Most companies aren’t struggling with AI because of model quality or tool availability. They’re struggling because their teams don’t trust the systems being rolled out. The pattern is consistent across industries: people hesitate to use AI when they don’t understand how it works, how it affects their role, or what guardrails are in place.
This isn’t a fringe concern, it’s becoming the defining barrier to AI productivity. And it’s starting to show up everywhere, from frontline teams to the biggest institutions in the world.
Today’s stories highlight that shift: the need for trust inside teams, the IRS deploying AI agents with strict oversight, robots stress-tested on factory floors, the music industry embracing licensed AI, and Google turning raw notes into full decks.
Across all of them, the through-line is the same: adoption only scales when people trust the systems behind it.
Topics of the day:
Why team trust is the biggest AI roadblock
The IRS deploys its first AI agents
Figure's humanoid robot trial at BMW
Major labels sign their first AI music deal
Google auto-generates slide decks from notes
The Shortlist: Google hires Boston Dynamics’ former CTO to push Gemini into robotics, Physical Intelligence raises $600M for a general-purpose robot brain, researchers jailbreak LLMs with hidden prompts in poetry, and Uber Eats rolls out sidewalk delivery robots in the UK.
Study: Why your team isn’t using AI
What’s happening: A new study from Edelman reveals a major roadblock to AI adoption: your team's trust. The 2025 Trust Barometer shows employees are more likely to reject AI when they lack clear information and training from their employers.
In practice:
To boost productivity, don't just roll out tools, create a clear policy on how AI will augment, not replace, jobs to build confidence.
Drive efficiency by offering role-specific AI training that shows your team exactly how to automate their most tedious tasks.
Turn adoption into a growth lever by communicating early and often about AI goals, which helps turn hesitant employees into advocates.
Bottom line: Getting your team to use AI isn't a tech problem, it's a people problem. The companies that nail the internal communication and training will see the biggest gains.
The IRS deploys AI Agents
What's happening: The IRS is deploying Salesforce’s Agentforce AI agents across its Chief Counsel, Taxpayer Advocate Services, and Appeals divisions to speed up case handling after a 25% workforce reduction. Humans still make the final calls.
In practice:
AI agents are taking on routine work like case summarization and internal search so teams can focus on higher-complexity issues.
This shows how large organizations can scale output and handle more customer volume without adding proportional headcount.
The IRS setup, where agents cannot make final decisions or move money, provides a practical template for using AI in regulated environments without increasing operational risk.
Bottom line: Government agencies are starting to formalize AI-assisted operations, not as a replacement for staff but as a force multiplier. It offers a realistic playbook for leaders looking to extend their teams’ capacity with AI while keeping human oversight intact.
Humanoid robots get real-world factory test
What’s happening: Figure AI’s humanoid robots just finished an 11-month pilot at a BMW factory, where they worked 10-hour shifts on a live assembly line. The company is now retiring the battle-worn bots and using the data to build its next-gen model for scaled deployment.
In practice:
This is what real-world stress testing looks like for robotics, moving beyond polished demos to see how hardware holds up to grime, long hours, and unexpected issues.
The data on what broke, like specific forearm components, is directly informing the next generation of robots, creating a fast feedback loop for more reliable hardware.
The initial job of loading sheet metal shows a clear path for automating repetitive tasks that require more dexterity than a simple robotic arm but are physically demanding for humans.
Bottom line: Humanoid robots are officially leaving the lab and entering the proving grounds of real industry. It’s a signal for operators to start mapping out which physical, repetitive workflows are next in line for automation.
AI Music gets industry green light
What’s happening: AI music just got its first major industry endorsement, with startup Klay Vision signing landmark licensing deals with all three major record labels to legally train its models on their catalogs.
In practice:
This gives creators a legal path to generate high-quality background music for videos and podcasts, automating a costly bottleneck.
Klay's license-first approach offers a blueprint for other AI startups building on copyrighted material, from images to text.
For established companies, it shows a new way to monetize existing IP by partnering with AI developers instead of just suing them.
Bottom line: This signals a major shift from litigation to collaboration in the AI space. Legally licensed models will likely become the gold standard, creating more predictable and scalable business opportunities.
Google's NotebookLM now makes slide decks
What’s happening: Google’s AI-powered research tool, NotebookLM, now auto-generates slides from your notes and source documents, turning dense material into a customizable first-draft presentation.
In practice:
This kills the “blank page” problem by instantly creating first-draft presentations from your research, reports, or meeting transcripts.
You can accelerate your workflow by turning project briefs or performance data into client-ready decks in a fraction of the time.
Use it to quickly distill key insights from multiple documents into a cohesive narrative for team updates or stakeholder briefings.
Bottom line: This moves AI from a simple research partner to an active collaborator in creating professional assets.
The Shortlist
Google hired the former CTO of Boston Dynamics to help turn its Gemini AI into a universal operating system for robots.
Physical Intelligence raised $600M at a $5.6B valuation to build a general-purpose AI "brain" that can be licensed to power different types of robots.
Researchers found a way to jailbreak large language models by hiding harmful instructions inside poems and metaphors, successfully tricking them 62% of the time.
Uber Eats partnered with Starship Technologies to deploy sidewalk delivery robots in the UK, with plans to expand to the US in 2027.
This newsletter is where I (Kwadwo) share products, articles, and links that I find useful and interesting, mostly around AI. I focus on tools and solutions that bring real value to people in everyday jobs, not just tech insiders.
Please share any feedback you have either in an answer or through the poll below 🙏🏽
