Manny Medina is in the midst of his next bet on how artificial intelligence should make money. His newest startup, Paid, has raised an oversubscribed $21.6 million seed round to enable “results-based billing” for AI agencies so companies can pay only when their automated workers deliver real-world impact. The London-based company has now raised a total of $33.3 million, including a €10 million pre-seed, and at a valuation reportedly “well north” of $100 million.
Instead of seats, or token packages sold over time, Paid offers the rails on which to meter outcomes—margin saved, tickets resolved, invoices matched, leads qualified—and turn those into invoices. The pitch for agent builders squeezed by model and cloud costs is straightforward: match the revenue you take in with the work your agent does verifiably.
Why Outcome Pricing Complements Agentic AI
Per-user SaaS pricing does not survive in an agent-first world, because your agents (not their logins) consume the compute. Unlimited use deals can come back to bite you when model calls and GPU time grow faster than revenue, a phenomenon well known to teams experimenting with autonomous coding or support bots.
Your trust gap is not completely cured by usage caps through ‘credits.’ Enterprise buyers want to pay for specific business results, not probabilistic outputs. Gartner and MIT Sloan analysts have observed that most AI pilots hit the brakes before they make it out of the pilot phase, partly because value can be difficult to measure or tie to financial results.
Results-based billing addresses this incentive mismatch. An agent writes a hundred emails few customers read? The meter doesn’t budge. If it settles invoices that would have taken hours of manual work or averts stockouts that eat away at margin, the company writing the software gets paid. The transaction model focuses on accuracy rather than quantity.
How Paid Measures Value and Verifies Real Outcomes
Paid’s platform is a layer between the agent and the customer systems of record, instrumenting workflows with deterministic event data. It logs when an agent starts, completes, or reactivates a task; reconciles outcomes against entries in ERPs, CRMs, or data warehouses; and computes agreed-upon unit economics (e.g., cost-to-serve averted or revenue reclaimed).
Picture a collections agent who reaches out to late payers. With Paid, you can link an automated sequence to the actual payments received and differentiate what would have occurred with no intervention based on historical baselines. For a support agent, it may contribute value to ticket deflections or first-contact resolution enhancements evidenced in the help desk.
Preliminary traction at early customers includes ERP vendor IFS, an interesting signal since outcome-based pricing is most effective in environments where data integrity is high and processes are well defined. With regular KPIs and audit trails, finance departments can manage agent spend as another performance-based supplier contract.
What’s Behind the Funding Round and Key Supporters
Lightspeed led the round with FUSE and also saw existing investor EQT Ventures return for its second investment in Paid to date. Infrastructure is far from the end of the story… The pain of extracting value out of AI strategies is very much centered at the application layer, and that’s where Paid fits in. Lightspeed has talked openly about its heavy investment in both AI infrastructure and corresponding application layers in the last few years, while partners, including John Vrionis, have voiced how tough it can be for industry to convert pilots to ROI — exactly what Paid hopes to remove.
Medina, who built Outreach into a multibillion-dollar sales platform, is now applying his familiar go-to-market muscle to a new pricing substrate. It takes on the operational overhead of dealing with telemetry, pricing logic, contracts, and billing so that startups don’t have to cobble together their own internal tools to be able to charge for outcomes.
The oversubscribed round and a nine-figure valuation imply that investors believe the business of commercial plumbing — how agents prove value and collect cash — will be as important as model quality in the next phase of AI adoption.
Competition and the Stakes for Outcome-Based AI
Usage-based pricing is common in cloud and data platforms, but linking price to business outcomes adds new challenges. Sellers need defensible attribution, guardrails against metric gaming, and human-in-the-loop confirmation for edge cases. Paid’s bet is that a pretty rigorous set of standardized outcome definitions, third-party verification, and transparent audit logs will become table stakes for agent contracts.
The opportunity is sizable. As businesses experiment with agentic systems in finance, supply chain, and customer operations, the winners will make cost-to-value math easy for CFOs. If Paid can be that neutral layer of translation between events in operations and the things that need to get paid for, then it might exist at or near an intersection point of a quickly emerging class of market.
What to Watch Next as Paid Scales Results-Based Billing
Near-term benchmarks include increasing integrations with ERPs, CRMs, and data lakes; prebuilt outcome templates for common workflows; and compliance frameworks to satisfy audit teams. And expect early customers to focus on high-volume, easily measured actions — think invoice matching, returns processing, lead routing — where attribution is clean and pay-for-performance can shine.
If it could turn out to be the case that results-based billing is a means of driving AI waste down and margins up, that raises the possibility for more widespread deployments than pilot projects of the first generation. And with its new funding, Paid has a long enough runway to test that thesis at scale — and in so doing, provide a concrete answer to the question that’s dogged AI agents from day one: what exactly are we paying for?
