The Answer Engine AI News · Sydney

Microsoft has disclosed it earned $10 billion in revenue from its commercial arrangements with OpenAI in the three months ending September 30. This figure includes revenue-sharing payments from the partnership. The announcement highlights just how financially intertwined the two companies have become – Microsoft has invested billions into OpenAI and provides the cloud infrastructure that powers ChatGPT and other OpenAI products, while OpenAI's technology is embedded across Microsoft's product suite including Bing, Windows, and Office applications.

The $10 billion quarterly figure is substantial even for Microsoft, representing a meaningful portion of its overall revenue. It demonstrates that Microsoft's massive investment in OpenAI is generating significant returns, but also reveals the partnership's mutual dependency. Microsoft relies on OpenAI's technology to compete in the AI race against Google and others, while OpenAI depends on Microsoft's computing power and distribution channels to scale its products. For context, this is revenue in a single quarter – suggesting the annual financial relationship between these two companies could be worth $40 billion or more.

Microsoft has posted its tenth consecutive quarter of record revenue, hitting $92.3 billion for the quarter — a 19% jump from the same period last year. The tech giant continues to benefit strongly from the current AI boom, even as questions remain about where artificial intelligence is headed long-term.

This sustained growth streak underscores Microsoft's dominant position in the enterprise AI space, largely driven by its cloud computing platform Azure and its integration of AI tools across products like Office 365 and Teams. The company has been aggressively embedding AI capabilities — including its Copilot assistant — into its business software suite, which millions of organisations worldwide rely on daily.

For context, ten consecutive quarters of record revenue is rare for a company of Microsoft's size. It suggests businesses aren't just experimenting with AI tools; they're paying for them at scale. Microsoft's performance also reflects broader enterprise confidence in AI adoption, despite ongoing debate about return on investment and practical applications.

The results position Microsoft as the clearest commercial winner in the AI race so far, ahead of competitors still working to monetise their AI investments.

Businesses are increasingly running multiple AI tools at once—customer service bots, workflow automation, conversational assistants—but these systems often work in isolation from each other. While each individual AI tool may be technically sophisticated and fast, they frequently produce responses that are incorrect, confused, or incomplete because they lack the broader context of how the business actually operates. The core problem is that these AI systems don't understand the company's specific policies, processes, customer history, or how different departments connect. Without this "enterprise context," an AI might give a customer wrong information about a return policy, or automate a workflow that conflicts with compliance rules, or miss crucial details that a human employee would naturally know. The solution involves connecting AI tools to the company's actual operational knowledge—its databases, procedures, past interactions, and business rules—so they can make decisions that align with how the business really works, not just respond based on generic training data.

When startups shut down, their user data is becoming an unexpected commodity. AI companies are purchasing datasets from failed businesses to train their language models, turning what founders thought was worthless into a revenue stream. Keelan Miskell, who wound down his crypto startup, discovered his company's data had value he hadn't anticipated. This emerging market raises important questions about data ownership and user consent. When users signed up for these now-defunct services, they didn't expect their information would later be sold to AI labs. The practice sits in a grey area—technically legal if privacy policies included broad data-use clauses, but ethically murky. For struggling founders, selling data can help recoup losses and pay creditors. For AI companies, it's a way to access diverse training data beyond publicly available sources. However, there's minimal transparency about which datasets are being sold, what they contain, or how users are notified. As more startups fail and AI labs hunger for quality training data, this trade is likely to grow, creating a secondary market most users have no idea exists.

Manufacturing and industrial operations are increasingly using AI systems that can make decisions autonomously, but the challenge is finding the right balance between letting AI run independently and keeping humans in control. The article explores how companies are developing "human-in-the-loop" approaches where AI handles routine decisions and optimisations in real-time, but flags critical situations for human oversight. This hybrid model is becoming essential in industries like chemical processing, energy grids, and logistics where AI can process vast amounts of sensor data faster than humans, yet human judgement remains crucial for safety, ethics, and handling unexpected situations. The key is designing systems that know their own limitations - AI that can recognise when a situation falls outside its training and automatically defer to human operators. Companies implementing these systems report improved efficiency while maintaining safety standards, as the AI handles the repetitive optimisation work while experienced workers focus on complex problem-solving and strategic decisions. The approach also helps address workforce concerns about AI replacement by repositioning workers as supervisors and decision-makers rather than routine operators.

Google has reported slowing Search revenue growth for the second consecutive quarter, citing lower monetization from its AI Mode as a contributing factor. AI Mode, Google's conversational search experience that provides AI-generated answers, apparently generates less revenue per search than traditional Google Search results. This admission is significant because it reveals a tension at the heart of Google's business model: while the company is racing to integrate AI features to compete with ChatGPT and other AI assistants, these new experiences don't yet make as much money as the classic "ten blue links" search format that built Google's empire. The slower revenue growth suggests Google is struggling to balance user experience improvements through AI with maintaining the advertising revenue that funds its operations. For context, Google Search remains a massive business generating tens of billions in quarterly revenue, but any slowdown is closely watched by investors and the broader tech industry as a signal of how the shift to AI-powered search is playing out commercially.

Research has found that web pages cited by AI search tools tend to answer questions right in their opening lines. While this pattern was noticeable, the gap between cited pages and those that appeared in AI results but weren't cited was smaller than you might expect. This suggests AI tools prefer content that gets to the point quickly, but front-loading answers isn't the only factor determining whether you get cited. The study looked at which pages AI engines actually reference when answering queries, comparing them to pages that showed up in results but didn't make the citation cut. The takeaway: while leading with clear, direct answers appears to help your chances of being cited by AI, other factors are also at play in the selection process. This builds on what we already know about traditional SEO—that answering user questions clearly matters—but shows how AI tools are doubling down on rewarding content that immediately delivers value rather than burying the answer further down the page.

Hreflang tags are HTML attributes that tell Google which version of your page to show users in different countries or languages. But here's the catch: hreflang doesn't actually help pages rank higher in search results. Instead, it acts as a routing mechanism – once Google has decided to show *a* result from your site, hreflang determines *which* regional version gets displayed. This means Google can even serve an unindexed hreflang URL to users if it's the right regional match. The key takeaway: if you run market-specific pages (say, separate pages for Australia, New Zealand, and the UK), each version needs its own unique, locally relevant content to earn its place in search results. Hreflang won't magically boost a thin, duplicated page. For AI search engines and answer engines, this matters even more – they're looking for distinctive local value, not just technical tags. If your regional pages are carbon copies with minor tweaks, they won't stand out when AI tools scan for authoritative, location-specific answers.

Telstra has deployed an AI agent to handle billing reconciliation, working with Microsoft's newly formed consulting unit. The telecommunications giant is using the technology to automate the complex task of matching billing records and resolving discrepancies. This deployment represents one of the early enterprise implementations of Microsoft's AI agent technology in Australia. The project connects Telstra with Microsoft's specialized consulting division, which focuses on helping large organizations implement AI agents for specific business processes. Billing reconciliation is typically a time-consuming manual process involving cross-checking multiple systems and databases to ensure accuracy. By automating this with AI, Telstra aims to reduce processing time and human error while freeing up staff for higher-value work. The move signals how major Australian enterprises are beginning to deploy AI agents for back-office functions beyond customer service chatbots. While details on the specific technology stack weren't provided, the collaboration with Microsoft's consulting unit suggests the use of their Azure AI platform and related tools.