I told a room of CEOs last week that AI will split software companies into two groups: the accelerated and the disrupted. Three ideas shaped how I framed that divide for the room. Accelerated versus replaced. The Red Queen race we’re all running. And a test I now apply to every AI initiative at SecurityScorecard: show me the before and after.
Accelerated Versus Replaced
AI will hurt software as a service (SaaS) companies that amount to thin workflow wrappers with weak proprietary data and low switching costs. It will accelerate companies with proprietary data, deeply embedded workflows, and systems of record.
I challenged our team with an uncomfortable question: what stops a customer from pointing Claude, Copilot, or another foundation model at a vendor questionnaire and bypassing SecurityScorecard entirely? Every SaaS CEO should ask some version of that question about their own product.
Our answer: foundation models become far more powerful when they operate through SecurityScorecard’s data, workflows, permissions, integrations, and context. We plan to expose TITAN AI everywhere customers already work, including Claude, the Model Context Protocol (MCP), APIs, plugins, and customer-native agents. The durable value sits in the system underneath, not just in the agent sitting on top of it.
AI does not kill SaaS. It kills lazy SaaS.
Our moat compounds across three layers:
- Telemetry. Thirteen years of outside-in cyber data, extended by Driftnet signals. A horizontal large language model cannot recreate 13 years of proprietary telemetry.
- Activation. TITAN AI and our managed services (MAX) turn that data into action: policy reasoning, validation against external telemetry, vendor engagement, remediation, and verification.
- Vendor flywheel. Vendors contribute through, trust centers, vendor responses, reviewer corrections, and remediation outcomes.
A generic model gives you a one-time answer. SecurityScorecard owns the continuous loop around that answer: observe, reason, engage, remediate, verify, learn, repeat. That loop, not a single query, is the architectural shift third-party risk management is making with TITAN AI from point-in-time ratings toward a system of record and system of action for Third-Party Risk Management (TPRM) and supply-chain cyber hygiene.
Launching agents isn’t the finish line, though. Companies still need to point to specific revenue, retention, productivity, and operating improvements that AI actually produced.
The Red Queen Race
We’re also living through an AI Red Queen race. In Through the Looking-Glass, the Red Queen tells Alice: “it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that.” That’s corporate AI today.
Every company is buying frontier model licenses, deploying copilots, automating support, shipping more code, and telling the board it’s becoming “AI-native.” Everyone is running. Few can prove they’re getting anywhere.
Did annual recurring revenue (ARR) per full-time employee (FTE) improve? Did gross margin improve? Did retention improve? Did sales cycles shrink? Did more valuable software actually reach customers? The honest answer, for many teams, is no.
Show Me the Before and After
This is the test I care about most. If developers generate twice as much code, is twice as much useful software actually reaching customers? If support adds automations, did resolution time improve, did satisfaction improve, and did cost-to-serve decline?
We apply that test across the company.
Take sales and marketing, for example. We use AI to continuously extract insights from calls, so we understand what prospects and renewing customers are actually saying. We survey business development representatives (BDRs), solutions architects, and other teams with one question: what wastes their time the most? Then we send people in to automate it and elevate them away from drudgery work. We also built daily leaderboards tracking leads generated by account executives (AEs) and sales development representatives (SDRs), because AI productivity should eventually show up in output.
People, Culture, and Operations: Make AI Fluency Non-Negotiable
We built an internal AI portal where employees share their best AI skills and use cases, and we created a head of AI transformation role under our chief technology officer (CTO) to rethink how work gets done. AI fluency now factors into quarterly performance reviews, from interns through the C-suite, and every candidate completes an AI fluency assessment before receiving an offer.
We also hired a Chief Information Security Officer (CISO), Mike Mylrea, who is also our Chief AI Officer (CAIO). Instead of security saying no to AI, he helps us accelerate adoption with sensible guardrails. Culture matters here. If one leader mentions AI once at an offsite, little changes. But if leadership talks about it, uses it, measures it, and challenges assumptions every week, the organization actually shifts.
We also tell employees directly: the goal isn’t to replace you with AI. If you get great at using it, AI should remove grunt work and free up time for higher-value problems.
There’s a cost lesson, too. You can’t blindly point the most expensive model at every task. You need model routing, open-weight alternatives, and clear economics for where intelligence actually creates value.
Tech and Product: Burn the Boats
Don’t be afraid to burn the boats. The phrase traces back to Hernán Cortés, who had most of his ships disabled or sunk after landing in Mexico in 1519, removing retreat as an option. AI transformation sometimes needs the same mentality. Preserve every old process, organizational layer, and way of working, bolt AI on top, and you end up with the old company plus an expensive AI bill.
We saw this directly with TITAN MAX, our managed services for supply chain risk management. Rather than bolt AI onto the existing workflow, we redesigned the process itself around it. That let us focus the team on higher-value work and output roughly doubled. Don’t use AI to make the old process 20% better. Redesign the process.
Engineering tells a similar story. We leaned into building with AI, and roughly doubled the velocity of GitHub pull requests (PRs). That surfaced the next bottleneck: humans reviewing all the AI-generated code. The problem just moved upstream, into automated quality assurance (QA), testing, deployment, monitoring, and rollback. AI removes one bottleneck and reveals the next one.
Meanwhile, nearly 80% of our engineering research and development (R&D) investment is going into our new platform, TITAN AI, even as many customers continue to use our original platform. We’re serving today’s customers, but funding the future while operating the present may be the hardest part of AI transformation.
Four Ideas Worth Running With
These four ideas aren’t a checklist to close out. They’re a discipline to keep running, because the Red Queen race never truly ends. The companies that get this right will keep asking the uncomfortable question about what stops a customer from bypassing them with a foundation model. They’ll keep measuring the actual before and after. They’ll keep redesigning processes instead of decorating them with AI.
See how TITAN AI turns proprietary telemetry into a continuous action loop, or start a free SecurityScorecard assessment to see where your own organization stands.
