London Startup Paid Raises $21.6 Million to Solve How Companies Bill for AI Agents That Replace Workers

The New Economics of Digital Labor: AI Agents Challenge Software’s Old Billing Model

A London startup’s $21.6 million raise signals the end of seat-based pricing—and the rise of outcome-driven billing for autonomous AI.

LONDON — Back in the 1990s, software firms hit on a simple idea: charge customers per employee “seat.” More workers meant more value, so per-seat billing felt fair. Fast forward three decades, and that tidy math is falling apart. Why? Because AI agents don’t just help employees—they replace them.

Paid, a London-based startup, thinks it has the answer. On Monday, the company revealed a $21.6 million seed round led by Lightspeed Venture Partners, pushing its total funding to $33.3 million. Paid’s mission is ambitious: build the financial plumbing for a world where software agents, not humans, do the bulk of the work.

“The old per-seat model doesn’t make sense when software removes the seats altogether,” said Manny Medina, Paid’s founder and CEO.

And the timing couldn’t be sharper. By 2030, autonomous AI agents could pump nearly $20 trillion into the global economy. Yet the billing systems companies rely on today were built for a world where software made humans faster, not redundant.

When Seats Disappear, So Does Revenue

For software vendors, the problem is brutal and obvious. Imagine a customer service platform that charges $50 per human agent each month. If AI agents take over, that same company suddenly bills far less, even though its software is delivering better results. Customers win with efficiency gains, but vendors get punished for being too effective.

Paid’s early adopters are testing ways out of this paradox. Industrial software provider IFS, which acquired AI startup TheLoops, uses Paid’s system to scale agent-driven solutions in manufacturing and asset management. Artisan, a startup selling “AI employees” for sales development, now ties its fees directly to outcomes like meetings booked rather than licenses sold.

Of course, outcome-based pricing creates its own headaches. What exactly counts as a “resolved ticket” or a “qualified lead”? Without clear, trusted definitions, disputes between vendors and customers are inevitable. The companies that figure out how to verify outcomes fairly will separate themselves from those offering flashy dashboards with little substance.

The New Infrastructure Beneath the Agents

Paid isn’t just tinkering with billing—it’s trying to redefine it. Its platform offers five key tools missing from legacy systems:

ROI portals that prove customer value
Flexible pricing options like revenue-sharing and success fees
Hybrid billing models that blend subscription with outcome-based fees
Real-time tracking of AI agent costs
AI-focused analytics for forecasting and scenario planning

It’s hardly alone in the race. YC-backed Skope brands itself as “the billing system for AI products.” Nevermined focuses on real-time settlements for AI work. Orb positions its platform as “revenue design for the AI era.”

The broader ecosystem is moving too. Google has floated an Agent Payments Protocol to govern transactions between autonomous systems. Academic groups have pitched frameworks like PACT, which tackles quality and liability in AI services. And the Coral Protocol imagines decentralized coordination between agents themselves.

The Seat–Usage–Outcome Continuum

Despite all the hype around outcome-based billing, most companies in 2025 are blending it with traditional models. Data from Orb suggests that pure outcome billing is still rare, while hybrid setups—mixing subscriptions with outcome fees—dominate.

This makes sense. Finance teams crave predictability, so they prefer stable subscription revenue with performance bonuses layered on top. Services firms are moving in this direction too. Globant, for instance, recently shifted from hourly billing to “AI Pod” models that combine reserved capacity with consumption-based charges.

The technical challenge is proving outcomes without loopholes. If a vendor claims a “qualified lead,” the CRM (say, Salesforce) must back it up. If a “ticket resolved” is logged, platforms like Zendesk or ServiceNow—and even customer satisfaction scores—need to confirm it. Without these guardrails, billing systems are too easy to game.

And then there’s accounting. Outcome-based fees complicate revenue recognition, an area already loaded with regulatory nuance. If Paid can’t deliver clean data to ERP and finance systems like NetSuite, adoption will stall no matter how slick the tech looks.

Where the Risks Really Lie

Pricing isn’t the only hurdle. Attribution is another. When multiple AI agents collaborate on a task, who gets credit for the value created? Tracing those contributions in enterprise environments where tools from different vendors overlap is far from simple.

Liability makes things even messier. If an AI agent makes a costly mistake, who’s on the hook—the vendor, the customer, or both? Industries like manufacturing and supply chain demand tight governance, with human checkpoints, audit trails, and clear accountability baked into the system.

Paid has chosen to target CFOs and revenue operations leaders, not engineers. That focus shapes its pitch: less about tech wizardry, more about margin control and audit readiness. Competitors coming from other directions—say, observability tools pivoting into billing—face uphill battles convincing finance leaders they’re trustworthy partners.

What to Watch Next

How will we know if outcome-based infrastructure really takes off? A few signs stand out:

Vertical playbooks: Pre-built catalogs of outcomes—say for customer service, sales, or manufacturing—that reduce friction.
Verification services: Neutral third parties that use cryptography to confirm whether outcomes really happened.
Adoption by services firms: Consultancies shifting their own billing models under pressure from clients could be the strongest signal yet.
Standardized protocols: Industry-wide frameworks, like Google’s payments protocol or Model Context Protocol, that let agents transact seamlessly.

The Investor’s Angle

For investors, the winners in this space will be the companies that can prove outcome integrity while plugging seamlessly into enterprise finance stacks. Deep ERP integration, strong anti-gaming mechanisms, and hybrid billing flexibility will all be critical.

Domain expertise may also be the moat. A one-size-fits-all billing system is too vague. But providers that tailor their models to healthcare, manufacturing, or finance could establish defensible advantages early.

Consolidation seems likely too, as observability and billing vendors merge to connect the dots from system trace to invoice. The first major software suite to roll out outcome pricing as its headline feature would mark a watershed moment. On the flip side, if early adopters stumble with disputes or verification failures, the market could retreat back to safer, usage-based models.

In the bigger picture, infrastructure for AI agent billing looks like the classic “picks-and-shovels” play in a gold rush. There’s clear demand, but execution risks—from technical complexity to regulatory uncertainty—are sky high.

Disclaimer: This article reflects current market data and trends. It isn’t financial advice. Readers should consult licensed professionals before making investment decisions. Conditions in the market may shift quickly.