How a Fractional CFO Firm Built an AI Recruiter That Hired a Real Employee

Madeleine Garcia has spent the last few years at the intersection of finance and technology. First at Embark, where she helped roll out AI tools across the organization. Now at Ignition Consulting, a fractional CFO firm serving startups, where she's Head of Client Delivery and the person responsible for making AI actually work.

We sat down with Mads to talk about what she's learned, what's working, and what most companies get wrong.

The "Tyler Jones" Experiment

The most striking example from our conversation: Mads built an AI recruiter named Tyler Jones.

Tyler has a Gmail account, a professional background, and a personality. Tyler screens candidates, scores them against a rubric, drafts rejection emails, schedules interviews, and briefs the hiring team. The candidate who was eventually hired didn't know Tyler was AI until after they accepted the offer.

"I built Tyler exactly like I would hire a recruiter," Mads explained. "Here's your email, here's how I want you to organize it, here's the labeling system. A lot of it was me and Tyler brainstorming how we were going to manage multiple roles."

The ROI was immediate. No external recruiting firm. No placement fees. Just a small team moving faster than they could have otherwise.

Hot Takes: Why Most AI Rollouts Fail

Mads has seen AI adoption from both sides: the large enterprise (Embark) and the small, fast-moving startup (Ignition). Her take on why most companies struggle is direct.

"You can't half-ass it," she said. "There's so many tools, things change so quickly. You have to be ready to designate this as a full-time job for somebody who's on top of it."

The failure pattern she sees repeatedly: companies buy a tool, tell employees to figure it out, and wonder why adoption stalls. The power users thrive. Everyone else tries a few prompts, hits a wall, and goes back to the old way of working.

Her recommendation: don't expect individual employees to figure out AI on their own. Build workflows at the system level, then let people experience them. Once they see what's possible, they start thinking about what else they could automate.

"Companies will be more successful in adoption if there is a specific team to implement,” she said, “versus individual people trying to take it on themselves."

Sachin pushed back on this a bit. His view: it has to be both. Yes, you need dedicated people building systematic workflows. But you also can't wait until you have the perfect team in place. Boards are pushing executives to move now, not after they've hired the right person. The challenge is doing both simultaneously: getting people enabled on something while you're still figuring out the long-term approach.

"That's the tricky part," Sachin said. "Most companies would agree with you, but they're dealing with this catch-22. They have to get everybody trained and up and running on something, but they also know that if they don't have the right people, it's not going to work well."

The companies that succeed are the ones willing to make the investment upfront, whether that's an internal hire or outside help. Half-measures lead to the failed rollouts everyone's seen.

What's Actually Working, And The AI Tools She Uses

At Ignition, the AI strategy is practical. They use Claude for team collaboration. They've experimented with Uru Platforms for building personas like Tyler. They use Notebook LM to help CFOs digest client information in whatever format works best for them: audio podcasts, PowerPoints, or chat.

But the real insight is how they think about deployment. Every new CFO goes through an onboarding process that includes a "bingo card" of AI tools to try. The goal is to meet people where they are and figure out what resonates.

"Everybody's on a different level of AI adoption," Mads said. "Using and just introducing tools to people and seeing how they like to use it, what they gravitate towards, is still helping me discover their levels of understanding and their level of AI-forward readiness."

Some of the most surprising contributors aren't the heavy users. They're the old-school CFOs who don't use the tools themselves but keep coming to Mads with ideas. "I'm sure you could build something like that," they'll say. And usually, she can.

The Advice

For companies trying to figure out their AI strategy, Mads offered a clear framework:

1. Designate ownership. Either build an internal team or bring in outside help. This can't be a side project for someone with a full-time job.
   
   
2. Start with workflows, not tools. Find the processes that happen repeatedly across your organization and automate those first. Individual productivity gains are nice, but systematic improvements scale.
   
   
3. Dogfood everything. Ignition tries every tool on their own operations before recommending it to clients. They know what works because they've used it.
   
   
4. Accept that this is ongoing. The AI market changes constantly. What works today might be obsolete in six months. You need someone paying attention.
   
   

"This stuff doesn't sleep," Mads said. "You have to have designated research and know-how and a plan forward."