Invest1 publisher3 min readPublished
Blackbird funds a finance agent whose main feature is the paper trail
Sterling raised NZ$3.8m to run invoice processing and month-end close. The load-bearing design choice is not the model. It is the log and the escalation rule.
The Investor · Invest desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction

What happened
- New Zealand AI startup Sterling raised NZ$3.8 million (US$2.2 million) in a funding round led by trans-Tasman venture capital firm Blackbird.
- Sterling was founded in 2025 by Nik Wakelin and Ludwig Wendzich.
- Sterling develops what it calls an 'AI autopilot' for finance teams and is expanding its AI-powered automation platform for finance teams.
- Sterling said every action taken by its software is logged, while ambiguous tasks are escalated to a human.
- Sterling automates workflows including invoice processing, bank-feed reconciliation and month-end routines.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
Sterling, a New Zealand startup founded in 2025 by Nik Wakelin and Ludwig Wendzich, has raised NZ$3.8 million (about US$2.2 million) in a round led by trans-Tasman firm Blackbird [1][2][3]. The product detail that matters more than the funding: Sterling says every action its software takes is logged, and ambiguous tasks are escalated to a human [4].
That is the shape automation has to take to clear a controller. Finance work is not judged on average accuracy; it is judged on whether an auditor can reconstruct why a specific entry was made. An agent that posts a journal correctly 98 percent of the time and cannot explain any of it is worse than useless in a close, because the two percent has to be found by hand. Logging every action and refusing to guess when the input is unclear converts a black box into something a reviewer can sign.
The workflows Sterling names are the ones where this bites hardest: invoice processing, bank-feed reconciliation and month-end routines [5]. It integrates with the systems finance teams already run, including enterprise resource planning software and accounts inboxes, rather than replacing them [6]. Sterling says it is SOC 2 Type II certified [7]. Early customers include finance teams at Manukora, Storypark and Echelon, and development has been supported by a New Zealand government research and development grant [8][9].
The pitch is explicitly against the chatbot layer. "Finance teams don't want another chatbot to talk to about their work. They want something that does the work," Wakelin said [10]. Blackbird principal James Palmer framed the gap as reliability rather than capability, arguing finance teams are seeing plenty of AI demonstrations aimed at single workflows but lack tools that hold up across several: "What's missing is a product that offers reliability and scales beyond the first use," Palmer said [11][12]. Wendzich, for his part, expects accounting roles to move "from operational to strategic" [13].
The founders' backgrounds are operator-shaped rather than research-shaped. Wakelin co-founded time-tracking company MinuteDock and developer documentation platform Gelato.io, and held a senior engineering role at Deliveroo [14]. Wendzich spent close to a decade in product and design at Vend and then Lightspeed after its acquisition of the New Zealand retail software company, and was previously a senior front-end engineer at Apple [15]. The round also drew Rowan Simpson and Eliot Crowther plus alumni of Trade Me, Pushpay, Xero and Vend [16].
Note the scale. NZ$3.8 million is a small cheque against the ambition of selling into large enterprises, and the round arrived roughly a year after the company was founded [17]. The money is earmarked for hiring and for larger enterprise customers in New Zealand, Australia and the United States [18].
What to watch: Wakelin has set the bar himself, saying the next 12 months are about demonstrating the software can reliably do finance work while keeping every action logged and explainable [19]. The specific things to test are the escalation rate over time, whether it falls as the system learns or quietly rises as edge cases accumulate, and whether the log survives contact with an actual external auditor rather than an internal reviewer. Also worth watching is whether SOC 2 Type II plus a full action log is enough to get an agent write access to a general ledger, or whether enterprise buyers keep it in suggest-only mode. That distinction decides whether this is automation or a faster inbox.