Humanity Labs gives wealth management firms an AI Workforce that makes growth profitable. In a people-based business, serving more clients means hiring more people, so cost scales with revenue. The AI Workforce breaks that trap.

It works inside the systems your team already runs and delivers done work across the front, middle, and back office, while your people stay focused on the decisions that matter.

Every job it runs writes into one shared memory, building the firm's Organizational General Intelligence (OGI) so it gets smarter as it grows.

The result is the Firm of the Future: a firm that serves more clients and grows without adding cost at the same rate.

Humanity Labs works with some of the world's leading wealth management firms, which together oversee more than $750 billion in client assets, and is backed by leading venture firms including Google Ventures and Vestigo Ventures.

Learn more at HumanityLabs.ai

Organizational general intelligence is the question of how a company knows what it knows. This period it became a question about storage. The record a company's work leaves behind is the raw material its systems now learn from. It is also what regulators and opposing counsel ask for.

On Wednesday OpenAI previewed a system it calls Private Safety Processing, built to spot misuse spread across a customer's separate conversations without keeping any of them. (OpenAI, Aug 19) The next day Bloomberg reported that Anthropic plans to let business customers hold their required 30 days of logs on their own cloud rather than on Anthropic's. (PYMNTS, Aug 20)

Why it matters: The object in dispute is the same in both cases. It is the record of what an AI system did while it worked. OpenAI is building to eliminate it. Anthropic is building to move it somewhere the customer controls.

Between the lines: The case for keeping the record rests on harm that no single exchange reveals. Aleah Houze, OpenAI's head of product policy, said the company is seeing risks that emerge not from one prompt and response pair but across several, and gave the example of someone asking about a weakness in a company's software in one conversation and about remote access in another. (Axios, Aug 19) Anthropic reached the opposite conclusion from the same observation. It has required 30 days of retention on Claude Fable 5 and Mythos 5 since June, and told customers it expected them to dislike it.

A third position holds that the record should not exist. Writing in Law.com this month, Chad Ergun argued that firms should demand zero retention from AI vendors, on the grounds that nothing preserved is nothing that can be waived or produced. His premise is that the model is no longer the advantage: every AmLaw 100 firm, meaning the hundred largest law firms in the United States by revenue, can now buy access to the same frontier models. (Law.com, Aug 18)

What is next: Europe would have settled part of this in August. Articles 19 and 26 of the AI Act put a floor of at least six months on the automatically generated logs that both the provider and the deployer of a high-risk system must keep, and those obligations were due to apply from August 2. Regulation (EU) 2026/1744, the Digital Omnibus on AI, entered into force on July 27 and moved the deadline for standalone high-risk systems to December 2, 2027, and for AI built into regulated products to August 2, 2028. (EU AI Act, Article 26) The floor survived. The date moved.

Yes, but: The deferral does not reach everything. Transparency duties under Article 50, the obligations on general-purpose model providers, and the sanctioning framework all became enforceable on August 2 as scheduled. A company that fine-tunes a bought system on its own data or ships it under its own brand can also be reclassified from deployer to provider, which moves it into the fuller set of obligations without anyone deciding to take that step.

Until December 2027, the record of AI work is governed by whichever vendor a company happened to buy from. One keeps nothing, the other keeps 30 days, and the six-month floor arrives later at the deployer's own expense.

Amazon Web Services published a system on Wednesday whose premise is that the useful knowledge is already written down, in the wrong place. KnowledgeForge reads resolved IT service tickets and turns them into knowledge base and root cause articles, because, as consultant Anmol Dhankhar and three colleagues write, each ticket holds a symptom, a root cause, and the fix an engineer applied. (AWS Machine Learning Blog, Aug 19)

By the numbers: The design puts a person at the end. Finished articles go to ServiceNow, the IT service management platform where the tickets originated, for knowledge-manager review, and the approval itself is written to a database. The record of who signed off becomes a second record the company keeps.

Yes, but: The record of what agents themselves do is thinner than the record of what people did. Salesforce documented that its agent audit logging captures only new events going forward and that historical usage from before the toggle is not backfilled. (Salesforce, Aug 18) A company can mine years of human tickets and still hold no account of last quarter's agent work.

The big picture: Vendors are placing the human at opposite ends of the same task. Chief executive Chuck Robbins and his finance chief told investors that 145,000 support cases were resolved entirely by AI with zero human intervention in Cisco's 2026 fiscal year, a company claim about its own product made on an earnings call. (CX Today, Aug 13) Under Amazon's design the equivalent number would be zero, because a knowledge manager signs every article before it publishes.

Joseph Conroy, who founded Harford Retirement Planners, has a rule for people in their first years at the firm. "We pretend AI doesn't exist in the early learning stages of a new advisor," he said. (InvestmentNews, Aug 11) The work he withholds the tool from is data gathering, report assembly and documentation, which is the work that used to constitute the job for three to five years.

The big picture: Conroy drew the distinction that makes this more than nostalgia. "Not plugging in the inputs is different from not knowing them," he said. His argument is that someone who never assembled the report can still direct a system that assembles it, and will not recognize a wrong result when it arrives.

Catch up quick: The same question is reaching professional rules. The New York City Bar Association's policy paper this month lists supervise among six duties for lawyers using AI tools, alongside understand and verify, which places the obligation on whoever signs the work rather than on the tool. Chris McMahon, who runs MFA Wealth and Aquinas Wealth Advisors, described the mechanism in ordinary terms: mentorship and team development develop real judgment and foster accountability.

Between the lines: The projected advisor shortage of roughly 100,000 by 2034 reads as a hiring problem. Set against Conroy's rule it also describes a supervision problem, because the people who would review AI-produced work in 2034 are the people learning the job now. Reporting this month on the same pattern in office work found the tasks disappearing first are research, data cleanup and preliminary memos. (IBTimes UK, Aug 2026)

This magazine, and all the magazines in the Humanity Labs AI Magazine series, were created using Hawkeye, an AI system built for people who work in fields too large and too fast-moving to keep on top of alone.

The stories here are based on a thematic cluster analysis of thousands of articles each week. Those articles range from mainstream news to individual company newsletters. They're analyzed using an inductive reasoning process for trend discovery, a data feature selection process, and a novelty filter for determining which developments are most significant in their respective field.

The analysis is performed and the first draft is written with Claude Opus. It's then fact checked using OpenAI. It's read over and reviewed by series editor and Hawkeye founder Marshall Kirkpatrick, a serial entrepreneur, data journalist, and the first writer hired at TechCrunch. Marshall's writing has been cited twelve times in the WSJ, by ABC, NBC, CBS, the BBC, the American Medical Association, the UN, and many more.