Trump’s “Super Intelligence” order, small-business AI training, Airbnb’s trip tools and a federal safety investigation.
 |
| Thursday, October 1, 2026 · Edition 323 |
Good morning. President Trump has ordered federal agencies to call AI “Super Intelligence.” Small-business advisors are getting AI training, Airbnb is making trip comparisons easier, and the FTC has confirmed a safety investigation. Also inside: hiring, accessible websites, local services and an editable PowerPoint flowchart tool. |
|
TOP 5
Five stories worth your time.
|
 Airbnb’s new comparison view summarizes homes side by side. Image: Airbnb.
|
|
01
Trump signed an executive order Tuesday directing federal agencies to use “Super Intelligence” and “SI” in place of artificial intelligence and AI. The change covers executive-branch communications, websites, reports and policy documents, where the law allows.
The order keeps the existing legal definition of AI for now and does not require rewriting old regulations, contracts or grants. Proposed legislative language for a federal definition is due within 60 days. The terminology change does not establish that the technology has reached superintelligence.
Separately, companies signed a voluntary safety accord calling for internal controls, outside reviews and board oversight.
White House / signed executive order · DW / Sept. 30 reporting · Safety accord · Sept. 29 order; Sept. 30 reporting
|
|
|
02
America’s SBDC and OpenAI are starting a pilot that trains business advisors to lead hands-on AI workshops for more than 1,000 small businesses. Participants who complete the workshops will receive free ChatGPT Plus access; the announcement does not specify how long it lasts.
The rollout is planned over the coming months. Local dates and participating centers are still to come. The partners also plan sector-specific guides, giving owners a way to learn with an advisor who understands their business.
America’s SBDC / partnership announcement · Sept. 30
|
|
|
03
Starting in the U.S., travelers can describe a trip by text or voice, find relevant filters and compare homes from their wishlist with AI summaries. Listing descriptions also highlight a property’s main features.
Airbnb says AI chat support now covers more than 50 languages and is expanding to phone support. New host pricing and earnings tools are due later this fall. Travelers should still check the listing’s actual amenities, total price and cancellation terms before booking.
Airbnb / fall update · Sept. 30
|
|
|
04
The Federal Trade Commission confirmed Wednesday that it is investigating OpenAI, Anthropic and other AI labs over consumer risks. The inquiry began during the summer.
Bloomberg reports that the agency is preparing formal requests for information about product safety and compliance with consumer-protection law. An investigation is not a finding of wrongdoing. It adds federal scrutiny as companies release agents able to take actions across software and websites.
The Next Web / FTC confirmation and reporting · Bloomberg / information requests · Sept. 30
|
|
|
05
In HP’s survey of 19,506 desk-based workers across 15 countries, including the U.S., 47% reported using AI agents at work. Nearly half said their employer lacked a clear AI strategy. Among knowledge workers, 73% said investment in their development would make them more likely to stay.
The survey was fielded in April and May and released Wednesday. These are workers’ reported experiences, rather than proof that adding AI causes better retention. The findings give managers a useful question: do employees know which tools they can use and how to build their skills?
HP / Work Relationship Index · Sept. 30
|
|
|
AI WORKFLOWS | THURSDAY: NEW-OPPORTUNITY IDENTIFICATION
An AI skill can find the customer need your next service should address.
Anthropic’s September 25 plugin announcement gives developers a new way to distribute reusable skills alongside app connections. One useful example already exists: its public Competitive Intelligence skill, which compares customer evidence with competitors’ public moves.
A skill is a reusable set of working instructions. This one can examine won and lost deals, preserve customer quotes with their sources, and identify recurring gaps. Thursday’s application: use that evidence to choose a service idea worth testing.
Say you manage accounts for a cleaning company. You suspect small property managers need Saturday move-out cleaning. Before asking for staff or advertising money, delegate a check of the past 90 days in one service area. The idea remains a hypothesis until customers and your operating numbers support it.
Give the agent a method and access
In a paid Claude account, review the Sales plugin under Customize → Plugins. Use its Competitive Intelligence skill in Cowork, then customize the output into an opportunity brief. If your account has the new unified Claude experience, start from its main conversation. Connect only company-approved customer systems and documents; keep research access read-only. Workspace rules may limit available plugins and connections.
