"For 20 years, enterprises have rented software that forces them to operate around it. With AI, that ends." — Adam Jafer, CEO of Pit.
When Pit — an AI startup from the founding team behind Swedish e-scooter platform Voi — closed a $16M seed round led by a16z, this declaration started getting real. For 30 years, enterprises bent to SAP, Salesforce, and Oracle. Is that relationship finally reversing?
30 Years of Software Calling the Shots
HR teams designed onboarding the way SAP HR modules supported it. Marketing teams defined sales stages to fit Salesforce pipelines. Finance teams built approval criteria around their ERP invoice flow. "Adopting software" almost always meant "reshaping your work to fit the software" — not the other way around.
RPA tried to fix this, but bots broke whenever the UI changed, and humans had to intervene for every exception. Traditional RPA achieved only 20–30% of the automation potential enterprises expected — AI agent approaches extend that to 60–80%.
"The shift from RPA to agentic AI is not incremental. It is architectural."
— Lasting Dynamics, 2026 RPA vs AI Agents Analysis
Pit is making waves because they're proving this structural shift works in actual enterprise back-office processes — not in a demo. The Voi founding team's DNA (real operations scale-up experience) combined with engineers from iZettle and Klarna hit this moment at exactly the right time.
Studio Learns, Cloud Governs
Pit has two layers. The first, Pit Studio, is where enterprise employees directly teach AI their workflows — which data flows where, what conditions trigger which decisions. Studio learns all of it, then converts those processes into custom automation software.
The second, Pit Cloud, is the infrastructure that makes that software work safely in real enterprise environments: ISO 27001 certification, tenant isolation, SSO, role-based access control, and full audit trails — all built in from day one. "Can we trace the reasoning behind this automated decision later?" That is the first question every CIO asks. Pit answers it through design, not afterthought.
These are not pilot numbers. They come from a live deployment replacing legacy contract and invoice validation systems at a major European industrial company. Agentic AI achieves straight-through processing rates of 85–92%, versus 65–75% for traditional RPA.
| Traditional SaaS/RPA | Pit (AI-Native Custom Software) | |
|---|---|---|
| Adaptation direction | Company adapts to software | Software adapts to company |
| Automation coverage | 20–30% | 60–80% |
| Straight-through rate | 65–75% (RPA) | 85–92% |
| Handling changes | Rebuild bots | Model updates |
| Governance | Requires extra configuration | ISO 27001 built-in |
Not Doing Things Was the Whole Strategy
Pit's most distinctive move was deciding what NOT to build. No customer-facing AI. No chatbots. No conversational AI. Back-office only. Here is why that constraint was strategically brilliant:
- Risk elimination: Cuts out the biggest hazards of customer-facing AI — wrong answers and brand damage
- Measurable ROI: Invoice acceptance rates, contract validation speed, campaign execution time — processes that prove themselves with numbers
- Trust in regulated industries: Lower adoption barriers in telecom, healthcare, logistics, and industrials where compliance matters most
a16z partner Alex Rampell nailed it: "Every AI company is selling speed. Pit is selling speed that holds up for years, that's secure, governed and built to last."
The EU AI sovereignty angle reinforces this even further. Pit is fully vendor-agnostic — not locked into any AI or cloud provider. Jafer noted that "EU models running on EU compute is top of mind for almost every CIO we're meeting." In a world of GDPR and the EU AI Act, that positioning is both a competitive moat and a sales accelerator.
Finding Your First Back-Office Automation
The same sequencing Pit's embedded solution engineers use with clients applies when you're starting internally. According to Hypatos, accounts payable (AP) processing delivers the highest ROI and shortest time-to-value for a first automation.
- List your repetitive, predictable internal processes
Build a list of tasks handled the same way every week or month: invoice validation, contract processing, report compilation, campaign execution requests, onboarding paperwork. - Prioritize processes needing unstructured-to-structured conversion
AI agents excel at converting PDFs, emails, and spreadsheets into structured outputs. Start with processes that follow this pattern. - Document your exception cases
Those "a human has to decide this" moments usually collapse into a few repeatable patterns on closer inspection. Cataloging them upfront dramatically expands automatable scope. - Define audit and compliance requirements upfront
Whatever you automate, someone will ask "can we trace this decision later?" Build audit logging in from the start — not as an afterthought. - Start with one process, one team
Establish the pattern small, then expand. Pit's landmark results — 85%, 99%, 10,000 hours — all came from single-process deployments first.
Want to Go Deeper?
Voi founders' new AI startup Pit has become the latest rising star out of Stockholm TechCrunch original with founding story and full product details techcrunch.com
A16z leads $16m seed round in Voi cofounder's new AI startup Pit Sifted in-depth coverage: investment context and Stockholm AI ecosystem sifted.eu
AI product team startup Pit raises $16M from a16z SiliconANGLE technical breakdown of Pit Studio and Pit Cloud infrastructure specs siliconangle.com
From RPA to AI Agents: Why 2026 Is the Year Enterprise Automation Gets Real Source for 20–30% vs 60–80% automation coverage comparison data lastingdynamics.com
Enterprise guide to agentic AI for back-office automation Hypatos data on straight-through rates and highest-ROI automation processes hypatos.ai
Agentic AI vs RPA in 2026: The Definitive Guide to APA Technical comparison guide for agent-based process automation in enterprise settings ampcome.com
