Ariel Harmoko

Cofounder and CEO @ArtifactOffcial | Prev prof racing driver champion | Prev ML @ Cambridge and JPMC

[Ariel Harmoko @ariel_harmoko]

Every AI company in accounting is building point solutions. A copilot here. An auto-categorizer there.

@ArtifactOffcial is building the infrastructure layer that makes accounting firms AI-native.

The difference? Point solutions save a few clicks. Infrastructure changes what a firm can do. How many clients they serve. What services they offer. How they grow.

We're already unlocking 12,000+ hours and nearly 7 FTEs of capacity across firms on our platform. 92% reduction in processing time. 13x throughput improvement. Not so they can cut headcount. So they can launch advisory services, take on new clients, and build revenue lines that weren't possible when their teams were buried in manual processing.

And the adoption curve speaks for itself. 55x growth in active end clients. Businesses with real completed financial events on our platform.
Data Ingestion. Bank Reconciliation. Agentic Close Management. Anomaly Detection. Cross platform workflows. All without ripping out your existing tech stack.

We're building the system of intelligence for accounting. If your Accounting Firm is still figuring out where AI fits, we've built internal Agentic tools to assess your readiness and show you exactly where the leverage is.

Reach out or book a demo here: https://t.co/bIZgrhCu68

[Ariel Harmoko @ariel_harmoko]

@GEVS94 @astrange Congrats Gabe!! Thank you for all that you have done. 🙌

[Ariel Harmoko @ariel_harmoko]

We bet our product would be “self-learning” and “self-evolving” before Opus even shipped.

We sell to accounting firms and automate their clients’ books. That means deploying AI into an organization with decades of process, senior people walking out the door, and SOPs no one reads — and doing it for every client on the firm’s roster.

The default playbook is to throw FDEs at it and hand-build prompts and workflows per client.

We wanted an intelligence layer that captures the entire organization and evolves with it. Every human correction — a changed account, tax rate, or counterparty — teaches the system. The next time it sees that pattern, it gets it right unprompted.

That’s the shift from copilot to autopilot. Copilot suggests and waits for a human on every line. Autopilot does the first pass correctly on its own, and the human reviews the exceptions.

So we pulled a small sample of end clients and tracked first-pass accuracy on the hardest work: complex journals, tax, revenue recognition. Same calculation every month.

Every client climbs. Not because someone wrote a better prompt — because the system learned their books. First-pass accuracy across the group went from 71% to 95% in three months.

The shift from copilot to autopilot is real. I

If you’re interested in deploying agents that performs real accountancy work, come join us at @ArtifactOffcial

[Garry Tan @garrytan]

This is the actual bottleneck. The models are smart enough already. What is missing is the company-specific context locked in senior people heads. Whoever cracks knowledge extraction at the company level unlocks the rest.

As you work on this, please consider using GBrain as your OSS retrieval layer

https://t.co/0F5uDQzPHu

[Ariel Harmoko @ariel_harmoko]

Everything, all at once.

[Ariel Harmoko @ariel_harmoko]

The defensible data is not the data you import. It’s the data your product uniquely causes to exist.

This reframes the entire AI-native SoR debate. The old playbook was: become the place where the work is logged. The new one is: become the place where the work actually happens, and let the record be a byproduct.

If your product is just a better front-end for the same data everyone else has, you don’t have a moat. You have a feature.

What a banger @seema_amble

[Seema Amble @seema_amble]

https://t.co/o3nHSGDlUb

[Ariel Harmoko @ariel_harmoko]

Our London team sits in the same building!!! We reached millions in ARR in <1 year. Come chat to @ArtifactOffcial if you want to be in the frontier of Accounting AI and share space with Jeff Bezos.

[etn. @etnshow]

BREAKING: Jeff Bezos's AI lab 'Project Prometheus' eyes London, Kings Cross for new office.

Project Prometheus is discussing signing a lease at the Jellicoe Building for three floors spanning 38,000 sq ft.

[Ariel Harmoko @ariel_harmoko]

Back at @Cambridge_Uni where we (@ArtifactOffcial ) hosted a builder showcase. The same grounds where Isaac Newton discovered gravity. Unreal founders and talent density. 10 Trinity college students came crawling to work for us.

Always impressed by the remarkable talent.

