AI Billing Costs, a Superintelligence Ban, and Amazon’s Seller Tools
Hi Friend,
In our AI Smart Lab session last week, we cleaned up our AI search tool. We went through every account we were paying for, removed what we no longer use.
Then Andrea walked us through what the tool was telling us. Visibility looked fine. Citation rate, which counts how often AI tools quote your pages instead of just crawling past them, could use some improvement.
We also set up a plain-text copy of a site on a separate subdomain so AI crawlers can read it without weighing down the main site and we connected an agent that drafts content and sends it to WordPress for a person to approve before anything goes live.
When did you audited your tools?
What to Demand From Your AI Consultant
If you are paying for AI help and cannot say what it has changed, the problem is usually the agreement, not the technology. Andrea wrote this article on the numbers to ask for before you sign: hours saved, cycle time, adoption rate, cost savings, and time to value. She also lays out the 30, 60, and 90 day milestones that keep an engagement honest after the kickoff excitement wears off.
The red flag she names is simple. If a firm leads with its favorite tool instead of your business problem, you are buying a demo.
Our Most Requested Service: AI Strategy and Consulting
The question clients bring us right now is almost always the same. They know AI can help. They do not know which tools to trust, how to protect their data, or how to get the team to actually use all of it.
We start with your workflows, not with software. We map where the time is spent, then build a phased roadmap with secure tools and clear owners, and we train your team to run it after we leave.
One client, a medical training company, worried the whole way through that AI meant cuts. Nobody was laid off. Every person was upskilled and moved into higher-leverage work, and the hands-on training led to automating 99% of their customer service operations, worth $177,000 in combined savings and repurposed labor. We have trained more than 20,217 professionals this way.
This week’s stories
Blue Cross Says Hospital AI Billing Tools Added $942 Million to Costs
The Blue Cross Blue Shield Association looked at its own claims and found that hospitals are coding more inpatient stays as medically complex. The share rose from about 37% in early 2023 to about 40% by the end of 2025, which it estimates added $942 million in costs across 2024 and 2025. About 70% of that came from secondary diagnoses, the extra conditions that move a claim into a higher-paying category. The insurer group says more than 60% of hospital systems now use AI coding tools, and it found no matching change in the care patients actually received. Hospitals dispute this, and the American Hospital Association says the evidence does not support it.
That bill does not stop with the insurer. It lands in the premium you pay for your team next year.
My Take: Both sides in this fight bought software, and you are the only one in it without a machine. Hospitals run AI to find every billable condition, insurers run AI to deny and downcode, and the employer paying for the plan sits between two systems arguing at a speed no human in your HR office can audit. Coding intensity moved 3 points and care did not move at all, which is a number that looks like progress until somebody checks what sits underneath it. We hit a smaller version of this in our AI Smart Lab session last week, where our search visibility score looked healthy and the citation rate under it did not perform at the same level. My call for 2027 renewals: the plans that hold their pricing will be the ones that started auditing coding patterns this year, not the ones that negotiated harder in November. When your renewal quote arrives, do you know which line of it is care and which line is coding?
Source: Blue Cross Blue Shield Association
Anthropic’s AI Found a New Enzyme System in About a Day
Anthropic says it pointed roughly 950 Claude agents at a biology problem for about 21 hours and came back with a previously undescribed enzyme system that has CRISPR-like repeats. The agents screened more than 200,000 reverse transcriptases, narrowed 3,500 candidates down to about 20, and wrote up the result. Feng Zhang of MIT and the Broad Institute called it intriguing and worth further investigation. Other scientists are openly skeptical that it rivals CRISPR, and the write-up sits on Anthropic’s own site without peer review.
My Take: The biologymay or may not not hold up, and the operations lesson will. What caught my eye was was 950 agents running in parallel against one question for 21 hours. We have built a couple dozen agents across ClickUp and Claude, and the hard part is not the model, it’s deciding what each agent owns and who checks its work. Within a year the thing companies brag about will not be which model or tool they use, it will be how many unique agents they can aim at a single problem without the output turning to mush. That said, a lab publishing its own unreviewed result and calling it a discovery is a press release with footnotes, so trust AND verify.
Source: Anthropic
Two Lawmakers Want to Ban Superintelligent AI in the US
Senator Bernie Sanders and Representative Greg Casar introduced a bill that would permanently ban artificial superintelligence and pause advanced AI development in the US until a new cabinet-level Department of Artificial Intelligence writes the rules. The penalties are steep: dissolution for companies, and up to 20 years in prison for individuals. The bill also directs the US to seek international agreements and export controls.
This is a proposal, not a law. It’s a long shot in a Congress that has not agreed on far smaller AI bills.
My Take: Washington is legislating against a machine that does not exist yet while the AI that is already taking money out of your business is dull as dishwater. Everything in this issue that touches a real budget is unglamorous: a coding tool that reclassifies a hospital stay, a state rule about claim denials, an agent that publishes something nobody read first. None of it needs superintelligence to cost you money. In our AI Smart Lab session last week we spent zero minutes on existential risk and a long time on which subscriptions renew next month and whether an agent’s output gets checked before it ships. My prediction: this bill goes nowhere, and the 20 year prison provision still shows up in some random state bill that does pass, because personal liability is the one idea in here that legislators of both parties can sell. Would your AI setup survive somebody asking who, by name, approved the output?
Source: Office of Senator Bernie Sanders
Amazon Sellers Can Now Pull Their Store Data Into Claude
Amazon opened up its Seller Assistant with a plugin that connects a seller’s Amazon data to Anthropic’s Claude and to Amazon Quick. Seller Assistant now reaches more than 90% of Amazon’s selling partners, and Amazon says sellers accept its recommendations more than 90% of the time. The new always-on workflows come with audit trails and actions the seller has to approve. The plugin is in US beta and connects in about 60 seconds.
My Take: Amazon just admitted its sellers are not going to stay inside Amazon’s tools, and handing them a pipe to a competitor’s model is the smarter play than pretending otherwise. The detail worth copying is not the plugin, it is that every always-on action still waits for a seller to approve it and leaves a record of what it did. We used this same idea in our AI Smart Lab session last week: the agent drafts the content and pushes it to WordPress, and a person clicks publish. Most of the agent disasters I expect over the next year will not come from bad models, they will come from teams that skipped that click. Is your product data clean enough that an AI tool reading it would describe you correctly?
If you need help with AI SEO, let us know. We’ll send you a free report to start and then help you get in front of your customers. Email [email protected] and send your URL as well as your competitors for the most robust info.
Source: Amazon
Tool picks of the week
ClickUp: A platform to manage tasks, notes, and chats in one place. We run most of our agents here.
Canva: A design platform for images, videos, and presentations with ready-made templates.
Gamma: Builds presentations, documents, and simple web pages fast.
Got a burning question, a fresh take, or just want to share your latest AI wins? Send us a message [email protected]. Your insights keep this community growing and thriving!
Have a great day!
-Nicole A. Donnelly
Founder, AI Smart Ventures
AI Strategy – AI Training – AI Consulting – AI Implementation
