A Prodigal Paradigm product Outside-in edition
November 1, 2026
Charles Schwab's
Kurultai.
Built from public sources alone: the FY2025 Form 10-K, the 2026 proxy, and Schwab's own careers board. No inside information was used or sought.
What this document is
Schwab is hiring a Director, Responsible AI in San Francisco: requisition 2026-127806, category Risk & Regulatory, pay range $248,300 to $370,000 a year plus bonus eligibility, application deadline 3 November 2026.
This is what the Kurultai produces from the public record alone. Schwab brought nothing to the instrument.
It is also the edition where the instrument finds the least to correct, and that is the point of including it. Schwab has done the thing most organizations get wrong. Reading how they did it is more useful than watching someone fail.
The unbundling, in their own words
Most organizations hiring their first AI leader write one posting that contains five jobs. Schwab wrote five postings.
Their careers page describes the team: "Schwab's AI Strategy & Transformation team, known as AI.x, is the central hub for Artificial Intelligence at Schwab. We are an integrated product, engineering, research, risk, and enablement team focused on advancing AI responsibly across the enterprise."
Product, engineering, research, risk, enablement. Five functions, named. And the board shows them staffed separately rather than described aspirationally - these were live on Schwab's own board on 1 November:
| Role | Requisition | Deadline |
|---|---|---|
| Director, Responsible AI | 2026-127806 | 3 November |
| AI Site Reliability Engineer, AI.x | 2026-127832 | 3 November |
| AI Engineer, AI.x | 2026-121216 | 7 November |
| AI Researcher, AI.x | 2026-123835 | 7 November |
| Sr. AI Platform Product Manager, AI.x | 2026-126348 | 10 November |
Alongside them, Senior Manager AI Risk and Supervision, Senior Manager AI Supervision & Controls, and Director Internal Audit AI Technology. This is a hiring programme, not a hire.
Which is the finding, and it is not a criticism. The split is correct, and it costs five salaries. The organizations that most need the split are the ones that cannot buy it, and they respond by writing one job description containing all five functions and hiring one person to fail at them. What Schwab purchased with headcount, a smaller organization has to purchase with clarity: knowing which of the five it actually needs first, which it already has in the building, and which can wait. That is the staffing map, and it is the reason this instrument exists.
What the posting gets right
It bounds the scope: responsible AI strategy "for client-facing AI solutions," not for everything. It names the partner functions rather than leaving collaboration to hope - "legal, risk, compliance, product, engineering, and research teams." And it asks for operational governance rather than principles, preferring experience "operationalizing responsible AI development practices through governance processes, risk assessments, model documentation, testing, monitoring, and issue escalation."
One detail deserves naming because it is true across this whole industry and almost never said out loud. At a ceiling of $370,000, the requirement is eight or more years in trust and safety, responsible AI, risk or a related field, with at least two focused on AI systems, and three or more building AI risk frameworks. Even at this level, with this budget, nobody can require more than two years of focused AI-systems experience, because the field is not old enough to have produced it. Any organization writing a job description that demands a decade of generative-AI expertise is describing a person who does not exist.
Engine one: the ceremony audit, outside-in
Instrumental,
and unusually concrete
The 10-K's model-risk governance is the real thing, and it covers AI explicitly:
"Schwab manages model risk, including use of artificial intelligence, through use of policies, standards, and controls which evaluate the conceptual and technical soundness of models used by the Company. Prior to the use of any artificial intelligence, we employ procedures which require cross-functional review from applicable oversight functions. We maintain a model inventory that includes a distinct record and risk rating for each model, and we conduct independent validations, annual reviews, and performance monitoring of the Company's models."
A model inventory with a risk rating per model, cross-functional review before any AI is used, independent validation. That is a control judged by whether it governs an outcome, not a statement of values. FY2025 Form 10-K, Management's Discussion and Analysis, accession 0000316709-26-000009.
The gap: governed,
but never described
Here is what the documents do not do. Across 623,636 characters, the 10-K uses the phrase "artificial intelligence" five times: twice as a threat (the sophistication of fraudsters and attackers), once in a forward-looking bullet listing "industry and competitive trends including artificial intelligence, digital assets, private company securities," and twice in the model-risk passage above. It contains no occurrence of machine learning, generative, algorithm or data science. There is no account of AI as a capability Schwab is building.
