Serval moved a carmaker off ServiceNow in 45 days. Can you show your AI is live?
A funded challenger now sells customers a move off a category leader in 30 to 45 days, and last week both the challenger and the incumbent defended themselves with production numbers. A board will test a vertical SaaS company the same way: are its AI features live, used and governed, with a number for each? Five numbers answer it: AI features in production, features and accounts live with paying customers, the share of accounts using AI each week, the share of runs traced and scored, and the share of sensitive writes held for approval.
What did Serval's raise actually price?
On 29 October 2026, Serval raised a $150M Series C at a $3B post-money valuation, led by Sequoia, which also led its $75M Series B at $1B in December. It is the company's third round in twelve months and brings the total raised to $277M, so the valuation has tripled in under a year. Serval told Fortune that its annual recurring revenue has doubled every quarter since the Series B, without giving dollar figures, and that it has grown from about 20 people in January to about 200.
The thesis behind the price is specific. In Fortune's account, Serval reads ServiceNow as a workflow automation company built on drag-and-drop tools and custom configuration, the kind of product that code written by AI now makes cheap to rebuild. Its chief executive, Jake Stauch, described taking on ServiceNow as a fight the company chose on purpose.
Swap the names and read it again. If your product is screens, rules and settings that customers have configured over years, the same thesis applies to your category, and the next challenger in it can raise money on that argument.
- $3B
Serval's post-money valuation after a $150M Series C on 29 October, its third round in twelve months and $277M raised in all.
Source: Fortune, 29 Oct 2026 - 45 days
A carmaker moved its whole organisation off ServiceNow, including service management, the CMDB and asset management.
Source: The Next Web and Serval, 29 Oct 2026 - 13x+
ServiceNow's agentic AI consumption in production, year over year, reported the same week.
Source: Fortune, 29 Oct 2026
How long would your largest customer need to leave?
The Next Web reported on 29 October 2026 that a carmaker moved its whole organisation off ServiceNow in 45 days, and that most Serval customers go live within 30 days. Serval's own post says the 45-day replacement, at an automotive manufacturer, covered IT service management, the configuration management database (CMDB) and hardware and software asset management.
Switching cost has long protected category leaders. A system of record that took months to implement took months to leave, and many customers never started. A 45-day migration turns that protection into a project plan with an end date, which a challenger can put in front of your customer at renewal.
So put a number on your own exposure. Take your largest account, list what it would have to move (records, workflows, integrations, reports, the people trained on your screens) and estimate the weeks. Say the number out loud in your next leadership meeting. If the honest estimate fits inside the time left on that account's contract, the account can leave before it renews. The AI features it uses every week are what it would have to rebuild elsewhere, and they count as retention evidence only when you can state that use as a number.
Moving off a category leader is now counted in days
Time to go live on Serval, in days, as Serval describes it
Why are both sides now arguing in production numbers?
ServiceNow answered Serval in the same week with usage. Its third-quarter figures, as Fortune reported them on 29 October 2026: agentic AI consumption in production up more than 13 times year over year, a 98% renewal rate, AI deal volume for IT service management up nearly 80%, and its 2026 target for AI annual contract value raised to $1.7B from $1.5B.
Sean Regan, general manager of Flow by ServiceNow, put the incumbent's case in one line: "Anyone can build a great demo." His argument, in Fortune's account, is that every agent depends on trusted data, workflows and governance, which he said cannot be easily replicated. Stauch turned the question around. He told Fortune he would be curious how many customers have actually put the AI products added to older platforms live.
Read the two together and both sides are arguing from usage: one says its customers go live in weeks, the other that its AI runs in production at scale and renews. A board that reads this coverage will put the same question to you, and someone in the room will want it answered with a figure.
Your customers' security team will ask who approved the last write before it asks to see the demo.
What does a "live" AI feature mean to a customer's IT team?
Serval's own product description, published with its raise on 29 October 2026, reads like a buyer's checklist for a live AI feature. Serval builds automations with one agent, used by administrators, and publishes them to a second agent that employees use. Fortune reports that each automation is first tested in a separate environment with branching, access controls and audit logs. Serval's post says it turns business requirements into deterministic tools, with guardrails, permission checks and approvals managed by admins. And The Next Web reports that customers still want a person to approve sensitive actions such as password resets and access to applications.
