Databricks Genie Scores 91% on Business Tasks

Databricks is positioning Genie as a business-focused AI assistant that draws on company context, not just database tables. In one Turing-built test, Genie scored 91% accuracy on business tasks involving tables, dashboards and notebooks; Databricks says teams including Zillow, Unilever, adidas, Atlassian and AstraZeneca use it in production. A separate test using the fictional energy supplier Calderwick Energy found Genie with Genie Ontology returned the correct FY24/25 planning rate—0.52 pence per kWh—in all five runs, using less than half the tokens of other agents. Databricks cautions that this advantage may shrink in workspaces with little existing context; the reported tests do not provide task counts, competitor details or per-run costs.

Databricks is also extending Genie to ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot through plugins, with an MCP server option for agents without a native plugin. Agents can ask Genie a complete question or retrieve a specific business fact, and supported MCP Apps clients can display charts and Ontology citations. Example uses include explaining sales performance, tracking product adoption and preparing reports. Microsoft Copilot Cowork requires approval before sending anything. Databricks warns that agents can still give confident but incorrect answers if shared business definitions become outdated. Unity Gateway adds controls for tool access, action tracking and spending under Unity Catalog permissions.

Constellation is building a shared Enterprise Data Platform on Databricks while keeping its existing systems. It selected the platform within weeks, citing its ability to work across those systems; Unity Catalog provides lineage and detailed access controls, while Data Vault is reserved for critical reusable information. Three pilots are underway: a Genie assistant for approved nuclear engineering information, tools combining outage schedules with asset health, and sales tools to prioritize opportunities and next steps. Constellation is also exploring Genie Ontology and Unity Catalog Pages for centrally managed business definitions.

AI investment is drawing scrutiny alongside growth. A September 2026 survey of 500 senior executives found nearly three in five organizations spent significantly more on AI than expected or budgeted in the previous year; among those with overruns, spending averaged 24% above expectations. Organizations could systematically track 61% of AI spending but connect only 45% to verified business value, and 60% lacked consistent enterprise-wide cost visibility. CFOs participate in approving major AI investments at 52% of organizations, up from 42% a year earlier. The report recommends tracking costs and outcomes, setting guardrails and selecting models and workflows carefully.

Companies are also reporting varied financial and product developments. Reply S.p.A. posted 9.2% year-over-year revenue growth to €1,987.9 million for the first nine months of 2026, with EBITDA of €354.9 million and a positive net financial position of €397.0 million; it lists AI, agentic systems and employee skills as priorities. ALMU reported fiscal 2027 first-quarter revenue of $0.6 million, unchanged from the prior quarter, and net losses of $3.8 million under GAAP and $2.4 million on an adjusted basis. Its development work targets high-speed photodetectors and quantum dot lasers for AI datacom.

AI-related demand is also supporting markets and infrastructure. Nvidia shares climbed 2.8% to $646.41 as the broader market was flat, giving the company the largest gain among major AI stocks and pushing its market value above $500 billion. Alphabet, meanwhile, is raising about US$3.8 billion in its first Hong Kong dollar and Singapore dollar bond sales, exceeding a US$3 billion target after receiving more than US$7 billion in orders. The seven-part sale spans two to 30 years; proceeds are for general corporate purposes and may include debt repayment. Settlement is expected Nov. 19, with expected ratings of Aa2 from Moody’s and AA+ from S&P. The articles also name Google and ChatGPT among the wider AI and technology developments.

Hardware growth is increasing the stakes around semiconductor patents. The Semiconductor Industry Association says global chip sales reached a record $795.6 billion in 2025, with AI chips accounting for roughly half of revenue but less than 0.2% of units. Memory bandwidth is a key engineering challenge, driving adoption of high-bandwidth memory (HBM): its market was valued at $2.93 billion in 2024 and is projected to reach $16.72 billion by 2033. DRAM patent disputes in U.S. district courts rose 119% in 2025. The article notes the International Trade Commission’s 15-to-18-month timeline and exclusion orders, and says the industry is moving toward chiplet designs as single-die processors face limits.

