One year after our Series C: what $85M bought
In March 2025 we raised an $85M Series C led by Northgate Partners. A year later, here's where the money went, what we got right, what we got wrong, and what comes next.
A year ago this week, we closed an $85M Series C led by Northgate Partners, bringing the total we've raised to $142M. At the time we said what the money was for. A year on, we'd rather tell you what it actually bought.
So this is that post: a plain accounting of the last twelve months, written for our customers, our team, and the people thinking about joining it.
What we said we'd do
When we pitched the round, our plan fit on one slide. We'd use the money to do four things:
- Make Pazau Agent resolve more. Not answer more questions. Resolve more conversations, under a definition strict enough to bill on.
- Become boring for regulated buyers. EU data residency, more certifications, and contracts that a hospital network's privacy office could sign without a six-month review.
- Make voice a first-class channel. Voice was our fastest-growing product and our least mature one.
- Stay small enough to move fast. Hire deliberately, especially in AI and security, and not at all in places where software could do the job.
Here's how each one went.
1. Resolution
Pazau Agent now resolves 67% of conversations end-to-end on average across our customers, with an enterprise median of 58%. That's measured under our billing definition: no human touch, no repeat contact on the same issue within 7 days, no abandonment. Both numbers are meaningfully higher than they were when we raised.
Most of that gain didn't come from better models. It came from giving the agent more ability to act and more structure to act within:
- Actions grew from a handful of prebuilt integrations to more than 120 connectors, plus the ability to call any REST or GraphQL API.
- Procedures replaced our old prompt-based architecture across every customer. Dr. Mei Tanaka wrote about why in Procedures, not prompts.
- Checkpoints and approvals let customers turn on Actions they'd previously kept human-only, like refunds above a threshold and claims-related changes.
The metric our customers care about most moved too. Across customers who've been with us a full year, median cost per contact fell 41%.
The best thing we did this year was refuse to change the definition of "resolved" when it would have made our numbers look better.
We're now deep into the next generation of Pazau Agent. It's built around one question our customers keep asking: how do I know a change is safe before it goes live? More on that later this year.
2. Regulated-ready
This was the least glamorous part of the plan and the one I'm proudest of.
EU data residency is generally available. We set up Pazau Ireland Limited in Dublin, and EU customers are now processed and stored entirely in the EU, with Dublin as the primary region and Frankfurt for failover. The full story is in EU data residency is generally available. Vasa Mobility, which runs support across 40 cities in 11 languages, was one of our design partners.
We added ISO 42001, the management system standard for AI, alongside SOC 2 Type II, ISO 27001, ISO 27701, and HIPAA. We believe we were among the first customer-support platforms to certify against it. We also completed PCI DSS SAQ-D for payments taken over Voice.
We rewrote our contracts. Our published Data Processing Addendum and AI Addendum now say, in plain words, that Pazau does not use customer data to train or fine-tune foundation models. That commitment was always true in practice. Now it's in the paper, and changing it would require sign-off from both our CISO and our Head of AI.
None of this shows up in a demo. All of it shows up in how long procurement takes. Enterprise security reviews that used to run three months now regularly close in weeks.
3. Voice
Voice was the bet we were least sure about. A year later it's the channel customers ask about first.
Meridian Mutual uses Pazau Voice for first notice of loss on property and casualty claims, with 100% of calls QA-scored for compliance. Tandem Credit Union handles card disputes over PCI-compliant voice. Both are workflows where a wrong answer is a regulatory event, and both are running in production.
The engineering work was mostly about latency. A voice agent that pauses for two seconds sounds broken, no matter how good its answer is. We rebuilt the voice pipeline around a strict end-to-end latency budget, and our Voice team will write up the details soon.
4. The team
We're now about 420 people, across San Francisco, New York, London and Dublin. We serve about 2,100 customers, handle more than 140 million conversations a year, and cross $60M in ARR.
A few things about how we hired:
- AI and evaluation got the biggest share of new engineering headcount. Better agents come from better measurement. We built an evaluation team whose only job is to tell us, and our customers, when something got worse.
- Security grew faster than the company. Omar Haddad's team roughly doubled. For the customers we serve, that's product work, not overhead.
- Dublin became a real office, not a legal entity with a mailbox. Our Data Protection Officer, Aoife Brennan, leads privacy from there, alongside engineering and customer success teams that support our EU customers in their own time zone.
- We said no to a lot of roles. We run our own support on Pazau, and our own contracts go through the same rigor we ask of our customers. When software could do the job, we didn't hire for it.
What we got wrong
A year-in-review with no mistakes in it isn't worth reading. Here are three:
- We launched Voice in too many languages at once. Quality in our long-tail languages lagged for months. We should have gone deeper in fewer languages first.
- We underinvested in admin tooling. Customers with hundreds of Procedures told us managing them was painful. That's being fixed, but we were slow to hear it.
- We let a few enterprise deployments go live without enough historical testing. In each case the agent handed off more than it should have, which is the safe way to fail. The customer still paid the cost, in human hours, and trusted us less. That experience shaped what we're building next.
Where the money is going now
We still have most of the round in the bank. We've been disciplined because we plan to be around for a long time, and because our customers are betting their front line on us.
The next twelve months look like this:
| Priority | What it means |
|---|---|
| Safe change | Test every Procedure change against real historical conversations before it ships |
| Deeper Actions | Multi-step Actions with approvals, spend limits and rollback |
| Voice parity | Everything Pazau Agent can do in chat, it should do on the phone |
| Insights | From dashboards to recommendations: which Procedure to fix next, and why |
| Regions | More residency options for customers who need them |
Thank you
To our customers: thank you for holding us to a strict definition of "resolved", and for telling us, loudly, when we fell short of it.
To Northgate Partners and our earlier investors: thank you for backing a company that would rather be trusted than be first.
And to the 420 people who do the work: this was your year.
If you'd like to help write the next one, we're hiring. You can read more about how we work on our company page.