Each of those is a question your customers' IT and security teams will put to your AI features before they let them near their records. Where was this tested before it ran on our data? Who can change what it does? Which actions does it take alone, and which wait for one of our people? Where is the log?
ServiceNow's other move points the same way. Flow, a standalone AI service desk launched on 1 October, costs $2 per conversation and can be bought with a credit card by companies with no ServiceNow contract. ServiceNow says it can be running in a day without an implementation project.
The go-live ladder
Five rungs from a demo to governed production, each with one number to fill
How do you measure AI feature adoption in SaaS?
Measure AI feature adoption with five numbers, one for each rung of the go-live ladder, that a CEO or CPO should be able to state on Monday, each with the person who owns it.
- Shipped to production. The count of AI features running for customers, leaving out pilots and sales environments. The CPO owns it.
- Live with paying customers. The number of those features that paying accounts use, and the number of paying accounts using at least one. Customer success owns it, from contract and usage data.
- Used every week. The share of accounts that use at least one AI feature in a given week, counted from run traces per customer. Product analytics owns it.
- Every run traced and scored. The share of runs with a full trace (what the agent read, which tools it called, what it wrote) and a quality score a domain expert would accept. The engineering lead for AI owns it.
- Sensitive writes wait for a person. The share of writes that change access, money or records which are held for approval before they execute. The CTO owns it, with security signing off. If a risk committee has to sign off the first write, here is how to ask it for one bounded action.
If you are the CEO or CPO
Put the five numbers on one page before your next board meeting. For each blank rung, pick the one workflow your customers already ask AI for, and set a date to have it live, traced and scored, with sensitive writes held for approval. Bring three people to that decision: product, a domain expert and the owner of the AI roadmap.
How many of your AI features are live with paying customers today?
Anonymous. One vote per browser.
Where SynOS fits
SynOS gives the AI features inside your product the instruments behind the top three rungs. The SynOS Agent Runtime gives every agent run a trace and a score, gives each agent its own identity, and sets token budgets and reports cost per customer. Writes are recorded rather than executed until someone approves them. It runs white-labelled inside your cloud, with a Brain for each of your customers, so the go-live numbers come from your own systems and are yours to report.
Questions
What counts as an AI feature being "live"?
A feature is live when it runs in production for paying customers and they use it in their own work. A pilot, a sandbox or a sales demo does not count, even when it runs on the same model.
How do you measure AI feature adoption in a SaaS product?
Count the share of accounts that use at least one AI feature each week, from run traces recorded per customer. Report it beside the number of AI features in production and the share of runs that are traced and scored, so adoption is read alongside quality.
Who should own AI feature adoption metrics?
Split them by rung. The CPO owns the count of AI features in production, customer success owns features and accounts live with paying customers, product analytics owns weekly use, the engineering lead for AI owns traced and scored runs, and the CTO owns the approval of sensitive writes, with security signing off.
Which AI actions should wait for a human?
Writes that change access, money or records. Keep them recorded rather than executed until a person approves. Serval's customers, The Next Web reported, still want a person to approve password resets and access to applications.
Sources
- Serval raises $150 million at a $3 billion valuation to take on ServiceNow, Fortune (Nick Lichtenberg) (29 October 2026; the raise, rounds and headcount, the configuration thesis, Catalyst's separate test environment, Sean Regan's quote and ServiceNow's case, ServiceNow's third-quarter AI figures, Flow's launch, price and time to run. Fortune discloses that ServiceNow sponsors its AIQ event.)
- Serval raises $150m at $3bn valuation to replace ServiceNow with AI agents, The Next Web (29 October 2026; the 45-day move by a carmaker, most customers live within 30 days, the separate builder and employee agents as told to Reuters, customers wanting a person to approve sensitive actions)
- We raised $150M to make everyone a builder, Serval (29 October 2026; deterministic tools, guardrails, permission checks and admin-managed approvals, the 45-day replacement at an automotive manufacturer and the systems it covered)