Have Cowork coordinate parallel checks, then reconcile them in one editable document:
Customer requests
Read approved account records and email. Count distinct customers asking for Saturday service; keep repeat messages from inflating demand. |
Existing alternatives
Check three named competitors’ current service pages. Mark unpublished hours or prices as unknown. |
Delivery constraints
Read approved staffing and cost assumptions. Show what a Saturday job would require, including travel and overtime. |
Give it this assignment
| Use the Competitive Intelligence skill to assess Saturday move-out cleaning for small property managers in our chosen service area. Use the approved sources and three competitors we named. Produce an editable brief with linked evidence, unique-customer counts, contrary findings, coverage gaps and a cost worksheet. Separate observed facts from estimates. Explain missing access. Recommend one small customer test and what result would make us stop. Keep customer details inside our approved workspace. Ask before contacting anyone, changing records, publishing or spending. |
Before approving a trial, open the strongest customer evidence and verify the cost assumptions. A competitor’s silence about Saturdays proves only that its website is silent. Ask real customers what they currently do, what fails and whether they would book a clearly priced trial.
For managers: Review the brief with operations before making promises. After a useful first run, authorize a weekly refresh of the same sources, with a short note only when new evidence changes the recommendation.
Start with one customer group, one service area and one decision about whether a trial is worth offering.
|
|
|
AI TOOL SPOTLIGHT · Lucen Flowchart
Lucen’s PowerPoint add-in can draft a flowchart from a plain-English description. The boxes and connectors remain ordinary, editable PowerPoint shapes. You can change a decision, group steps by owner and tidy the layout without rebuilding the diagram.
Try a familiar process, such as approving a customer refund. Describe the actual steps and who decides each one, then have the people doing the work check the draft. It requires PowerPoint 2016 or newer on Windows, including Microsoft 365 desktop. Mac, web and mobile PowerPoint are not supported.
Lucen / official product page · Setup and compatibility · Current product guide
|
|
|
MORE AI NEWS
Headlines
|
Work and skills
BCG’s new research puts the share of companies generating meaningful value from AI at nearly half. It also finds that 42% expect agents to have real decision-making authority by 2030, while only 5% have the necessary controls ready. These are the consultancy’s findings and respondents’ expectations.
BCG / Applied AI Index · Sept. 30
Draup’s analysis of Fortune 500 hiring puts AI-builder roles at 27% of technology hiring demand, more than twice their 2021 share. It also finds AI skills appearing in customer support, sales, finance and HR postings. The analysis measures advertised demand, rather than completed hires or the whole U.S. job market.
Draup / hiring analysis · Sept. 30
Among 1,500 technology executives, technologists and learning leaders surveyed in the U.S., U.K. and Australia, 96% said their organization had invested in AI for at least some staff. Only 52% considered most or all employees AI literate. The training provider’s survey measures respondents’ judgments, not tested proficiency.
Pluralsight / research release · Sept. 30
New research from the Center for Talent Innovation warns that cutting entry-level tasks and management layers can weaken learning, relationships and institutional knowledge. Its findings draw on surveys, interviews and executive discussions. The report urges employers redesigning work to preserve ways for less-experienced staff to develop judgment.
Center for Talent Innovation / research release · Sept. 30
The company says weekly voice-agent conversations have tripled since February. A completed $300 million employee share sale values ElevenLabs at $22 billion, twice its February valuation. The transaction lets existing holders sell shares; it should not be read as $300 million of new operating money for the company.
Reuters / WTAQ · Sept. 30
|
|
Tools for business operations
The partnership brings Ramp-powered bill payment and corporate expense cards into ServiceTitan. Card transactions sync into ServiceTitan, where purchases can be matched to jobs and receipts captured digitally, building on invoice review and approval tools. The aim is less financial re-entry for trade businesses; card access remains subject to credit approval.
ServiceTitan / partnership announcement · Sept. 30
Miter says more than 2,000 contractors use its workforce, payroll, field operations and expense platform. Its AI tools organize invoice and receipt data, flag job-cost trends and help draft safety reports from photos or voice notes. The new funding supports expansion; the announcement does not make every described feature a new launch.