Governance and deployment remain central themes. Vanta launched AI Governance, Risk Quantification and Autonomous Vendor Assessments, and upgraded its Vanta Agent. The governance tools are designed to discover and test AI agents, control behavior and document oversight; the agent can handle jobs and bring in staff when human judgment is needed. It is available through Claude, ChatGPT, Cursor and Slack. Vanta also named former Reddit and Gusto executive Fredrick “Flee” Lee as CISO, effective Nov. 16; the company has been valued at $4.15 billion since July 2025. At Homeland Security Investigations, official Peter Hatch says AI outputs must be auditable and supported by source documents. HSI has nearly a dozen AI uses and about five more in development or acquisition, including tools for lead generation, translated text, digital records and maritime data. Investigators must document use, review results and retain human decision-making.

Other coverage examines AI’s possible effects on productivity and company growth. A stock screen identifies 49 companies linked to automation, software and workplace efficiency, and highlights BILL Holdings, whose AI-enabled platform supports bill payment, invoicing and expense management for small and midsize businesses. BILL has about US$1.69 billion in software and programming revenue, including roughly US$1.66 billion from U.S. customers, and a US$4.6 billion market value. Hatz AI raised $24 million from Nadia Partners, Vestigo Ventures and The Private Shares Fund to expand its workforce, accelerate research and engineering, and strengthen go-to-market efforts. It plans to double staff, support managed service providers building customized AI solutions, and develop systems that identify and repair product issues with human supervision. The company says it has issued more than 15,000 AI certifications worldwide and recently created the SMB1 benchmark, based on about 500 real-world work tasks.

Key Takeaways

  • Databricks says Genie achieved 91% accuracy in a Turing-built business-task test; its separate Calderwick Energy test returned the 0.52 pence-per-kWh rate in all five runs using less than half the tokens of other agents.
  • Databricks is connecting Genie and Genie Ontology to ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot; Constellation is piloting the platform for nuclear engineering information, outage and asset-health tools, and sales prioritization.
  • A survey of 500 executives found nearly three in five organizations exceeded AI budgets or expectations; average overspending among those affected was 24%, while only 45% of AI spending could be connected to verified business value.
  • Reply reported 9.2% revenue growth to €1,987.9 million for the first nine months of 2026 and named AI and agentic systems among its priorities; ALMU reported $0.6 million in fiscal Q1 2027 revenue and continued AI photonics development.
  • Nvidia rose 2.8% to $646.41, taking its market value above $500 billion as the broader market stayed flat.
  • Alphabet is raising about US$3.8 billion in its first Hong Kong dollar and Singapore dollar bond sale, following more than US$7 billion in investor orders.
  • Global semiconductor sales hit $795.6 billion in 2025; AI chips generated roughly half of sales while representing less than 0.2% of units. DRAM patent disputes in U.S. district courts rose 119% in 2025.
  • Vanta launched AI governance, risk and vendor-assessment tools and an upgraded agent available through Claude, ChatGPT, Cursor and Slack; Fredrick “Flee” Lee becomes CISO on Nov. 16.
  • HSI says AI outputs used in investigations must be auditable, backed by source documents and reviewed by people; the agency has nearly a dozen AI uses in place and about five more in development or acquisition.
  • Hatz AI raised $24 million and plans to double its workforce; BILL Holdings is highlighted for AI-enabled financial software, with US$1.69 billion in software and programming revenue and a US$4.6 billion market value.

Databricks Genie reaches 91% accuracy on business tasks

Databricks says Genie achieved 91% accuracy in a Turing-built test of business tasks that involved joining tables, dashboards and notebooks in one workspace. The company says thousands of teams use Genie in production, including Zillow, Unilever, adidas, Atlassian and AstraZeneca. Unity Gateway provides controls for tool access, action tracking and spending under Unity Catalog permissions. The reported result does not include the task count, competing products or cost per run, and Databricks has not described how the system handles jobs that need repeated approvals.

