Welcome back to LLM Decode 👋
AI is moving deeper into the workplace while the business behind it faces tougher questions. Google is pushing Gemini toward becoming an AI coworker that can execute multi-step work, even as investors scrutinize whether the enormous cost of building AI can translate into sustainable revenue.
Here’s what matters today.
Google Wants Gemini to Become Your AI Coworker
Google Cloud has introduced the Gemini Agent, a universal workplace agent designed to handle everything from research and content creation to coding and multi-step knowledge work. Instead of asking AI to complete isolated tasks, users can give it a goal and let it plan the work.
The agent can use skills and tools, connect to business systems and return completed work inside the environments employees already use. Google says it can also maintain business context and coordinate work across applications, making it closer to a digital coworker than a traditional chatbot.
Another notable feature is model choice. Gemini Agent can select the model that best fits a particular task, including Google's Gemini models and, initially, Anthropic's Claude models. Google is also adding security, administration and cost controls for enterprise deployments.
Why it matters
AI is moving from answering to executing: The value shifts from generating responses to completing outcomes.
One agent could replace multiple AI tools: Businesses can increasingly give AI access to several systems instead of managing separate assistants.
Model choice may become invisible: Users may care less about which model is running and more about whether the task gets done.
Enterprise governance becomes essential: More access and autonomy means permissions, security and spending controls become part of the product.
The AI Boom Is Getting a Financial Reality Check
OpenAI's annualized revenue reportedly reached nearly $50 billion in September 2026, according to people familiar with the company's figures. That's substantially below the roughly $70 billion annualized figure previously indicated, although the comparison is complicated by differences in how OpenAI and Anthropic count revenue generated through cloud partners.
The timing matters because AI companies are spending enormous amounts to expand computing capacity. OpenAI, Anthropic and other frontier labs are simultaneously pursuing rapid revenue growth while committing to massive infrastructure and compute expenses. Investors are increasingly asking whether that spending can eventually produce sustainable profits.
The pressure is also spreading beyond AI labs. Companies including Oracle, Broadcom and SpaceX are pursuing very large financing arrangements connected to AI infrastructure and chip purchases. Morgan Stanley estimates that the broader AI infrastructure buildout could require around $1.5 trillion in external financing by 2028.
Why it matters
Revenue growth isn't enough anymore: Investors are starting to examine margins, infrastructure costs and cash requirements.
AI infrastructure is becoming a financing story: Building compute at today's scale requires enormous amounts of capital.
Lower AI prices could increase usage: Cheaper inference may create more demand, but companies still need that usage to translate into profitable revenue.
Enterprise buyers should watch the economics: Falling model prices could make AI adoption cheaper while increasing the number of tasks businesses automate.
4 AI Workflows to Try
Give AI an Outcome, Not Just a Question: Instead of asking for research, ask AI to research a topic, organize the findings, create a brief and identify what still needs human review.
Build a Multi-App Workflow: Choose one recurring process that requires several tools. Map which information AI needs from each system and identify where human approval should remain.
Measure AI by Cost per Outcome: Don't evaluate a model only by its token price. Track the total cost of completing a real task, including retries, tool calls, human review and processing time.
Create an AI ROI Dashboard: For important AI workflows, track time saved, cost per task, output quality and human intervention. This gives you a clearer picture of whether AI is actually creating business value.
Ready to level up your AI skills?
Explore Our Courses →That’s it for today.
The AI space doesn’t slow down - and neither should your thinking.
See you in the next drop.