🌟 Vasilij’s Note
Last week I had a chance to speak to around 40 business leaders in Cardiff (picture above). It was a great event organised by Help To Grow. I have learned a lot while talking to the business leaders there. Jump to my section below to learn more.
This Week in Agents | What Changed
Google AP2 (agent payments): Open protocol for AI agents to shop/pay with two-step consent (intent + cart), optional auto-buys under limits, and signed audit trails; 60+ partners. Read more below>
Notion 3.0 (personal agents): Agents execute multi-step tasks across your workspace with memory, run for ~20 minutes, and can be scheduled—speeding planning, documentation and coordination. Read more below>
We are using Notion at AiGentic Lab. I will be doing a video on Notion3 agents in the upcoming week or two. If you are interested to see it in action, subscribe here now so you don’t miss it: https://www.youtube.com/@vasilijnevlev
Upskilling Spotlight: Skills Bootcamps
Free, flexible training (up to 16 weeks) for adults 19+ across in-demand areas – digital (data, software, marketing), technical (construction/engineering), green skills (EV, heat pumps), health & social care, HGV driving, and business & admin. Programmes run online or at local providers and come with a guaranteed job interview on completion. skillsforcareers.education.gov.uk
Why this matters for you:
Close priority skills gaps fast without long course commitments; most Bootcamps fit around work and family.
Low cost for employers: when upskilling existing staff, employers contribute 10% (SMEs) or 30% (large firms). Individuals and the self-employed are fully funded.
Direct hiring pipeline: candidates who complete a Bootcamp receive an interview, making it useful for entry-level and career-change roles.
How to take action:
Find live courses and apply via the National Careers Service.
Select relevant tracks (AI/data, digital marketing) that align with your 90-day priorities.
Line up interviews for graduates or nominate team members for funded places.
For Scotland, Wales or Northern Ireland, use the devolved links provided on the Skills for Careers page (bottom right). Link.
Maker Note | What I built this week
My team and I have been building a voice agent using n8n (orchestration) and ElevenLabs (voice AI). The work is in progress and too early to share anything, but here are key takeaways so far:
Latency is the key to interaction; machines don’t say “ahm” or give you “give me a sec, to look it up”, which makes it uncomfortable for the caller
We are experimenting with asynchronous handling of connections and webhooks to make sure the interaction is as seamless as possible; otherwise, it is an uncanny valley all the way.
Context engineering and tool configuration for the voice agent are crucial, but ensuring the correct application of each tool is challenging.
Monitoring the agent’s following statement provides a clear record of interaction for improvement.
Top Moves | Signal and Impact
Google Unveils Open Protocol for AI-Driven Transactions
Google Unveils Open Protocol for AI-Driven Transactions
What happened: Google launched the Agent Payments Protocol (AP2), a new open standard for purchases initiated by AI. Backed by over 60 banks, merchants, and tech players, AP2 sets out rules for how AI agents can securely shop, negotiate, and complete purchases on behalf of users. It introduces a two-step approval: an “intent mandate” (the user’s go-ahead to find options) and a “cart mandate” (final user approval for a specific purchase), with an option for fully automated buys under strict preset conditions. All transactions produce an auditable, cryptographically-signed trail for security. google.com
Why it matters: This is a foundational step toward an agent-to-agent economy. By standardising how AI agents authenticate payments and interact with retailers’ systems, Google is preparing for a near future where your digital assistant might routinely book travel or order supplies by talking directly to seller agents. For businesses, AP2 promises interoperability – any compliant agent (not just Google’s) can transact across platforms – and builds trust through transparency and control. Strategically, it signals that big industry players expect autonomous shopping and B2B procurement by AI to become mainstream. Companies should anticipate new commerce models (e.g. spontaneous AI-negotiated offers) and consider how their products and pricing appear to algorithmic buyers, not just human customers. google.com
Notion 3.0 Brings Personal Agents to the Workplace
What happened: Productivity platform Notion released Notion 3.0, transforming its built-in AI into fully-fledged Agents that can perform multi-step tasks across your notes, databases, and integrations. Users can now instruct their personal Notion Agent to handle projects like creating a content calendar or drafting a meeting brief – the agent will generate plans, assign tasks, update pages, and even pull information from connected tools (e.g. Slack, Google Drive) as needed. These Agents have a memory (they retain context from your workspace) and can operate for up to 20 minutes continuously, executing actions that span dozens of pages or records. Notion is also introducing Custom Agents that teams can set to run autonomously on schedules or triggers (e.g. a Support Agent that triages incoming tickets every hour).