Databricks says business context helps Genie outperform larger models

Databricks says Genie with Genie Ontology reached 91% accuracy using less than half the tokens of other agents in a test, because it could draw on notebooks and other workspace content, not only tables. The test used a fictional UK energy supplier called Calderwick Energy, with hundreds of tables and a table of about 120 million rows. For a question about the FY24/25 Commercial planning rate, Genie found a notebook record of 0.52 pence per kWh and returned that value in all five runs. Other agents returned 0.31 or said the year-specific rate was not documented. Databricks says the advantage may be smaller in new workspaces with little context, and it has launched Genie plugins for ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot.

Constellation builds a shared data platform on Databricks

Constellation is building its Enterprise Data Platform on Databricks to give teams a shared understanding of their data without replacing their existing systems. The company said it chose Databricks within weeks because it could work across those systems. Unity Catalog provides data lineage and detailed access controls, while Constellation uses Data Vault only for critical, reusable information. Three pilots are running: a Genie assistant for approved nuclear engineering information, tools that bring together outage schedules and asset health, and sales tools that prioritize opportunities and next steps. Constellation is exploring Genie Ontology and Unity Catalog Pages to manage business definitions centrally.

Databricks gives AI agents shared business definitions

Databricks is connecting agents such as ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot to business information through Genie and Genie Ontology. Agents can ask Genie a full question or request a specific fact while handling the rest of a task themselves. Agents without a native plugin can use the Genie MCP server, and supported MCP Apps clients can show charts and Ontology citations. Examples include tracking product adoption, explaining a regional sales shortfall, preparing reports and reviewing costs. Microsoft Copilot Cowork must wait for approval before sending anything, and the article warns that agents can all produce confident but incorrect results if a shared business definition becomes outdated.

Reply reports 9.2% revenue growth as it prioritizes AI investment

Reply S.p.A. reported that revenue for the first nine months of 2026 rose 9.2% year over year to €1,987.9 million. EBITDA reached €354.9 million, and its net financial position remained positive at €397.0 million. The company listed AI, agentic systems and employee skills among its strategic priorities. The figures come from its interim report dated November 12, 2026.

ALMU reports steady revenue as AI photonics products advance

ALMU reported revenue of $0.6 million in the first quarter of fiscal 2027, unchanged from the previous quarter. The company recorded a GAAP net loss of $3.8 million and an adjusted net loss of $2.4 million. Its development work focuses on high-speed photodetectors and quantum dot lasers for AI datacom.

AI stock gains help lift Wall Street

Nvidia shares rose 2.8% to $646.41 as AI stocks gained, while the broader market was flat. Nvidia had the largest gain among major AI stocks, and its market value topped $500 billion. Many AI stocks have more than doubled in value over the past year. The article links the sector’s strength to rising demand for artificial intelligence and machine learning technologies.

Survey finds AI spending overruns and weak cost tracking

A September 2026 survey of 500 senior executives found that nearly three in five organizations spent significantly more on AI than expected or budgeted in the past year. Among those reporting an overrun, spending was 24% above expectations on average, and 60% of organizations lacked consistent, enterprise-wide visibility into AI costs. Organizations could systematically track 61% of AI spending, but connect only 45% to verified business value. Among organizations with overruns, 53% responded by controlling spending and 20% reconsidered their tools or processes. CFOs now help approve major AI investments at 52% of organizations, up from 42% a year earlier; the report recommends tracking costs and outcomes, setting spending guardrails, and choosing models and workflows carefully.

