AI Weekly: A Shared AI Teammate and Cheaper Agents — June 22-28, 2026

Two useful things happened this week for small teams: Claude turned up as a permanent teammate inside Slack, and Google put agent abilities that can drive your screen into its cheapest model. Both make practical automation more accessible. There was also a signal about where costs are heading, with OpenAI unveiling its own custom chip to make running AI cheaper.

The Deep Dive: A Shared AI Teammate and Cheaper Agents

Claude becomes a permanent presence in Slack

On June 23, Anthropic launched Claude Tag, replacing its old Slack app with a persistent, shared @Claude that has team memory, an ambient mode, and can be assigned tasks like a colleague. It runs on Claude Opus 4.8 and is in beta for Team and Enterprise plans. The shift is from "a bot you message" to "a teammate that is just there", one that remembers the context of your channels.

For a small team that lives in Slack, this is the low-friction version of adopting AI. There is no new tool to open. You tag Claude the way you would tag a coworker, and it already knows what the channel has been discussing.

Google puts computer-use in its cheap model

On June 24, Google added computer-use to Gemini 3.5 Flash, letting its low-cost model control a browser and apps, at roughly a third of the cost of the pricier options. Agents that can actually operate software, at a price that makes sense for routine work, are what turn "interesting demo" into "worth automating".

OpenAI also revealed Jalapeño, its first custom inference chip built with Broadcom, aimed at cutting the cost of running AI over time.

Why this matters:

• A shared AI teammate in Slack removes the friction of adopting AI for a team already there all day.
• Screen-driving agents in a cheap model make routine automation affordable, not just impressive.
• Custom chips aimed at cheaper inference should feed through to lower AI prices down the track.

What This Means for Australian SMEs

If your team runs on Slack, Claude Tag is one of the easiest ways yet to bring AI into daily work. Because it sits in the channels you already use and remembers the conversation, people adopt it without changing how they work. Start by tagging it for quick jobs like summarising a long thread or drafting a reply, then expand from there.

The cheaper computer-use model is the bigger long-term story. Automation only makes sense when the cost of the agent is well below the cost of the time it saves. As these abilities land in low-cost models, more of your routine tasks cross that line. Keep a running list of the repetitive jobs you would automate "if it were cheap enough", because the price keeps falling toward them.

The practical takeaway for this week:

• If you use Slack, trial Claude Tag on low-stakes tasks like thread summaries and draft replies.
• Revisit routine jobs you once ruled out as too costly to automate. Cheaper agents may now make them worth it.
• Keep human review on anything an agent does on your screen until it has earned your trust.

Helping SMEs bring AI into the tools their team already uses is exactly what we do at IntelliAgent. If you want help finding the easiest first automation for your team, get in touch.