How do you differentiate? | enterprise accounting AI edition
Assorted notes from investing as a VC in the first-half of 2026
What I’m seeing
In 2024, saying you were the AI-native version of <Insert Incumbent> and having a slick UI was enough to at least have a conversation with a customer.
In 2025, as LLMs have become much better and AI is undoubtedly “hot” (more startups), every AI-native version of <Insert Incumbent> has 10+ competitors.
In 2026, some customers I talk to are trying to vibe-code their own versions of products as token spend is now normal outside of engineering. In other words: your competition is your customer. In the SaaS era, this was rarely credible outside regulated verticals and the technology sector; now it’s increasingly real across many verticals.
So what’s left to differentiate yourself on?*,**
The startups I have backed in accounting AI this year have the following enterprise ‘moats’:
Compound startup: enterprise product satisfies needs of multiple departments simultaneously. Ripping it out would greatly impact the other department, and vice versa.
Licensed, regulated entities: The code/software is a means to deliver X, of which it can only be done by licensed or regulated entities. You need this type of entity to be involved, whether it is a professional services firm or a new AI-native Service-as-Software firm.
Last mile: in a category I’m invested in, many new startups are trying to solve it using agents/software (undifferentiated). The company I backed recognizes that the problem is not in the software/agents, but at the last-mile (payments). There are also other last mile areas such as integrations to local or niche providers.
Other enterprise signals
Teams really matter (actually)
You are trying to convince an enterprise CFO or CIO, or their audit committee, about your new AI-native accounting product. The VC cliché is that teams are all that matter when backing a startup. This really rings true in the enterprise AI world. These customers are also buying trust in the (credentialed, pedigreed) humans behind the vendor when the trust in the product is new to them.
Investor brand matters more to founders (and probably customers)
Pre-AI, the name-brand firms were typically multi-stage generalist firms. They would mostly invest at Series A. If they invested at Seed, it was an option check. If they didn’t lead or go super-pro rata at Series A that would indicate signaling risk to the market and potentially put the Series A at risk for the startup.
Post AI, this is no longer the case. Founders and investors care far less about signaling risk. Many multi-stage firms still treat Seed as an option check, but in a world where valuations have expanded significantly, this “option check” it is also used to buy up ownership early. More founders seem to care about the positive signal of an investor’s brand as they’re competing for customer/investor attention far earlier in the company lifecycle than in the pre-AI world. In many cases, I’m seeing startups take lead checks from experienced generalist firms (whether Seed or multi-stage) and supplementing it with strategic checks from firms like my own with domain-specific expertise, networks, and signal.
So how do you differentiate?
++++
*I’ve left out an obvious one (better sales and marketing) as startups benefit from that in the pre/post AI world, albeit there are different techniques to make your sales and marketing stand out in the post AI world (perhaps a post for another day).
**Auditability and reproducibility of AI outputs are still an area that has not yet quite been elegantly solved IMO (see here and here)