

#LookerStudio CrUX Dashboard deprecation The Google Chrome team is deprecating the CrUX dashboard for a shutdown at the end of November. On this occasion, Barry Pollard wrote a very detailed article on the history of this project to provide Chrome technical performance data by site, its evolutions, as well as the possible alternatives for a migration that you must plan within the next 2 months if you are currently using the CrUX dashboard in Looker Studio. You have 2 options regarding this third-party data: 1️⃣ The "CrUX Vis" project 2️⃣ If you want to continue using the Looker Studio dashboard template: The underlying data, which was made easier to access in Looker Studio through the Chrome UX Report connector, has also been available via BigQuery for several years through a materialized view (chrome-ux-report.materialized.device_summary). You can query this view directly or ingest the desired data for the desired sites to store them in your warehouse against a query processing about 13 GB. The article includes a step-by-step tutorial to make this transition with a custom query. Up to you to choose the solution that best suits your needs...or start collecting first-party data in the meantime 😉 You will find the link to the article with the comments

46

5 commentaires

The emerging category of enterprise AI: Decisioning Agents In recent posts I described the long-term dream behind “agentic analytics.” It is the same dream as “intelligent automation” and “generalizable recommenders.” Agents that: • Sense continuously • Decide with sophistication • Surface relevant insights in the flow of work • Execute actions autonomously That's not analytics. That is Alan Turing’s 1950 dream of thinking machines. A better name for this dream in 2025 name is 𝐃𝐞𝐜𝐢𝐬𝐢𝐨𝐧𝐢𝐧𝐠 𝐀𝐠𝐞𝐧𝐭𝐬. Here is the emerging ontology of enterprise AI: 𝐈𝐧𝐭𝐞𝐫𝐚𝐜𝐭𝐢𝐨𝐧 𝐀𝐠𝐞𝐧𝐭𝐬 handle human-facing tasks via natural language. Example: the help.salesforce.com support agent. This is an established $12B+ market that predates LLMs. 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐠𝐞𝐧𝐭𝐬 create content for business tasks. Example: coding copilots or marketing copy generators. This is a fast-growing $15B+ market. 𝐃𝐞𝐜𝐢𝐬𝐢𝐨𝐧𝐢𝐧𝐠 𝐀𝐠𝐞𝐧𝐭𝐬 derive insights from data and execute business processes. Example: see churn risk early and trigger a retention play. This category is nascent, but as it matures, it can subsume analytics, BI, workflow automation, and recommendation systems into a $100B+ market. My predictions: • Decision Agents becomes the next hot enterprise AI category • Early wins will land in lead-to-cash where ROI is easiest to measure • Trust will come first, autonomy will come later. • Human will be in the loop for a long time. • Data teams will play that role. Expect the rise of the Decision Engineer. What do you think?

54

18 commentaires

Amazon Ads launched Creative Studio last week – a new agentic AI tool that acts as a creative partner for advertisers, powered by Amazon Nova. What stands out about this application: advertisers can have natural conversations with the AI to create professional ads from concept to completion. The system handles audience research, brainstorming, storyboarding, and produces video and display ads – all in just a few hours. It's a great example of how agentic AI can streamline ad creation while preserving the creativity and audience insights advertisers need. Looking forward to seeing what customers build with it and continuing our work with the Ads team! You can learn more here: https://lnkd.in/euYNRiMs

455

5 commentaires