Why it matters: This shift turns Notion from a passive workspace into an active collaborator. For teams and managers, routine planning, documentation and coordination tasks can now be offloaded to AI, potentially reclaiming hours. It demonstrates how AI agents can be embedded in everyday productivity software to augment knowledge work rather than replace workers – acting like a tireless project assistant that follows your preferences. Strategically, Notion’s move underscores a broader trend: workplace software is racing to integrate AI agents that truly act on information, not just answer questions. Businesses that leverage such tools may see faster project execution and more consistent organisational knowledge management. It’s an early glimpse of how “AI co-pilots” can scale across all levels of an organisation, from individual to enterprise-wide processes. notion.com
Operator’s Picks | Tools To Try
Wide Research (Manus) — A feature that spins up 100+ general‑purpose agents in parallel, each a full Manus instance, to tackle large research. Agents collaborate rather than follow rigid roles, unlocking high‑volume analysis for tasks like comparing the Fortune 500 or exploring GenAI tools. tasksmanus.im.
ElevenLabs Agents now support Chat Mode - text-only mode for its voice agents (enable via config or runtime override); pair with agent transfer to escalate to voice; bring your own LLM support. → One agent across chat + voice, fewer bots to maintain elevenlabs.io
Deep Dive | Here is what I have learned from the Help-To-Grow event in Cardiff, doing an AI Assessment
I have just returned from this event. It was a pleasure to speak to about 40+ business leaders in the room. As a part of my presentation, we have completed the AI Readiness Survey. You can see the overall results below.
AI Quiz Results from Help to Grow Wales National Event for Alumni.
TL;DR: Businesses are change-ready, but not execution-ready.
Small to medium-sized businesses have a strong culture and great readiness for change, but business leaders are struggling to cut through the hype and noise. They are ready to jump in and start implementing when given clear support and instructions on how to begin, free from sales BS.
Here is my interpretation of the scores:
Culture & Adaptability tops out at 71% — significant momentum to build on, indicating that businesses led by founders with minimal middle-management are well-positioned to capitalise on the change opportunity.
AI implementation is the lowest at 54%, with the lowest score, indicating that many businesses are in the middle, having started experimenting but are unclear on how to leverage experimentation into action.
Team & Skills at 62% — “good enough to start” with light enablement, by far with the highest middle maturity metric, which means with more enablement, could make a good start and progress towards higher maturity.
Data & Tech at 57% — which typically results in a scattered system and a lack of data integration — is a significant challenge. However, modern system integration and data integration systems can enable this quickly and easily.
Strategy & Ops at 59% — close, but needs crisper goals and metrics to focus minds and progress forward.
Overall: 60% — solid foundation, uneven follow-through and requires some support with expertise.
What I showed at the event that got the most traction
The attendees were particularly interested in the locally run AI, how easy it is to create automation with n8n and where to find more information.
In the breakout sections, most businesses focused on prompt engineering and building AI agents with n8n. We had the most questions about AI agents and their utility.
We did an informative session on what an AI agent is, the cost of a token, the length of context and how various models interact differently with the content.
My key takeaway:
Business leaders require clear guidelines on leveraging AI and AI agents to generate value. The focus is on creating value now and efficiency for their businesses, rather than grand speeches and ambitions or the big picture. They aim to deliver practical value that can be generated with their teams in the present moment.
Spotlight Tool | LM Studio
LM Studio is a cross-platform desktop app (Mac/Windows/Linux) for running and serving open-weight AI models locally, combining a clean chat UI with an intuitive interface. It matters that you get stronger data privacy (nothing leaves the device), predictable costs (free to run) and low-latency responses—perfect for confidential drafting, on-prem workflows and regulated teams.
Key features: a curated model catalogue (Gemma, Mistral, Qwen, DeepSeek, etc.) with one-click downloads and quantised builds for modest hardware; CPU/GPU acceleration; document ingestion for offline RAG; the ability to serve on localhost or your LAN, and drop-in use by simply changing the API base URL in tools like n8n or your custom apps.
I show how to set it up in one second-ever video on YouTube here (pardon the quality, it was the second-ever video).
Did you find it useful? Or have questions? Please drop me a note. I respond to all emails. Reply to the newsletter or write to [email protected]