Sofia Fatakhova (Chime): Building the finance function to go public

As Vice President of Finance at Chime, Sofia Fatakhova has had a front-row seat to one of the more demanding growth stories in recent memory. In four years, the company went from $700 million to $2.4 billion in revenue and launched into the public markets. For the finance function, that kind of pace forces a complete rethink of how the team is structured and what it needs to be capable of.

The work Sofia has done at Chime reflects all of that. She rebuilt a forecasting process that had previously sprawled across seven separate models and took two weeks to complete, collapsing scenario planning from three to four days down into seconds. Today, she is leading her team through a considered bet on what AI can (and should) do inside a public company's finance function.

Too often, broken architecture obscures hard work

At Chime, Sofia inherited a finance department that was stretched thin, with team members accustomed to working nights and weekends. The reforecast cycle consumed so much of the month that, by the time the team finished presenting to leadership, it was already time to start again.

Sofia knew what “good” looked like. During her time at Amazon and Salesforce, she had turned scenarios around in minutes. At Chime, the same task took three to four days. Because data lived in different places and was maintained by different people, there was no reliable way to know which version was current. The team could spend a week just pulling actuals and populating the models by hand before any real analysis even began.

The issue was that Chime’s forecasting process was built across seven separate models, each owned by a different person. When one finished their section, they handed it to the next. It was sequential, manual, and prone to error at every transfer. 

Rather than try to fix everything at once, Sofia set a specific and demanding target: take the two-week reforecast down to two days, with an eventual goal of half a day once the company went public. That clarity about what “done” looked like made everything else easier to prioritize.

Replacing the baton-pass model

Fixing the reforecast process required a shared environment where data was standardized and visible to everyone, and where models could be built simultaneously rather than handed off one by one. That meant getting out of spreadsheets entirely. 

Sofia brought in Pigment to do just that.

The shift happened just in time. The team had given themselves six months to get fully operational before the IPO, with everyone trained and all the models migrated. It was a tight window, and going public immediately tested everything they had built. Analyst models cycled through multiple versions, inputs kept changing, and the accuracy bar moved to an explicit target of 90% to 99%. 

At the same time, investors and analysts were forming their own views of Chime from the outside, and the finance team had to understand and actively respond to that perspective. Having a single environment where all of it could happen simultaneously was what made the timeline achievable.

The IPO also clarified how the finance team should allocate its energy. Sofia landed on an 80/20 model: 80% on strategic, insight-driven work with the business, and 20% on the operational machinery. Getting the infrastructure right was what made that ratio possible.

AI works best when it's nestled inside solid infrastructure

With the IPO behind them and the infrastructure solid, Sofia's attention shifted to what the finance function could become next. The efficiency gains were real, but the bigger question was how to use the time and capacity they had created. That's where AI entered the picture.

Sofia built a dedicated team inside the finance function that is charged with scaling AI across every part of team operations. Every one of her direct reports is writing a vision doc for how they will incorporate AI into their work and goals. She calls the current phase "AI tourism," a period of broad experimentation, with the explicit goal of moving toward something governed and scalable. 

One of the clearest examples of what that looks like in practice is how the team now handles vendor invoices.

Chime works with a large number of vendors, and the invoices they receive arrive in inconsistent formats that are difficult to analyze at scale. AI now ingests and normalizes them, surfaces trends, and flags places where the company may be paying more than expected. That output then flows into Pigment, where the actual modeling happens. The human work concentrates at the end, on the questions that matter: what drove this, and what should we do about it?

As a public company, Chime operates under strict financial controls and reporting requirements, which means the bar for how AI gets built into the function is high. Sofia is deliberate about that. The goal is to build something documented, transferable, and tied to real outcomes rather than excitement about the tools themselves.

Speed requires letting go of your attachment to completeness

In a way, the AI transformation Sofia is leading is an extension of the same lesson she has been learning since she joined Chime. A finance career builds strong habits around rigor and precision, and for good reason. But at the pace Chime has moved, waiting for complete information has never really been an option. The discipline has always been to act on the best available picture and trust the infrastructure to fill in the gaps.

That's ultimately what AI offers the finance function: not perfection, but speed and clarity at a scale no team could match on their own. Sofia has spent four years building the systems, the processes, and the culture at Chime to make that possible. The work of the next phase will test how far it can go.