Pit raises $16M to sell enterprise AI as a product team, not another chatbot

Pit, a new Stockholm AI startup founded by former Voi people, has raised a $16 million seed round led by Andreessen Horowitz. The company is building enterprise software that learns how a business runs, then generates internal tools and automations around those processes.

It’s a big pitch in a crowded market.

Pit is led by Adam Jafer, who spent seven years at Voi as the scooter company scaled to nearly 1,000 employees across 13 countries. Voi CEO Fredrik Hjelm is also a co-founder, though he remains in charge of Voi, which became profitable in 2024 and has been discussed as a possible IPO candidate. Filip Lindvall, another Voi co-founder, is a founding engineer at Pit. The team also includes former iZettle and Klarna engineers.

The round stands out because of who wrote the check. A16z has been actively looking for European AI companies, and Stockholm has become one of its more interesting stops. The city is also home to Lovable, another high-profile AI startup drawing attention from U.S. investors. Pit now joins a small group of European companies trying to sell AI into large businesses before the category hardens around a few winners.

The product bet

Pit calls itself an “AI product team as a service.” The phrase is a little heavy, but the idea is clear.

The company has two main pieces:

• Pit Studio, where enterprise employees describe and map the internal processes they want automated.
• Pit Cloud, the runtime and delivery layer meant to provide governance, certifications, auditability, and the operational controls large companies expect.

The target use case is back-office automation. Pit says its pilots, which started in mid-January, are in telecom, healthcare, logistics, and other sectors. It’s focusing on internal service, support, and administrative workflows rather than customer-facing chatbots.

That focus matters. Customer-facing AI agents carry brand risk, unpredictable users, escalation problems, legal exposure, and public failure modes. Internal workflow automation still has risk, but the blast radius is usually easier to contain. If an AI-generated tool mishandles a support triage flow or botches a compliance checklist, the company can often catch the problem before customers see it. Often, not always.

Pit is betting that large enterprises have too many processes scattered across spreadsheets, legacy SaaS tools, ticket queues, Slack threads, and tribal knowledge. That’s a fair diagnosis. Every big company has strange internal software because the official system couldn’t bend far enough.

The hard part is capturing the actual process accurately enough to automate it.

Agentic software still has to survive requirements gathering

Jafer told TechCrunch that the opportunity became obvious when models moved beyond text generation and became “more agentic,” meaning they could take actions rather than just produce answers. In practice, that usually means models can call tools, query databases, update records, trigger workflows, and coordinate multi-step tasks under some level of supervision.

For developers, the interesting question isn’t whether an LLM can generate a CRUD app. It can. The harder question is whether an AI system can safely turn a messy business process into durable software that behaves predictably under edge cases.

Enterprise workflows are full of hidden rules:

• “Approve automatically unless the vendor is new and the amount is above €20,000.”
• “Route this to the regional manager, except for German accounts, where legal reviews first.”
• “Use the old billing system for contracts signed before 2021.”
• “Never email this class of patient record outside the secure portal.”

Those rules often aren’t documented. They live with operations managers, senior support staff, finance teams, and people who have learned the exceptions by being burned before. Pit Studio appears designed to extract that knowledge from employees directly, then convert it into software.

That’s a reasonable approach, but it puts an old software engineering problem inside a new AI wrapper: requirements gathering.

AI can speed up implementation. It can propose schemas, generate UI components, write integration glue, and scaffold approval flows. But if the underlying process is ambiguous, political, outdated, or contradictory, automation can harden bad behavior into production systems.

A demo looks clean when the workflow is clean. A rollout gets ugly when the company’s actual process has 14 exceptions and nobody agrees which one is canonical.

Why Pit is hiring solution engineers

Pit isn’t pretending the product will sell and deploy itself. The company is hiring solution engineers, following the pattern set by AI firms that use forward-deployed engineers to push adoption inside large customers.

That’s a practical concession.

Enterprise AI tools often fail because the vendor ships a platform and expects the customer to figure out the use cases, integration points, risk controls, and internal adoption path. Large companies rarely have spare teams waiting around to rebuild operations around a new AI tool. They need help turning the product into working systems, especially when the software touches sensitive internal processes.

Forward-deployed engineers and solution engineers can bridge that gap. They sit close to the customer, understand the workflow, wire up systems, and often write the unglamorous integration code that turns a sales deck into something useful.

The trade-off is margin and scalability. A company selling “AI product team as a service” may need a lot of human product-team labor to deliver the service. That can work if the resulting automations become reusable patterns, or if Pit’s tooling gets better with each deployment. It gets harder if every enterprise customer needs bespoke process consulting, custom integrations, and weeks of hand-holding.