The 2026 proxy statement - the document where management tells shareholders what it is paid to care about - mentions artificial intelligence zero times in 348,634 characters. No AI, no machine learning, no algorithm.
So: a named enterprise AI organization, hiring across five functions at up to $370,000, called in Schwab's own words "the central hub for Artificial Intelligence at Schwab" - and the shareholder-facing record governs AI as a model-risk control and otherwise does not mention it.
This is not a scandal and nothing here suggests concealment. Filings are conservative by construction and a risk framing is the safe one. But it is a genuine gap with a cost: the investment is real, the governance is real, and the story that connects them exists only on a careers page. A capability nobody has described to the people funding it is a capability that has to be re-justified every budget cycle.
Engine two: the management-decision ledger
| The decision | Evidence | Belongs to |
|---|---|---|
| Whether AI becomes a described capability in the shareholder record, or stays a model-risk line | Five AI mentions in the 10-K, none in Item 1 as capability; zero in the proxy | Investor relations with the Head of AI Strategy & Transformation, before the next annual report |
| Where the responsible-AI function sits when it disagrees with delivery | The role is in Risk & Regulatory but embedded with product and engineering | Written escalation path, before the first contested launch |
| What the capacity AI frees inside operations becomes | The 10-K describes models used in fraud prevention and portfolio guidance; no public statement of where the freed capacity goes | Unassigned. The washing-machine clause: left alone it evaporates into rising expectations. |
| Which of the five AI.x functions a smaller peer should hire first | Not Schwab's decision, but the one their structure poses to everyone watching | The reader of this document |
A row with no owner is not a decision, it is a hope. This table is meant to be signed.
Engine three: the people arc
The posting asks the Director to "translate emerging AI risks, regulatory expectations, and enterprise standards into practical guidance for teams building client-facing solutions." Translation is the whole job, and translation fails silently when the receiving teams have not written down what they actually do.
The arc is the same at this scale as at any other. The teams named in the posting document their own work in their own words, see where AI already touches it, and the standards land inside those workflows instead of arriving as a policy nobody can apply. At Schwab's size the risk is not that the guidance is wrong; it is that it is correct and unread.
Engine four: the staffing map
Schwab's AI.x buys all five archetypes outright. Read against the map: the AI Engineer and Site Reliability roles are Builders and Maintainers, the Researcher is a Prototyper, the Platform Product Manager is a Grower, and the Director of Responsible AI is a Sweeper in the most literal sense - the person who finds what the others missed before a client does.
An organization that cannot buy five directors is not thereby excused from needing five functions. It is required to sequence them. Which one first depends on where the work actually is, which is what the ceremony audit establishes and what no org chart can tell you in advance.
The ninety days, outside-in
By day 30
Connect the governance to the story. The model-risk paragraph in the MD&A is strong and nearly invisible; the AI.x description is vivid and lives on a careers page. One of these audiences votes on the budget. Naming the capability in the shareholder record costs a paragraph.
By day 60
The written escalation path for the Responsible AI function: what happens when it says no and delivery says yes, named role to named role. Cheap now, retroactively impossible.
By day 90
The first decision-ledger review, with the freed capacity from AI-assisted operations traced to an assigned choice rather than absorbed.
A note on the front door, for completeness rather than emphasis: schwab.com has no llms.txt and no AI-crawler policy in either direction, though its homepage does carry Organization structured data and working Open Graph. It is in better order than most. For a firm whose clients increasingly ask an assistant before they ask an adviser, the absent piece is a stated position rather than a repair.
The commitment this document honors
The Kurultai does not produce headcount-reduction recommendations, ever. Nothing in this edition proposes one. Schwab's own documented AI value sits in governance, fraud prevention and client capability, and these recommendations follow that line, because that is where the evidence is.
Sources
- Director, Responsible AI, requisition 2026-127806, Schwab's own careers board, retrieved 1 November 2026; application deadline 3 November 2026
- AI.x roles open on schwabjobs.com, retrieved 1 November 2026
- FY2025 Form 10-K, filed 25 February 2026, accession 0000316709-26-000009
- DEF 14A proxy statement, filed 6 April 2026, accession 0001193125-26-142826
- schwab.com robots.txt and homepage markup, live inspection 1 November 2026
All via SEC EDGAR and Schwab's public web. No non-public information was used or sought. Every quotation was checked against the document it is attributed to, and the term counts above were taken from the full filings rather than an excerpt.