AI hardware boom raises stakes for semiconductor patents

The Semiconductor Industry Association reported that global semiconductor sales reached a record $795.6 billion in 2025, with AI chips driving roughly half of revenue while making up less than 0.2% of units. The article says large AI infrastructure investments, shared engineering challenges, and a maturing patent landscape could make AI hardware a major area for patent disputes. One key challenge is memory bandwidth, which has led the industry to adopt high-bandwidth memory, or HBM, in leading AI accelerators. The HBM market was valued at $2.93 billion in 2024 and is projected to reach $16.72 billion by 2033, while DRAM patent disputes in U.S. district courts rose 119% in 2025. The article also describes the International Trade Commission’s 15-to-18-month timeline and exclusion orders, and says the industry is moving toward chiplet designs because of limits on single-die processors.

Vanta introduces AI governance tools and an upgraded agent

Vanta launched AI Governance, Risk Quantification, Autonomous Vendor Assessments, and an updated Vanta Agent for security and compliance teams. AI Governance is designed to track AI agents through discovery, vulnerability testing, behavior controls, and proof of governance. Risk Quantification is intended to show leaders the financial impact of key risks, while the Vanta Agent can take on full jobs and bring in staff when human judgment is needed. Users can access the agent in Claude, ChatGPT, Cursor, and Slack. Vanta also named Fredrick “Flee” Lee, formerly of Reddit and Gusto, as its chief information security officer, effective Nov. 16; the company has been valued at $4.15 billion since July 2025.

Alphabet raises $3.8 billion in Hong Kong and Singapore bonds

Alphabet is raising about US$3.8 billion through its first Hong Kong dollar and Singapore dollar bond sales, above its US$3 billion target after strong demand. Investors placed more than US$7 billion in orders. The seven-part sale includes HK$16 billion in Hong Kong dollar bonds and S$2.28 billion in Singapore dollar bonds, with terms ranging from two to 30 years. Alphabet plans to use the money for general corporate purposes, which may include repaying debt. The bonds are due to settle on Nov. 19 and are expected to receive ratings of Aa2 from Moody’s and AA+ from S&P.

HSI says AI tools must produce results investigators can audit

Homeland Security Investigations official Peter Hatch says the agency needs AI tools whose results investigators can check and support with source documents. HSI investigates federal crimes including smuggling, cybercrime, human trafficking and intellectual property crime, and its case materials may be used in court. The agency has nearly a dozen AI uses in place and about five more in development or acquisition, including tools for lead generation, translated text analysis, digital records review and maritime data. Investigators must document AI use, review its results and keep people in charge of decisions. Hatch said AI can help skilled analysts work faster, but it cannot turn a poor analyst into a great one.

Three software stocks that could benefit from AI automation

The article looks at three companies from the AI Productivity Leaders stock screen and considers how AI automation could affect workplace efficiency and pricing power. It says concerns about retirement and Social Security may influence household spending and saving, while companies using AI to get more work from each employee could gain an advantage. The full screen identified 49 additional companies with similar links to automation, software and workplace efficiency. One company featured is BILL Holdings, which offers an AI-enabled platform for bill payments, invoicing and expense management for small and midsize businesses. BILL has about US$1.69 billion in software and programming revenue, including roughly US$1.66 billion from US customers, and a market value of US$4.6 billion.

Hatz AI raises $24 million to grow its platform and workforce

Hatz AI has secured $24 million from Nadia Partners, Vestigo Ventures and The Private Shares Fund. Josh Schertzer, a Nadia Partners partner and former Blackstone CTO, led the investment. The company plans to double its workforce, speed up AI research and engineering, and strengthen its go-to-market strategy. It will also help managed service providers build and deploy customized AI solutions, building on more than 15,000 AI certifications issued worldwide. Hatz AI recently created the SMB1 benchmark, which uses about 500 tasks to measure how AI handles real-world work, and is working on systems that can identify and repair product issues with human supervision.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

Databricks Genie Genie Ontology Enterprise AI assistants AI agents Model Context Protocol (MCP) Unity Catalog Enterprise data platforms AI governance AI spending and ROI AI cost management AI auditing Human oversight Constellation Energy AI in financial services AI infrastructure Semiconductors High-bandwidth memory (HBM) Chip patents AI photonics AI business applications

Comments

Loading...