The stronger version of Pit’s model looks like this: human engineers handle the first messy deployment, the system learns reusable workflow patterns, and later deployments get faster. The weaker version looks like a consulting firm with an AI wrapper and venture-scale expectations.

Investors are betting on the stronger version.

Governance is where enterprises will spend

Pit Cloud may be the less flashy half of the product, but it’s probably the more important one for serious customers.

Enterprises don’t just need generated software. They need to know who approved it, what data it touches, what model produced which action, how errors are logged, how access is controlled, and whether an auditor can reconstruct the system’s behavior six months later.

Pit will need strong answers on:

• Identity and role-based access control
• Data residency and tenant isolation
• Audit logs for model actions and human overrides
• Integration security for systems like ERP, CRM, HRIS, and ticketing platforms
• Model selection and fallback behavior
• Change management when generated workflows are modified
• Observability for failed automations, latency, and cost

This is where many AI agent startups sound thin. They show a slick builder, then wave vaguely at governance. Large companies won’t accept that for workflows involving healthcare, logistics, finance, or employee data.

Pit says it can work with different AI and cloud vendors based on customer preferences. That vendor-agnostic approach could play well in Europe, especially with current interest in sovereign technology. Jafer said “EU models running on EU compute” are top of mind for many CIOs the company meets.

That checks out. European companies, especially in regulated sectors, are asking harder questions about U.S. cloud dependency, data transfer, and model providers. An AI automation platform that can run on preferred infrastructure, including European compute and European model providers, has a better shot in boardrooms wary of sending sensitive process data through a black-box U.S. stack.

The caveat is engineering complexity. Supporting multiple clouds, model APIs, deployment targets, and compliance regimes gets expensive quickly. Model behavior also varies. A workflow tested against one frontier model may behave differently with a smaller EU-hosted model. Pit will need evaluation harnesses, regression testing, and clear reliability boundaries if it wants “bring your own model” flexibility without turning deployments into science projects.

The junior engineer controversy

Pit got attention earlier this year when Jafer posted on LinkedIn: “Yes, our team currently has no junior engineers. At Pit, agents now do most of what junior engineers used to do.”

He has since walked that back, telling TechCrunch that while it may have started that way, companies need a good mix as they scale.

The original post was predictably provocative, but it touched a real issue. AI coding tools are changing the shape of entry-level engineering work. They’re particularly good at tasks that used to train junior developers: scaffolding components, writing boilerplate, translating requirements into first-pass code, generating tests, and stitching APIs together.

That creates a training problem. If startups skip junior hiring because AI handles the simple work, they may save time now and weaken their senior talent pipeline later. Senior engineers don’t appear fully formed. They get there by debugging bad assumptions, maintaining code they didn’t write, learning production failure modes, and sitting through the boring parts of software delivery.

For Pit, the tension is hard to miss. It’s selling automation that moves workers “upstream” into higher-value work, while its own early messaging suggested AI could replace junior technical labor. Both can be true in different contexts, but enterprise buyers will still ask whether AI automation improves jobs or quietly removes them. Vendors need a better answer than a softened LinkedIn post.

Pit has also drawn scrutiny for its founder-heavy, male-coded “tech bros from Voi and Klarna” image, which Hjelm referenced on X while noting that women are also on the team. That may sound like startup gossip, but enterprise sales depends on trust. Culture, hiring, and credibility matter when a company asks customers to expose internal processes and let software generate operational tools around them.

Why Stockholm keeps producing companies investors chase

Pit’s $16 million seed round includes a16z, Lakestar, the founders themselves, executives from American tech companies, and wealthy Nordic families. Jafer said the company didn’t spend much time shopping the round around.

That’s founder-market privilege, and Pit has plenty of it. The Voi network gives the company operator credibility, fundraising access, and a story investors understand: a team that built and scaled a complex European business now wants to automate the internal machinery it had to build the hard way.

Stockholm helps too. The city has a dense pool of engineers and operators from companies like Spotify, Klarna, iZettle, Voi, and now newer AI firms. It also sits in a region where industrial companies, logistics operators, telecoms, and regulated enterprises can be serious buyers of internal automation if the compliance story holds up.

Pit’s European positioning could become an advantage. U.S. AI startups often lead with speed and model capability. European enterprise buyers increasingly want those things plus data control, procurement flexibility, and defensible governance. A startup that speaks that language natively may get meetings others don’t.

Still, Pit is entering a brutal category. AI agents, internal tool builders, automation platforms, workflow vendors, and consulting-heavy AI deployment firms are all circling the same budgets. The company has a strong team, a serious seed round, and a plausible wedge into enterprise operations. It still has to prove that generated internal software can survive the messy parts of real companies without becoming another layer of expensive software sprawl.