Before the board asks: 5 Q4 dates for your AI agents  Security patches, holiday traffic, state AI laws and budget reviews are landing together. Here is the calendar, and the five questions every AI agent should answer first.  Most leaders manage each Q4 deadline on its own: a patch window here, a holiday freeze there, a budget review in December. This year they overlap, and AI is the common thread. It speeds up how fast vulnerabilities are found, how many systems act on their own, and how closely finance and regulators look at the results.  In this article you will get the five dates that matter, why they land together, and a five-question test to run on every AI agent in production before the board asks about it.  The Q4 calendar: Five dates, one theme Oct 20: Oracle’s patching rhythm changes.Oct 25–28: Oracle AI World, in Las Vegas.Nov 26–30: Thanksgiving to Cyber Monday.Now to January: State AI laws take effect.Year-end: ROI review. Each is manageable on its own.Together, they are the order in which a board is likely to ask questions.  Patching and platform news: Oct 20 and Oct 25–28 Oracle began delivering monthly Critical Security Patch Updates (CSPUs) on May 28, 2026. Oracle’s stated reason is that AI has sped up how quickly vulnerabilities are found and fixed. The quarterly Critical Patch Updates continue as before, so teams now track more patch dates, not fewer. At Oracle AI World, the figures reported for the quarter were cloud infrastructure revenue up 121% to US$7.4 billion and cloud applications up 10%. Data Think’s reading: the money is going into AI infrastructure first, and applications receive the AI features later.  Cyber Week: The date you cannot reschedule Adobe forecasts US$47.5 billion in U.S. online sales for Cyber Week (Thanksgiving to Cyber Monday), up 7.4%. Cyber Monday, Nov 30, is projected at US$15.1 billion, which would be the first US$15 billion day. Black Friday is forecast at US$12.9 billion.  Data Think’s view: a patch window or an agent change scheduled in that week has no room for error, because a failure during the busiest days of the year is revenue you do not get back.  State AI laws and the year-end ROI review State rules are what applies today. The newsletter cites 145 state AI laws enacted across 38 states in 2025 and notes there is still no comprehensive federal law. Connecticut’s SB 5 begins phasing in on Oct 1, 2026, with later provisions staggered into 2027 and 2028. Then comes the ROI review. A WRITER survey reports that 59% of companies invest at least US$1 million a year in AI, but only 29% see significant returns. Expect finance to ask which agents earned their budget.  Who owns an AI agent? Traditional software has a clear chain: someone wrote it, someone reviewed it, someone owns the outcome. An AI agent makes judgment calls in real time and can act (send a message, approve a request, update a record) with no person in the loop at that moment. When it is wrong, “the model did it” is not an answer a board, regulator or customer will accept.  A pilot can run for months without anyone naming an owner. Production cannot. Naming one means: a business role accountable for outcomes (not only the engineering team), limits on what the agent may do alone, a record of what it decided and why, and a review as its scope grows. Teams that can answer this tend to get approval to scale sooner.  Answer these for every agent in production:1) Who owns it?2) What data does it use?3) What KPI proves it works?4) How do we turn it off?5) Where is the audit log?  What to do this week? 1) List every AI agent already running, including small ones.2) Name a business owner for each.3) Write down what each agent may do alone and where a human must approve.4) Confirm each one has a KPI, an off switch and an audit log.5) Add the patch dates week to the same calendar.6) Share the list with your finance and risk leads.  Linkedin Instagram