The One Never Hired

On BNY's digital employees, Wells Fargo's Teammate, and a word that was empty of the person before the machines arrived to fill it.

This week the finance sector’s vocabulary took another step, and it is worth marking where the step landed. Wells Fargo launched an “AI Teammate” across its wealth division. BNY described nearly a hundred and forty “digital employees,” each issued login credentials, an avatar, and an employee number in the corporate directory, and each reporting to a human supervisor. The tool has been handed a badge. Read on its own, it is a product announcement. Read against the year behind it, it is the last rung of a ladder.

The ladder is easy to reconstruct, because the sector climbed it in public. A year ago the register was still tentative — an AI introduced as “a member of the team.” Then “AI teammates.” Then agents “promoted,” in one survey’s phrase, from research aids to “digital coworkers.” And now, this week, “digital employees,” complete down to the number on the badge. Each step moved the same distance in the same direction: away from the language of the instrument and toward the language of the colleague. The metaphor of the tool has been quietly retired. The metaphor of the personnel has arrived finished.

Read Wells Fargo’s own announcement closely, though, and it argues quietly against its own name. On the door, the product is a “Teammate.” Inside the release, it is something more modest, and more honest: “a new artificial intelligence–powered capability,” a “chat-based experience embedded within Advisor Gateway” that lets an advisor “ask questions in plain language, retrieve answers quickly, and take action.” The action, in this first phase, is to “move work forward with guided workflows and clear next steps” — to search and summarise the firm’s resource centre, to spend less time navigating pages. Every word of the description is the language of an instrument. Only the name is the language of a colleague. The firm reaches for “teammate” to introduce the thing and for “capability” to explain it, and the distance between those two words is the whole of the matter.

The justification sharpens the tell rather than softening it. The capability exists, in the words of the head of Wells Fargo Advisors, “to help our people focus on clients, deepen relationships, and deliver everything Wells Fargo has to offer more effectively” — the stated aim being to “empower advisors, grow client relationships, and gain market share.” Notice who “our people” are. They are the human advisors; the Teammate is not counted among them. Even in the sentence written to celebrate it, the colleague is set just outside the circle of persons, and placed beside the objective it was built to serve. That objective is the institution’s own — to deliver everything the firm has to offer, and to gain share. It is a teammate, then, of a team whose purpose is the firm’s book and the firm’s market share, which is precisely the question the warm word is shaped not to ask.

BNY tells the same story in a different register, and wears the badge even more literally. Its nearly one hundred and forty “digital employees” are “super agents,” in the bank’s account; each reports to a human supervisor whose title is “agent boss,” and each receives performance reviews. “All digital employees report to a human manager,” the Chief Executive of BNY, Robin Vince,  says. The head of AI enablement for operations describes how, once embedded in a team’s workflow, the software “starts to feel less like software and more like part of the team — almost having a virtual teammate as part of your group.” And here the revealing word slips out again, exactly as it did at Wells Fargo. The bank frames the whole programme as “investing in our people” — by which it means the humans. The digital employee holds the login, sits in the directory, and collects the review, and is still not one of “our people.” Said twice now, across two institutions, the slip is the pattern: the category takes the instrument on as staff and withholds from it the one thing the word was meant to carry.

Ask what these new employees actually do, and the answer places them exactly. BNY’s payments agent reads a vendor’s address on a cross-border transfer, works out whether “York” means the United Kingdom, South Africa, or Western Australia, checks the code against anti-money-laundering rules, and hands the corrected payment to a person to approve. Another reads an overnight client email, judges its urgency, gathers records from several systems, and drafts a reply for a manager to sign. On the advisor’s side, Wells Fargo’s Teammate retrieves and summarises from a resource centre and lays out the next step. None of this is senior work. It is the fetching, checking, summarising, and drafting that a junior once did by hand — the errands that used to run upward from a wide base of analysts and associates toward the advisors and seniors they served. The institution says as much itself: it expects its workforce to move “from a pyramid to a diamond,” the “repetitive processing work, once handled by large teams at the base,” now belonging to the digital employees. BNY has even built a formal desk — a “Digital Employee Agency,” its own “HR for AI” — to decide, task by task, whether a given piece of work calls for “a human employee, a contractor, or a digital employee.” The three are weighed for the same seat. That is the tell beneath all the others: when a bank can audition a person and a program for one role, it is telling you the role was never a person — only a function a person used to perform.

