C-level corner
For CEOs, CIOs, CISOs and boards
Most of this blog is about how. This corner is about why. Every security and data decision should pay off in one of five ways: you get to market faster, you grow revenue, you cut cost, you reduce risk, or you free people from low-value work for work that matters. Each section below starts with the question leaders ask, gives the short answer, and links a brief you can reuse in your own discussions: no product pitches, no code. The technical articles behind each brief are linked quietly underneath, for your architects.
1Go to market faster
Why are our AI projects stuck in pilot?
Rarely because of the model or the budget. They stall in a security review that can't say what the agent is allowed to do. Build the controls into the platform once and every next project starts from a checklist instead of an investigation.
- AI Security Is a Business Accelerator
AI security is how a project gets approved, not the reason it's held up. Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, with inadequate risk controls one of the three reasons. Five controls, built once, turn every later security review into a checklist.
Technical background, for your architects: Keep Shipping, the full article
- From AI Pilots to an Operating Model for Agents
Treat agents as an operating model, not a technology project. One purpose-built agent for your most painful workflow, with its own identity, fixed tools and a test suite, becomes the template for the rest, and brings shadow AI into the open.
Technical background, for your architects: Building the agentic enterprise
2Grow revenue and unlock data
What stops regulated customers and data programmes from saying yes?
Usually a small slice of sensitive data, such as national IDs, bank details or patient records, that blocks the whole programme. Give that slice a stronger answer and the rest of the data can finally be put to work.
- Unlock the Data a Small Sensitive Slice Holds Back
For the sensitive slice, keep one decisive control in your own hands: the data, the key or the switch. Regulators, boards and customers get a demonstrable answer to "what is still under our control?", and everything else can move.
Technical background, for your architects: The framework on one page · Eight patterns, measured (Snowflake) · On-premise data, queried on demand (Snowflake)
3Cut cost
Where are we paying premium prices for work that doesn't need them?
Security and observability tools are often priced by volume, so teams pay real-time rates for data that is only read during investigations and audits. Route each kind of data to the tier it actually needs.
- Your Log Bill Is a Design Choice: Observability and Security Data in a Lakehouse, with Federated Queries New
Security and observability tools are often priced by the data they take in. Keep every log, trace and metric in an open lakehouse you own, send only the time-critical slice to the real-time tools, and query the rest where it lives. In the worked example, the monthly bill falls by 70 to 80% and years of retention become affordable.
The figures are estimates for one scenario, not a quote. Run the numbers with your own volumes and contracts.
Technical background, for your architects: Hybrid security data lake (Splunk, Snowflake) · Logs, traces and metrics into Iceberg (OpenTelemetry) · A governed data lake for under 10 euros a month
4Reduce risk
What will the board, the regulator or the auditor ask about AI?
Who is this agent acting for, what was it allowed to do, and can you prove what it did. With the EU AI Act, DORA and NIS2, that is no longer only a security team question.
- What the Board Will Ask About Your AI Agents
The three questions, the EU rules behind them, and the principle that answers them: restrictions on agents are enforced by access controls outside the agent, never by its instructions, and agent activity shows up in the security monitoring you already run.
Technical background, for your architects: AI governance architecture guide · Agent activity in the SIEM (Microsoft Sentinel) · AI audit and inference regions (Snowflake)
- Software Supply Chain: Why Signing Is Not Enough
In March 2026 a widely used AI library was backdoored through its genuine publishing account, so a signature would have looked valid. Expect three layers: control what comes in, contain what runs, and be able to answer "were we affected?" quickly.
Technical background, for your architects: Layered supply chain architecture (Python, conda)
- European Cloud and the CLOUD Act
Where data sits is half of sovereignty; who controls the provider is the other half. A curated list of European providers without a US parent company, with the nuances marked.
5Better work for people
How do we turn low-value toil into meaningful work, and scale scarce expertise?
Let AI take the repetitive work, and keep people for the judgement calls: deciding how much the automation may do, approving the risky steps, and dealing with what's genuinely new. The same idea lets a few specialists make hundreds of colleagues more capable, instead of joining every meeting.
- A Few Experts, Hundreds of Sales Engineers: Scaling Security Expertise With AI and Just-in-Time Enablement New
How a small group of specialists can support hundreds of sales engineers: an AI skill that researches and writes the way the specialists do, a champions team with a clear career deal, answers in writing first, and helping the engineer run the session instead of taking it over.
- Let AI Do the Routine, Keep People for Judgement
A pattern for handing routine work to AI without losing control: autonomy earned per task from a measured track record, every AI change logged and undoable, human approval for risky steps, and one switch to stop it all.
Technical background, for your architects: The automated SOC prototype (Snowflake) · AI-built slides and sheets (Google Workspace)
Taking this to your team
- Slides. Most articles come as a slide deck you can open in a meeting or share.
- The CISO version of any article. On every article, newsletter subscribers can have Ask the blog rewrite it for a CISO: the risk, the business impact, the decisions leadership owns and three questions to ask the team. The newsletter is free.
- Building the team. Hiring for AI, data or security? The talent board lists people looking for roles, reviewed by hand, with the ones I have worked with and vouch for marked with a star.
- A conversation. For a talk or a discussion with your leadership team, get in touch.
Snowflake-related articles explore the art of the possible and are not official Snowflake solutions or endorsed by Snowflake unless stated. Figures come from the articles and their sources; check them against your own situation before you rely on them.