FEATURE: INVESTMENTS when compliance requirements evolve( and they always do) or business priorities change( and they usually will). That means CIOs keep their organizations flexible instead of getting stuck with a massive model that no longer aligns with what they need. Running AI in-house or on private infrastructure with SLMs gives organizations more control over their data, reducing the chances of exposing sensitive information or violating data residency rules.
It’ s why we developed our own tool, Generate Enterprise, to let enterprises keep Generative AI securely behind their own firewall while still offering competitive performance.
Another major benefit of SLMs is security. Public models make it difficult to know exactly where data goes or how it might be used in the future. By prioritizing investments in SLMs with the ability to run fully air-gapped, CIOs maintain tighter oversight of their AI environment. That helps protect against prompt injections, data leaks and attacks on connected systems.
SLMs also support modular, adaptable architectures. Instead of relying on a single massive model, companies can deploy multiple specialized models that each serve specific departments or workflows. If one needs an update or faces an issue, only a small part of the system is affected. That limits disruptions and encourages experimentation, which is essential for driving continuous innovation without risking stability. center: AI is explicitly not allowed to answer customer questions. It might be fast, but if it hallucinates or omits something critical, we pay the price. That’ s not a risk we accept lightly.
To keep that discipline in place, we’ ve established an internal AI council. They help evaluate both the products we build and the tools we adopt internally, making sure we have clear, responsible guardrails. If we can’ t codify those guardrails simply – if it takes a policy manual to explain – then we wait. Simplicity isn’ t just elegance; it’ s enforceability.
That said, we’ re bullish on experimentation. We want our engineers and teams to test and play, so we provide sandboxes where they can try things without touching production or regulated data. Most experimentation doesn’ t need permission – it just needs boundaries. And boundaries create freedom. They make it safe to take risks and learn without unintended consequences.
For non-engineers, we apply the same philosophy. If a tool doesn’ t integrate with our systems or access sensitive data, they’ re free to explore. If they find something worth using more broadly, they bring it to our onboarding process, which includes legal, privacy and security review. If any one of those three says‘ no’, it doesn’ t move forward. But if we all agree, we move fast and help operationalize it safely. It’ s a gate, not a wall.
Henric Andersson, CIO, Intellistack
CIOs who choose targeted, specialized AI capabilities instead of chasing the largest and most complex models will be far better positioned to deliver meaningful innovation while keeping their organizations secure, agile and financially sound. Investing wisely in technologies that truly solve business problems is what will set successful leaders apart as AI becomes an everyday part of enterprise operations.
Henric Andersson, CIO, Intellistack
At Intellistack, we’ re operating in a world where AI is reshaping how we build, ship and support software. But as CIO, I’ ve learned that the best way to stay ahead isn’ t to chase every new capability – it’ s to balance innovation with resilience and to do so intentionally.
That balance starts with one simple rule: AI should amplify systems that already work – not compensate for ones that don’ t.
The key to keeping this process lightweight is training and guidance. When people understand the risks and trade-offs, most don’ t need to be policed. They self-regulate. And that keeps the onboarding process focused and efficient. A little knowledge goes a long way, especially when paired with trust.
If there’ s one piece of advice I’ d offer to other CIOs trying to strike this same balance, it’ s this: Give people a playground. Hype dies quickly when people can test things hands-on. Empower them so they don’ t have to sneak around – because nothing fuels shadow IT like a culture of‘ no’. Be a partner. Educate your teams. And when it’ s time to go live, always weigh the risk and the blast radius.
That’ s how we do it at Intellistack. And so far, it’ s kept us moving fast – without losing control.
Jeremy Brown, CTO, GitGuardian
We don’ t deploy AI into areas where the fundamentals aren’ t solid. For example, while we use AI to assist in writing code, every change still goes through human review. Same with our security questionnaires and trust
We’ re in the middle of an AI explosion, and frankly, CIOs are caught between a rock and a hard place. Everyone wants to experiment with AI tools, but you’ re worried about security disasters waiting to happen.
www. intelligentcio. com INTELLIGENTCIO NORTH AMERICA 37