Content
Topic
Building
Building DualEntry

SaaS no more → Agents as a Service

Benedict Dohmen
Benedict Dohmen
Co-founder
Benedict Dohmen
Benedict Dohmen
Co-founder
Benedict is the co-founder of DualEntry. He previously co-founded Benitago, a digital consumer products group that raised $380 million in funding, grew to over 300 team members, and achieved $100M in annual revenue before its acquisition in 2024. Benedict has been featured in The Tim Ferriss Show, Forbes, The Wall Street Journal, and more. He graduated with a BA from the University of Cambridge and was raised in Germany.
Learn about our editorial policies.
Last updated: 
November 3, 2026

Business software has been sold much the same way for 25 years: per seat. You pay for a login, and a person does the work inside it. 

‍

ERP is the clearest example. Finance teams have been running on cloud ERPs for two decades. And yet the close still takes 15+ days, implementations cost $100k+, and nothing moves until an accountant keys it in and ties it out.

‍

Santi and I ran a $100M company on a legacy ERP, and it was blank-screen software. It stored whatever our team typed in, and that was about it.

‍

NetSuite and Sage are the best version of that model, with billions in revenue and pricing structures built around user licenses. It's a great business, and I don't expect them to walk away from it. 

‍

When you earn per seat, the last thing you want to ship is software that needs fewer of them. So the temptation for incumbents is to bolt a chatbot onto the existing product and leave the business model alone.

‍

Agents change that, and we call the new model AaaS: Agents as a Service.

‍

With SaaS, you bought a tool for your AP clerk. With AaaS, you buy accounts payable itself. The bill gets read, coded, and put in front of someone for approval. That's what you pay for, not the login.

From the software budget to the labor budget

SaaS was only ever competing for the software budget. Agents compete for the labor budget, and in finance, that's the much bigger line. A single accountant can cost $120k+ a year, often more than the ERP they work in.

‍

Customer support got there first. For example, Intercom lists its traditional software at $29 per human seat per month, while its Fin AI agent starts at $0.99 per outcome. You can see the whole shift in those two numbers.

‍

I'd argue finance is an even better fit than support. The work is structured and rule-bound, and it gets measured to the penny: days to close, invoices processed, accounts that tie.

‍

We made our first move on pricing early. Every DualEntry plan comes with unlimited users, and the broader idea is that what customers pay should reflect the shape of their business, not simply how many people log in.

Why now?

So why is the business model changing now?

‍

The ERP industry has been through two major platform shifts: from on-premise to cloud in the 2000s, and now from cloud to AI.

‍

The first changed the business model as much as the technology, turning perpetual licenses into subscriptions and giving rise to SaaS. The second is doing the same, with subscriptions giving way to work. The cloud changed how software got delivered, and AI is changing what it does once it's there.

‍

Two things make the timing real. The models can finally carry a task end to end, so an agent can match deposits or chase a late payer without someone prompting it through every step.

‍

And the accounting talent pipeline is under pressure. CPA exam candidates fell from roughly 50,000 in 2010 to 30,000 in 2022, while experienced accountants continue to retire. Finance teams aren't going to hire their way through the next decade if the talent isn't there.

Selling the work raises the bar

A SaaS vendor could ship a blank screen and leave accuracy to your team, and most of them did. If you're selling the work itself, you own the accuracy.

‍

In accounting, that means two things have to be true:

‍

1. The agent has to work from the whole ledger. An agent bolted on through an API only sees what the API hands it. An agent that lives inside the system of record sees the transaction, its history, the contract behind it, and how the same vendor was booked last period. That's the difference between proposing the right entry and proposing a plausible one.
‍

2. The controls have to live in the data model, not the screen. Agents don't click through a UI, so approvals, period locks, and the audit trail have to sit underneath it. In DualEntry, every agent proposes transactions under its own identity, approval is the posting event, and nothing reaches the ledger without a human signing off.

‍

We call this CTRL+AI: control first, then AI, with a citation behind every entry an agent proposes.

‍

It's also why you cannot “vibe-code” an ERP and why an AI lab can't just build one. The agents have to live where the work happens, and that's the system of record.

What changes for accountants?

Is it working? So far, the numbers say yes.

‍

Slash runs $600M of ARR on a finance team of 2. Graphite took its close from 20 days to 1. Customers tell us that without DualEntry they'd have hired one or two more accountants, and those are expensive hires in a market where accountants keep getting harder to find.

‍

So what happens to the accountants?

‍

Their job moves up a level. The accounting team of two is a preparer and an approver with a fleet of agents underneath them, where the agents do the work and the humans supply the judgment and the signature.

‍

Imagine an AP clerk who works every hour of the month, never quits in close week, and remembers how you coded every vendor last period.

‍

SaaS gave every accountant a login. AaaS gives every accountant a team.

Want to build the future of ERP?

Join the team building a modern ERP for companies that have outgrown the old way of doing finance.

Join us