Miter / funding announcement · Sept. 30
The integration lets UiPath read and model Snowflake data where it already lives, using existing access controls. Teams can use the data when designing and running automated work, while Snowflake’s assistant can call UiPath skills. It is an enterprise connection for organizations already using these platforms.
UiPath / integration announcement · Sept. 30
Stravito’s partnership with Langston gives brands simulated consumer perspectives based on 24 motivation segments or their own research. Teams can ask questions and explore reactions to an idea, with supporting evidence intended to remain traceable. These simulations can suggest questions for customer research; they do not establish that real people will buy.
Stravito and Langston / partnership announcement · Sept. 30
|
|
Health and local services
The company says its agent now serves more than 240 local governments, and 63% of this year’s conversations occurred outside office hours. The service answers from government content and connects with municipal websites and service tools. The figures describe vendor-reported usage, without an independent accuracy audit.
CivicPlus / usage report · Sept. 30
Researchers analyzed Fitbit data from 134 women with chronic pelvic pain and 61 comparison participants. With about ten days of personal data, models could predict activity an hour ahead and identify sedentary periods. Future trials must test whether well-timed movement prompts improve symptoms; this study establishes prediction performance, not a treatment benefit.
Mount Sinai / research announcement · Sept. 30
|
|
Trust, rights and public decisions
New reporting on Google’s existing AI contribution pilot finds some small and midsize publishers receiving less than 0.1% of their advertising revenue. Other participants receive much more. The pilot pays for content used in AI answers, but the early results do not show a consistent replacement for search-driven advertising income.
Search Engine Roundtable / The Information reporting · Sept. 30
In a commissioned survey of 2,000 U.S. adults online, 63% said they would or might let AI choose brands if it consistently saved them 20%. But 64% would not let it spend without approval. These are stated preferences, highlighting a distinction retailers must respect between recommending a purchase and authorizing one.
Spreedly / Talker Research survey · Sept. 30
AudioEye asked five AI coding tools to build three websites each using explicit accessibility requirements. Its testing found 306 issues across the 15 sites, with 91% rated medium or high severity. The vendor-run test is small, but it shows why an accessibility instruction alone is an inadequate acceptance check.
AudioEye / study announcement · Sept. 30
The Alliance for Critical Infrastructure now includes nearly 50 companies across sectors such as finance, energy, water and communications. Its agenda includes safer open-source software, secure adoption of AI language models and coordinated responses to attacks. The coalition addresses both physical and cybersecurity risks to essential systems.
Alliance for Critical Infrastructure / announcement · Sept. 30
The court recognized that an actor’s voice can fall within publicity-right protections in a dispute involving alleged AI imitations of Kenjiro Tsuda. It dismissed the request to remove the TikTok videos because they had already been deleted. The ruling is a Japanese legal development, with no automatic effect on U.S. law.
ITmedia / Tokyo court reporting · Sept. 30
Andrew Bailey argues for rigorous testing and workable intervention points as more capable AI raises risks for banking and payment systems. His remarks focus on how institutions would respond if a system became unsafe. They are a call for safeguards, rather than a new binding rule.
Bank of England / Andrew Bailey · Sept. 30
|
|
The computing behind AI
HPE says the Vultr deal is its first order for AMD Helios AI Rack systems. The equipment is intended for U.S. cloud data centers running AI training and serving model responses. The announcement describes an order and planned deployment; it does not establish that all the systems are already operating.
HPE / company announcement · Sept. 30
The multiyear agreement covers chip-design technology and engineering software for Amazon’s custom silicon, including Trainium AI chips. Synopsys will also expand its use of AWS infrastructure and Bedrock. The deal spans Amazon’s broader chip portfolio, including computing and networking products.
Synopsys / official announcement · Sept. 30
DeepSeek has open-sourced computing and communication tools optimized for Huawei’s Ascend chips. The companies also describe a system linking 128 Ascend 950 chips. The work gives Chinese AI developers another route for building on domestic hardware, as they try to reduce reliance on Nvidia’s software ecosystem.
Reuters / Yahoo · Sept. 30
|
|
|
mAIn Street gives nontechnical readers useful AI news they can put to work.
mAIn Street #323 · Thursday, October 1, 2026
Reporting checked through September 30, 2026.
|
|
|
|
