SiteMinder’s CTO on doubling AI code output every month
AI coding agents have spooked SaaS investors enough to wipe more than a billion dollars off the sector’s valuations. The theory: if AI makes building software trivially easy, why pay for someone else’s product when you can just build your own?
Tom Varsavsky, CTO at hotel software platform SiteMinder, thinks that logic is incomplete. His take is worth reading closely, because SiteMinder is not sitting on the sidelines debating AI’s potential. They’re in the middle of a real deployment with measurable numbers.
The “Build Your Own SaaS” Trap
Yes, AI makes writing code faster. That doesn’t mean rebuilding your CRM from scratch with a few Claude prompts is a smart use of your time.
“Building the software has never been the hard part in building a business. It’s just one part of the value chain, and all the things around it are probably more important. Businesses need to figure out what customers want in order to build it, they need to find customers, they need to service them, and they need to deliver on the value proposition. All those things are much bigger than the code.” — Tom Varsavsky, SiteMinder
His point about opportunity cost is direct: a software company’s time is better spent clearing its customer backlog than retooling internal systems. “I’m not spending any time thinking about how to replace Salesforce with my own vibe-coded solution,” he said. “I’m spending all my time trying to figure out how my engineers can now go faster with AI to deliver our roadmap so we can grow the business.”
He does carve out one exception: SaaS products that handle fringe business processes rather than core systems of record are more vulnerable, because experimenting with AI alternatives there is lower risk for the customer. “I think there will be winners and losers, but it’s not a catastrophe that’s playing out,” he said.
The Numbers SiteMinder Is Actually Seeing
Varsavsky pinpoints Claude Opus 4.5, released last November, as the inflection point. Since then, the volume of AI-generated code at SiteMinder has been doubling every month. Claude has become one of SiteMinder’s top five vendors by spend.
The practical results show up across the software lifecycle:
- Initiatives that previously took months are now completed in weeks.
- Projects that required a full team can now be delivered by two or three people.
- The feature delivery team has recorded examples of working 50% faster with AI.
- QA processes, including code reviews, have been accelerated by AI, saving half an hour per day per person.
Product managers are using AI tools to code prototypes before ideas go to engineering, which means customer feedback gets collected earlier in the process.
️ A Real Example: The Library Upgrade Problem
Tech debt and version management are the unglamorous work that slows every engineering org down. Normally, asking every team to upgrade to the latest version of a particular library pulls them away from shipping customer value. It’s the kind of task that sits on the backlog for months.
SiteMinder’s developer experience team spent a couple of weeks building and testing a Claude skill specifically for that task. Once built, the skill was deployed across every codebase in the company. The upgrade completed an order of magnitude faster than the traditional approach would have allowed.
Varsavsky frames the economics clearly: a dollar invested in the developer experience team yields ten dollars in benefit across the other engineering teams, because that team owns the tooling and processes everyone else depends on.
A separate example: one of SiteMinder’s engineering managers used Claude Code to build a dashboard that pulls data from GitHub, Jira, Confluence, email, and Slack to surface team progress, blockers, and who needs help.
How Agentic Coding Actually Works in Practice
Varsavsky pushes back on companies that announce they’ve stopped manual coding entirely. He calls this a “vanity metric” used by organizations that see AI as a panacea rather than a tool.
That said, the balance has shifted. Twelve months ago, developers used AI as an assistant: write the code yourself, let the AI suggest changes. Over the last six months, that workload has moved toward agentic coding. With Claude Code, the developer interacts mostly with the agent while code is generated in the background and inspected infrequently.
“I find that when I’m coding with Claude, I’m spending less than 10% of the time looking at the code, and 90% of the time I’m interacting with the agent.” — Tom Varsavsky, SiteMinder
SiteMinder still runs full QA: code reviews, quality gates, the works. Varsavsky is explicit about why: “One of our key value propositions is the robustness and reliability of the platform; we just can’t afford to lose reservations so we take a more cautious approach to that than maybe you would if you were a startup with no customers.”
Token Economics: The Next CTO Headache
With Claude usage doubling monthly, the cost line is moving fast. Varsavsky draws a direct parallel to the early days of cloud computing, when engineering leaders had to learn to manage variable consumption costs after decades of predictable on-premises hardware bills. He expects the same learning curve with tokens.
Right now, SiteMinder is prioritizing adoption over optimization. The cost is growing, but the focus is on getting the most out of AI while maintaining quality and security standards. The next phase, which Varsavsky describes as a future priority, will shift toward measuring which token spend actually delivers value.
The honest admission: “We know that some of our tokens are not going to the most valuable outcome, but I can’t tell you which ones.” The work ahead involves building the analytics and metrics to distinguish between teams and individuals using the tools in different ways, both to control costs and to amplify what’s working across the company.
Engineering productivity measurement has never had a clean universal metric, and Varsavsky doesn’t expect AI to solve that problem. “The industry hasn’t come up with a universal measure of engineering productivity just yet, and I’m not sure we will with AI either. So, it will rely on leadership judgment in the end, but we need some data to back that up.”
How SiteMinder Is Driving Adoption Internally
Technology people are naturally agile learners, Varsavsky observes, so organic adoption is already happening. His role as CTO is to accelerate it. The mechanisms SiteMinder is using:
- A developer experience team that triages new tools, figures out what works, deploys them, and encourages widespread use.
- A guild of champions that shares successful recipes across teams.
- An internal Slack channel for surfacing the latest AI developments.
- A company-wide AI hackathon held in 2025.
Some engineers are slower to adopt, whether from skepticism about output quality or simple inertia. SiteMinder deals with those cases individually rather than waiting for everyone to self-select in.
The longer-term structure Varsavsky is building toward: smaller, faster feature teams running at high throughput, supported by a growing platform team that maintains consistent architecture, easy deployment, security guardrails, and QA infrastructure. The platform team is the force multiplier that lets the feature teams move fast without supervision.
“In the end, businesses that provide good value for the customers at a reasonable cost will win. That’s what we’re focused on. In technology shifts, there’s always risk and opportunity, and the companies that come out of this well will have taken up the opportunity.”
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