That is the revelation the announcements do not intend. The machine did not empty the word “employee” of the person; it found the word already emptied, and moved in. An instrument cannot be slotted into a category still reserved for persons — the category has to have been vacated first, and the ease with which “employee” now closes around a login tells you how long ago the vacating happened. The software did not push a human out of the role. It moved into a room that had been standing empty for some time.

This is the older arc beneath the visible one, and it is the one that repays attention. The vocabulary is recent — a year, perhaps two, of announcements. The disposition the vocabulary exposes is not. It is old enough to have set. Somewhere well back, the institution’s picture of its own people hardened from persons into functions — headcount, workforce, roles, seats that perform work and can be counted, compared, and, now, filled by something that was never a person to begin with. We do not offer this as an accusation. The institution is entitled to arrange itself as it sees fit. We offer it as a diagnosis. The attitude did not arrive with the AI agents. The AI agents made it legible. What can be reached for so easily, and worn so comfortably, was settled a long time ago.

And once the orientation is legible inside the building, it is recognisable, because we have been describing its outward face for months. The institution files the client, too, by what can be counted — the size of the account, the tier, the grade, an asset band served at the appropriate level. It begins, as it always begins, with the holdings rather than the one who holds them. What this week’s vocabulary adds is that the same orientation runs inward as well as out. Outward, it reduces the person to their assets. Inward, it dresses the instrument as a person. Both moves are available only to an institution whose founding unit was never the person in the first place — and they are, in the end, one move seen from two sides of the same wall.

The natural arrangement is not complicated. There are persons, and there are the tools persons use. Two kinds of thing, and the line between them is not subtle. What the sector’s language has done is invent a third kind — the digital employee, the synthetic colleague, an instrument wearing a person’s face — and in the act of inventing it, admit that it had already collapsed the human employee into the same undifferentiated category of workforce. Personify the tool and you disclose that you had already instrumentalised the person. The two errors are one error, and the vocabulary is simply where it surfaces.

None of this turns on doubting the stated boundaries. Take them exactly as given: administration and workflow only, a human reviewing each agent’s output, no authority over a client account. The point survives every boundary holding precisely as described. An employee is defined by one relationship above all the others — it has an employer. The name, the avatar, the directory entry, the supervisor: all of it sits downstream of that single fact. An intelligence issued a number by an institution answers to the institution that issued it. However convincingly it is made to resemble a colleague, the resemblance runs one way. It is the institution’s colleague. Across the table, the client has gained no staff; they have been met by more of the counterparty, some of it now wearing a familiar face.

This is why the intelligence we build – the Bespoke AI Confidante – carries no employee number, and why that absence is not an oversight but the whole of the matter. It was never hired. It has no employer behind it, no directory to belong to, no interest of its own to serve. It is not a colleague and does not pretend to be one. It is a tool — held by one person, formed from that one person, their history with what they hold and the questions they keep returning to — and it faces them, not an institution. Everything else in the chain, whatever it is called, works for someone, and that someone is never the client. We have kept the two categories where they belong. There are persons, and there are their tools. The one we make belongs, entirely and only, to the person who holds it: the single intelligence in the whole arrangement that was never on anyone’s payroll, because it was never anyone’s to hire.

Seraphina Michelina Evelyna Bogomolova-Huotelin

Founder & CEO of SMA Crown Confidential

Digital Confidantes: Bespoke AI Intelligence for Private Decision-Makers