Four operators, sixty days, the same sentence in different words. The unit they want to buy has changed, and most vendors still cannot hear it in the room.
Last week I was in a healthcare operator’s office, my deck up on a big screen, his leadership team around a conference table that had not been used for a pitch in a while. It was the fourth room like that one in two months.
My third slide was up. Twenty-plus people for his back office, compliance, claims. The standard team-build answer under the old playbook.
He pushed his laptop closed and looked up at me.
The specifics change every time but the idea does not. Paraphrased, across four rooms now: stop sending more people, send Virtual Employees.
I closed the deck.
We spent the next two hours at the whiteboard, trading the marker back and forth, the cold coffee still in the cups between us. I walked out with a different contract than the one I walked in to sell.
Same Word. Two Different Things.
He was not telling me AI is replacing his team. He was telling me the unit he wants to buy is a Virtual Employee, and the unit I was selling was a person.
Same word. Two different things.
A team is priced per seat, hired per role, planned on an org chart. A Virtual Employee is a persistent, governed system that holds business context across months of work and performs defined tasks inside a regulated workflow. A buyer who asks for one and gets the other has been told they are the same thing. They are not.
Same word. Two different things.
Four Rooms. Sixty Days.
The first time it happened I treated it as the operator’s idiosyncrasy. A buyer who had thought about AI longer than most. Interesting, probably not general.
The second time I started to pay attention.
The fourth time I stopped bringing the old slide.
Four operators. Sixty days. Same sentence in different words. Different industries, different scales, different stages in their own AI journey. The idea shows up unprompted now, in the part of the meeting where the buyer decides what they are actually trying to buy.
The Two Hours at the Whiteboard
Once the deck is closed, the conversation stops being a pitch and starts being a working session. The things that come up are not abstract. They are the scars any operator already carries from trying to run a Virtual Employee in production.
There are four of them, and they show up every time.
Token costs that quietly double on the wrong prompt structure. A workflow that worked at the demo stage runs inside a regulated back office and the spend doubles or triples before anyone notices. Scoping the engagement requires knowing which prompt patterns are expensive, which are cheap, and which drift over time. This is an operator problem, not a procurement problem. It does not appear in a rate card.
Governance around who approves what the Virtual Employee does, and who answers when it gets something wrong in a regulated workflow. A human employee has a manager, a scope of authority, and a record of decisions. A Virtual Employee needs the same thing. Who reviewed the output. Who signed off. What the escalation path looks like. Governance is not a slide; it is a set of documented controls that survive an audit.
Persistent memory, which is where most of these efforts actually break. A Virtual Employee that holds your business context across months of work is a system. One that starts from zero every morning is a chatbot. The difference is not the model. The difference is the architecture that carries context: what gets remembered, what gets forgotten, what gets written back to a system of record. The memory layer is the thing that compounds and the thing that fails.
Pricing for a deliverable that is not a shift or a seat. If you are buying a workflow that an AI agent runs under human governance, what is the invoice line? Per transaction? Per decision? Per hour saved? A monthly retainer with a token allowance bolted on? Nobody has converged on an answer, and every operator who asks about Virtual Employees asks this question inside the first thirty minutes. The honest answer is not a rate card. It is a unit-of-work contract with a value share at the top end. Vendors whose comp plan is built on per-seat billing cannot answer it without dismantling the comp plan.
Easy to say. Hard to do. That is the work.
Why Most Vendors Cannot Answer
The shape of the company decides the shape of the conversation.
A company whose reps are paid per head opens the meeting with a rate card. A company whose QBRs count seats closes the meeting with a staffing plan. When the operator says what he said, the rep nods, promises to help, and sends a proposal the following week priced per person. The deal does not die in the room. It dies slow, over the next two quarters, as the operator quietly hires somebody who already figured out how to sell Virtual Employees.
This is not a personality problem. It is a structural one. The comp plan, the pricing model, the deck template, and the reporting cadence are all built for an era when the thing being bought was a headcount. Pivoting any one of those pieces without pivoting the others produces a vendor that claims to sell Virtual Employees and a proposal that still charges per seat. Operators notice.
A company whose product is a team can deliver a Virtual Employee. A company whose product is a seat cannot.
What It Takes to Hear It
I walked into four of those rooms with a team pitch and walked out of four of them because I was willing to put the deck away. That is not a personality trait. It is the shape of the company.
A company whose product is a team can deliver a Virtual Employee. A company whose product is a seat cannot. The difference is not marketing. The difference is what the company does when the operator redirects the meeting.
If the deck closes and the conversation accelerates, the company was built for this. If the deck stays open and the conversation slows down, the company was built for something else.
What This Does Not Mean
Humans are not going away. Humans are what build, govern, and retrain the Virtual Employee. A Virtual Employee without the team that owns it drifts, breaks, or quietly starts producing nonsense. A team without Virtual Employees is a cost center competing against companies that have both.
The compounding thesis is intact and arguably stronger. The team compounds. The Virtual Employees the team builds compound on top of the team. Two layers of compounding replace the one layer of the old model. Any operator who has actually tried to run this has felt the second layer.
The Level 2 Trap still applies. A portfolio that buys Virtual Employees without building the team that designs and governs them collapses inside eighteen months, when the first drift hits and nobody knows how to retrain. Ownership of the Virtual Employees has to sit with a team that can defend them, not with a vendor whose contract can be canceled.
Three Questions for the Next Meeting
If you are an operator evaluating who can actually build a Virtual Employee, three questions separate the firms that can from the firms that cannot.
Who inside your team runs Virtual Employees today, and in what workflows? Firms that have not run one internally cannot credibly scope one for you. Ask for specifics. Ask for the four scars. If the answer is abstract, the firm is still selling headcount.
How is the engagement priced when the deliverable is not a shift or a seat? If the answer is a per-seat retainer with a token allowance bolted on, the firm has not made the pivot. If the answer is a unit-of-work contract with a value share, the firm is operating on the same clock you are.
Who governs the Virtual Employee once it is live, and who signs off when it gets something wrong in a regulated workflow? If the answer is a screenshot, the firm has not done this in a regulated back office. If the answer is a named reviewer, a documented approval path, and an audit trail, the firm has.
One of those sets of answers is the engagement you want. The other is the engagement you will quietly regret six months in.
The Reliable Group Position
We run twenty-plus Virtual Employees inside our own business today, across finance, operations, recruiting, and client service. The scars above are our scars. The governance, the memory architecture, the pricing, and the human review layer are not slides. They are the operating model.
When an operator closes the deck and asks for a Virtual Employee, we do not need to pivot a deck. The pivot already happened inside our own company, one workflow at a time, over the last eighteen months.
Four operators. Sixty days. Same sentence in different words. If you have been in a version of that room, or you are about to be, we have been on both sides of the conversation.