Five AI workflows running inside a small M&A brokerage

The AI workflows earning their place inside a small M&A brokerage are email drafting, morning meeting briefs, a valuation skill that gets roughly 80% of the way, video editing with captions, and AI-search visibility that produces leads. All five run at The Magnolia Firm, the Austin sell-side brokerage Christine McDannell owns, on subscriptions she puts at about $100 a month against $40,000 a month of labour.

What follows is what one operator reports running, named by product, with the setup cost attached. It is reporting on her stack, not a recommendation to buy any of it.

Which AI workflow pays back first?

Outbound email, on her account, but not immediately. The agent checks her inbox every 30 minutes and drafts replies rather than sending them.

"It's doing all of our email, but again it's just drafting. So it checks every 30 minutes and it drafts email, and it's nailing it."

The number worth planning around is the setup, not the output:

"It took a good four to five days. It's like training a brand new employee. It was super frustrating and I almost gave up. And then it was just dialed to the point where me and Tyler were like: did you write that email? Did you? Oh wait, Claude did."

Four to five days of frustration is the honest entry price for this category, and it is the part most vendor material leaves out.

What does an AI meeting brief do for a broker?

It arrives before the seller does. The same agent reads her calendar every morning and researches whoever booked.

"It's also checking my calendar every morning ... when they book, it gives me this amazing brief on the person. It's combed everything. It tells me all about the person, all about the company."

For a sell-side broker whose week is spent meeting prospective sellers, walking in already briefed on the person and the company is a structural advantage rather than a convenience.

Can an AI valuation skill be trusted?

Only with the expertise already in the prompt. McDannell's firm runs a valuation skill built on its own industry knowledge, current market conditions and past deals.

"It'll get us about 80% of the way. And then the next 20 is me still fixing the valuation and jazzing it up a little bit and making sure it covered everything."

Her conclusion is the boundary condition for the whole stack: "It does a good job if you have the right input. And you have to know the industry to know the inputs in the prompts."

Which tools does she name, and for what?

Four products and one piece of hardware come up on the record:

• Claude — the email drafting agent, the morning calendar brief, the valuation skill, and the marketing work, run through one main agent she calls Anna.
• Claude Cowork — phone footage in a Google Drive folder turned into individual reels, a sizzle reel and branded captions from a single request. Output roughly 80% finished, then re-prompted.
• Canva — split-tested against Claude Cowork on caption transcription and, in her test, the one that made mistakes.
• YouTube — how she learned, following one creator, Brock, who teaches Claude to non-technical users. "Coffee and YouTube is my go-to."
• A separate Mac mini — dedicated hardware bought for the firm's work, with two separate accounts.

Tools named by an operator on the record. Inclusion is reporting, not endorsement.

How does an M&A firm handle client financials safely?

By stripping the identifiers before anything goes in, which she believes is not standard practice among advisors.

"What we're doing is needing to strip out — it's sensitive data. And again, I know no advisor's doing this, they're just throwing it all in there ... So we strip the company name off when we're throwing in their financials in the system."

Her stated reasoning is training data and searchability. The dedicated Mac mini and the split accounts belong to the same habit.

Does any of this bring in deals?

She says it does, and by an unexpected route. After a year of deliberate work on GEO and AEO — the AI-search equivalent of SEO — her last five leads came through LLMs, and the firm asks every inbound where they came from.

"The last five leads — amazing leads — came from the LLMs ... we always ask. So it's like a GPT, or right now it's Claude."

One arrived through a buyer's own agent: "This last guy put Claude in an agentic agent. And so he has his agent that's now searching." She has since built a personal-name site around keyword-loaded blog content, on the reasoning that it is "exactly what the LLMs are going to crawl."

What should a small firm try first?

The order McDannell's own results imply is drafting, then briefing, then anything client-facing — and the discipline that matters more than the tool choice is fewer agents doing more. She keeps one main agent rather than the 40 a friend at a tiny company built, and she blocks out four to five days for training it. Which task in your week would you notice most if it were drafted for you by tomorrow morning?

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.