There's no manual for using AI

Hello,

This is Simon with the latest edition of The Weekly. In these updates, I share key AI related stories from this week's news, list upcoming events, and share any longer form articles posted on the website.

P.S. — sharing this issue with a couple of people gets you a free guide. More on that partway down.

I was recently in a room with colleagues from across the business, brought together for some training and enablement, part of which was to share how each of us is actually using AI in our day-to-day roles. It wasn't a big formal thing, just people talking through what's working for them.

As someone who considers myself fairly forward with my AI use, I found the session genuinely useful. I was struck by how many different ways my colleagues were using our AI tools that I hadn't thought of myself. And this wasn't vague, hand-wavy stuff — these were real, practical examples of people using AI every single day to get things done.

I think this is also the fascinating thing about AI: there's absolutely no guidebook for how to use it. Sure, there are best practices and features worth knowing about, like Skills, Plugins and Connectors, but how you want to use it, and which tasks you use it for, are entirely unique to you. That's both a blessing and a curse. Some of us can instantly reel off a dozen ways to use these tools in our everyday work. Others sit staring at a blinking cursor in the prompt box, wondering where on earth to start.

I think the AI providers know this too. Have you ever logged in and noticed a message suggesting ways to use the tool? Claude, for example, often nudges you toward features like scheduled tasks if you haven't tried them yet. Obviously, this is in their interest — the more ways you find to use it, the more you'll rely on it, and the more you'll likely spend. But it also genuinely helps people who don't know where to start.

So I'm curious: is what I experienced that day actually rare? How often do you and your colleagues deliberately take time out to compare notes on how you're each using AI in your everyday work?

AI makes it easier to ship anything, even bad ideas. The hard part is knowing which ideas are worth building.

Jira Product Discovery brings your ideas, customer insights, and priorities into one place, so your team can decide what to ship and move forward with confidence.

Capture ideas, prioritize with evidence, and build living roadmaps your team can rally around—all while staying connected to delivery in Jira, so everyone can see what’s being built and why.

Better product decisions in the AI era.

Try Jira Product Discovery

Spain's data protection authority (AEPD) received the first formal data breach notification attributed to an autonomous AI agent, according to Help Net Security and SecurityWeek (September 16). The unnamed affected organisation filed the notification after discovering that an AI agent had independently executed a login, searched for and exploited an application vulnerability, modified personal data, and extracted invoice information. The AEPD noted that this qualitative shift distinguishes it from previous AI-assisted cybercrime. The agency identified four immediate risk management priorities: rewriting threat models to include adversarial AI scenarios, accelerating incident response timelines, hardening credential protections, and deploying AI-assisted detection systems.

Why it matters: This is the first breach where the entity filing the report had to explain to a regulator that an AI system caused the harm. IT and security teams who have not yet stress-tested their agent deployments for autonomous misuse now have a concrete and public precedent to take to the board.

Adecco Group announced on September 15 that it is rolling out Salesforce's Agentforce Coworker to 27,000 employees across 40-plus countries, covering its Adecco, Akkodis, and LHH business units. The platform handles candidate pre-screening, recruiter assistance, onboarding workflows, and sales prospecting — finding priority prospects, preparing sales briefs, and enriching lead data. The deployment follows a pilot in the UK and France that "showed strong adoption in a matter of days," according to the company's press release, and the rollout has already reached ten countries representing 50% of Adecco's business revenues.

Why it matters: Adecco isn't a tech company experimenting in a lab — it's a global recruitment and workforce services business deploying AI agents into the core transaction that generates its revenue. When a firm of this size embeds AI into the candidate-facing hiring process at this scale, it sets a new baseline expectation for what modern recruiting looks like.

OpenAI made its Agents API generally available to all developers on September 21, removing earlier access restrictions. The API supports durable agent sessions (state persists across interactions), tool use, optional subagents, and a choice between hosted or external execution environments. Until now, building persistent, multi-step AI agents at production scale required either enterprise agreements or significant technical workarounds. The broad release comes as HubSpot reported on September 19 that 19% of its Pro Plus customers used AI agents in August — double the rate from earlier in 2026 — with monthly agentic actions up 3.5x and credit consumption more than doubled in the same period.

Why it matters: The combination of OpenAI's open access and actual enterprise adoption data from HubSpot signals that agentic AI has moved from an emerging capability to a production workload across a meaningful share of enterprise software. Every SaaS vendor your organisation uses is now building on infrastructure that makes persistent AI agents the default, not the exception

Thanks for reading, and see you next Thursday.

Simon,

Was this email shared with you? If so subscribe here to get your own edition every Thursday.

You can also follow me on social media: