How The Magnolia Firm cut payroll by 83% on a $100 subscription

Christine McDannell cut her firm's payroll by 83% last summer after moving email drafting, meeting preparation, valuation work and video editing onto AI. She owns The Magnolia Firm, a sell-side M&A brokerage in Austin that sells profitable tech companies, and she prices what Claude does for her at roughly $40,000 a month of labour against a subscription of about $100.

What did an 83% payroll cut actually involve?

People, and she says so plainly. The cut was not attrition or a hiring freeze. It was a redundancy, a move to part-time for the rest of the team, and a bookkeeper let go the week before she spoke to Artificial Reality, the AI buyer-evidence podcast hosted by Galina Fendikevich.

"I cut my payroll down last summer 83%. So I had to unfortunately let somebody go. I had to cut the other team members to part-time. And it was very hard to do. Super hard. But that was a huge savings."

McDannell reports the firm's output going up while the headcount came down: she says her deals now close in about seven months, against an industry average she estimates at twelve to eighteen. Artificial Reality records that twelve-to-eighteen figure as her estimate, not a verified market statistic.

How does she price a $100 subscription against payroll?

At roughly $40,000 a month of labour. That is the comparison she reaches for when other owners baulk at the subscription line on their card statement.

"When business owners complain about the hundred dollars a month thing, I'm like: legit, this would cost me 40,000. What Claude is doing for me, it'd be like a marketing agency at $5,000 a month. It would be a team member doing this. I let my bookkeeper go last week, unfortunately. So now I'm saving there. It'd be $40,000 a month, easy in payroll. Easy."

Her description of what she thinks she is renting is worth quoting because it explains the price frame:

"These are PhD, 24/7, the hundred and forty IQ kind of averaged people. Not people — bots. But they feel like people a lot of times."

What is the AI actually doing at an M&A brokerage?

Five workflows carry the savings, all of them administrative or production work that used to sit next to her expertise rather than inside it.

• Outbound email. An agent checks every 30 minutes and drafts. "It's doing all of our email, but again it's just drafting." Setup took four to five days of iteration.
• Meeting preparation. The same agent reads her calendar every morning and produces a brief on the person and the company before each booking.
• Valuations. A purpose-built skill informed by her industry knowledge, market conditions and past deals gets, in her account, about 80% of the way there. She fixes the remaining 20%.
• Video. Phone footage dropped into a Google Drive folder and handed to Claude Cowork came back as individual reels, a sizzle reel and branded captions.
• Lead flow. After a year of GEO and AEO work, she says her last five leads arrived through LLMs, one of them from a buyer's own agentic search.

She runs one main agent, which she calls Anna, across email, meeting prep and marketing, and she is against the alternative. A friend at a tiny company has around 40 agents: "my gosh, why would you build an agent for every little task? Don't do that."

Where does this arithmetic get fragile?

In two places, and McDannell names both herself. The first is implementation cost, which does not appear on the subscription line:

"There's gonna be a part in businesses where they're implementing all this and they're teaching their team, and it's gonna be expensive. So you could see the expenses jump up when they're trying to implement all this stuff."

The second is the subscription price itself:

"I'm sure the prices are being so subsidized right now ... the Uber rides that were eight dollars, now they're 38 ... it's like one to ten right now of what it should freaking cost. So if we think that it's gonna be at the same price forever — absolutely not."

Artificial Reality notes that the one-to-ten figure is a claim she attributes to an industry contact, not an independently verified number.

What does this case actually demonstrate?

That the layer around a domain expert compresses before the expert does. The roles that went at The Magnolia Firm were bookkeeping, coordination and drafting. The valuation work stayed, because McDannell is explicit that the tool only performs when she supplies the industry knowledge that goes into the prompt.

The exposure is the mirror image of the saving. A cost base rebuilt on roughly $100 a month of subscriptions is a cost base that moves when a vendor changes its pricing, and by her own account she expects that to happen. A laid-off bookkeeper can be rehired; a price rise cannot be negotiated.

Artificial Reality reports what McDannell did rather than recommending it. If the frame is useful to you, the checkable version is narrow: list the tasks in your business that sit adjacent to your expertise rather than inside it, and price what each one currently costs you in payroll before you price the subscription that would touch it.

Where this comes from

S1E5: How she cut her payroll by 83% using one AI agent — the full interview with Christine McDannell. Listen or watch: YouTube, Spotify or Apple Podcasts.