[Ariel Harmoko @ariel_harmoko]

This is what it’s all about💙

[CPA Practice Advisor @cpapracadvisor]

ArtifactAI has launched Omni, a new AI-powered platform designed to target the connective tissue sitting between tools and manual accounting workflows- https://t.co/Q4DsD7plak

[Accounting Today @AccountingToday]

Artifact @ArtifactOffcial, an AI-focused accounting process automation solutions provider, announced the launch of its new Omni agentic workflow solution meant to help firms orchestrate complex, multisystem and cross-platform work. https://t.co/3vkrH0tuB9 #AI #accounting

[Ariel Harmoko @ariel_harmoko]

@Tocelot @ArtifactOffcial Thanks Jon!!! Missing the SR5 era

[Ariel Harmoko @ariel_harmoko]

The accounting profession has a double bind that nobody talks about. Today, @ArtifactOffcial is launching Omni, your firm's new agentic orchestration layer, in response.

The prevailing narrative is that "AI handles routine tasks, accountants go from doers to reviewers". But this breaks how accountants actually develop judgment.

You learn to reconcile correctly by reconciling incorrectly and being corrected. Remove the doing and the learning mechanism breaks. If junior accountants are just reviewing AI outputs, they won't develop professional expertise.

But the traditional path was already failing. 300,000 accountants have left. The doer work is gruelling. The apprenticeship demands years of tedious production before the payoff arrives.

One path breaks the traditional learning path. The other breaks the people. Neither is sustainable. This is the double bind. It's what Omni undoes by presenting a way of working that is different enough to sustain the learning, but compelling enough to retain the people.

Omni connects across 1000+ platforms and lets accountants describe multi-step, cross-platform workflows in natural language. Build a workflow for one construction client on Sage, templatize it, deploy it across every construction client with client-specific logic applied automatically. The people closest to the work become the ones building the systems.

Omni sits on top of Arti, our intelligence layer, already live across Top 250 firms in the US and Top 50 firms in the UK. 99% reconciliation accuracy. 92% reduction in processing time. 13x throughput improvement. 12,000+ hours of capacity unlocked and counting.

But the numbers aren't the point. The architecture is. Arti learns continuously from every interaction. When a senior corrects a categorization, the correction and its reasoning become institutional knowledge. When anyone overrides an automated match, the override and the why get encoded into the firm's knowledge base.

Together, they capture what we call "wet code," the unwritten business rules, client-specific logic, and contextual knowledge that live in senior accountants' heads and nowhere else. With 75% of CPAs eligible to retire within the decade, that knowledge is walking out the door.

The goal is not to turn accountants into reviewers. It's to make them a different kind of doer.

Join the Omni waitlist and read our manifesto at the link in the comments.

[Ariel Harmoko @ariel_harmoko]

https://t.co/pVBx1S1xhJ

[Ariel Harmoko @ariel_harmoko]

🤘🤘🤘🤘

[Ariel Harmoko @ariel_harmoko]

Hmu!

[Ben Lang @benln]

Startups best known for hiring former founders:

Cursor, Notion, Rippling, Applied Intuition, Supabase, Ramp, Exa, Chroma, Corgi, Lovable, Mintlify, Browserbase, Instawork

Who else? Refreshing my shortlist.

[Ariel Harmoko @ariel_harmoko]

@PpetersonPaul Which team do you support? 😄

[Ariel Harmoko @ariel_harmoko]

.@MeekMill after discovering vibe coding and GLP-1

[Ariel Harmoko @ariel_harmoko]

Distribution is the moat

[John Coogan @johncoogan]

TBPN has been acquired by OpenAI!

The show is staying the same and we’ll continue to go live at 11am pacific every weekday.

This is a full circle moment for me as I’ve worked with @sama for well over a decade. He funded my first company in 2013. Then helped us fix a serious logjam during a critical funding round a few years later. When I took my second company through YC, he was president at the time, and then when I joined Founders Fund, the first deal I saw in motion was the post-ChatGPT round in late 2022. And as we started growing TBPN last year, he was the very first lab lead to join the show.

Thank you to everyone that has been a part of TBPN until now. The last year has been the most fun and rewarding part of my career and we’re excited to have more resources than ever going forward.

[Ariel Harmoko @ariel_harmoko]

@PpetersonPaul We’ve accidentally created fragmented “system of knowledge” through the software we’ve been feeding our expertise for decades. The opportunity now lies in making sense of those chronological enterprise/human patterns to unfold value.

[Ariel Harmoko @ariel_harmoko]

If you’re not profit-maxxing this quarter, and just getting corporatemogged by the competition, stop slackmaxxing or you’re getting frame-mogged by high-tier chads