The strategy stories enterprise IT leaders actually read.
Volume 342, Issue 1 · 20 articles from 5 issues
Regulated industries like travel need composable architecture and explainability processes to ensure AI agent accountability and decision transparency, writes a CTO.
Paul Curtis, CTO of easyJet, argues that enterprises deploying autonomous AI agents must establish robust monitoring and interpretability strategies before deployment.
He emphasizes that organizations need to log agent reasoning, build decision dashboards, and create feedback loops for retraining, particularly in regulated industries like travel and hospitality.
Curtis highlights that a 2025 MACH Alliance survey found accountability concerns among 38% of enterprise decision-makers, and advocates for composable, API-first architecture as essential infrastructure for effectively managing multiple agents across customer-facing and operational functions.
Aerospike's Srinivasan Seshadri says AI cost overruns live in the data tier, not model or token spend.
IDC projects AI infrastructure costs at Global 1000 companies will run 30% above current budgets by 2027, a gap the author attributes to the data layer rather than model size or token pricing.
Pilots run on narrow datasets with few users, while production agentic workflows expand one prompt into tens or hundreds of parallel lookups across CRM, session and vector systems under sub-100ms latency budgets.
The overrun accumulates from locally rational defensive choices: over-provisioning for spikes, duplicating data and layering services to insulate against slow dependencies.
Recommended fixes include matching each data tier to its access pattern, engineering for fan-out and tail latency, treating write consistency as correctness, and making KV cache reuse first-class infrastructure.
Risk assessments become powerful culture-building tools when their findings are communicated beyond the security team.
Organizations often conduct risk assessments but fail to share findings beyond security teams, leaving business stakeholders to view controls as barriers rather than protections.
CISOs can use assessments to drive cultural change by translating cyber risk into business language, securing executive buy-in, and reinforcing security-aligned behavior through repeated messaging.
Key strategies include assessing baseline security attitudes, rewarding positive behavior, treating failures as learning opportunities, and staying accessible to employees. Success requires years of effort involving leadership alignment, transparent communication, and consistent reinforcement.
CIOs risk losing strategic influence over AI decisions when they lead with caution rather than confident positioning.
CIOs are increasingly bypassed in critical AI strategy conversations as CEOs seek confident answers from junior data leaders, vendors, and outside specialists.
The article argues this is a positioning problem rather than a competence issue: the cautious, risk-focused stance that once built credibility now reads as resistance to progress.
Technology leaders who retained influence developed defensible AI positions before being asked and presented solutions that address risks instead of merely flagging obstacles.
By orchestrating the AI conversation and bringing specialists into their own framing, these leaders stayed strategic decision-makers rather than implementers of someone else's strategy.
Post-quantum migration is now an operational priority, with major providers targeting 2029 for full post-quantum security.
Google and Cloudflare have both set 2029 as a target for full post-quantum security across their infrastructure, driven by rapid advances in quantum research and a shrinking margin for delay.
The article argues that what was once treated as a future problem is now an immediate operational priority for IT leaders, with major tech companies and national security agencies setting concrete migration timelines.
Post-quantum migration sits inside asset management, supplier governance, and sound operational planning rather than standing apart as a cryptography project. Organisations that start early, set priorities clearly, and coordinate across their supply chain will have more options and face less disruption.
Legacy systems are more than just a drag on efficiency or merely older programs; they are strategic inhibitors that prevent resilience, innovation and growth.
Quantum computing remains in development, but quantum annealing is already delivering value for complex optimization problems in logistics, telecommunications, and transportation sectors. While gate-based quantum computers could eventually revolutionize molecular simulations for pharmaceuticals and aerospace, they remain years from commercial deployment.
CIOs should consider partnerships with quantum technology providers to build expertise early and address post-quantum cryptography threats. Industries with significant optimization challenges should prioritize quantum annealing exploration today, treating it similarly to high-performance computing initiatives.
Seven essential books help CTOs develop leadership, strategy, and organizational management skills beyond technical expertise.
Modern CTOs require more than technical proficiency to succeed; they must align engineering with business goals, lead high-performing teams, and make strategic decisions.
This curated reading list includes The Manager's Path by Camille Fournier for leadership progression, The Hard Thing About Hard Things by Ben Horowitz for crisis management, Accelerate by Nicole Forsgren et al. for evidence-based engineering decisions, Team Topologies for organizational structure, Good Strategy Bad Strategy by Richard Rumelt for business alignment, Inspired by Marty Cagan for product development, and Radical Candor by Kim Scott for effective communication.
CIO Interview: Frederic Laurent, Deputy CIO, Starling Bank
Computer Weekly interviews Frederic Laurent, deputy CIO at Starling Bank, about how he applies AI both personally and professionally. Laurent describes AI as an exceptional educational tool, using models to explain unfamiliar concepts in simple terms while always requesting references because models can hallucinate.
He uses Google's NotebookLM to build an outsourced memory from articles and research papers that he can query directly.
Laurent also submits problems to AI to surface alternative perspectives, then asks the model to contradict its own answer as a check. He argues that anyone claiming AI has not changed how they work has missed the boat.
CIO argues work breakdown determines both merge quality and how much agent capacity a team can actually use.
When several agents work the same codebase, the article argues, the way assignments are divided determines the outcome. Done well, merges fit together and align to team goals; done badly, you get changes that each look right alone but do not add up.
Work breakdown means splitting work into clean pieces, naming what depends on what, and keeping that plan current as things move.
The article also connects breakdown quality to capacity utilization, because you can only run as many agents as the decomposition supports. It is aimed at engineering leaders scaling past a handful of agents.
Argues business alignment is an org-design problem, and proposes a business-facing CSO sitting above the technical CISO.
JC Gaillard of Corix Partners argues the modern CISO job has become unsustainable, bundling technologist, strategist, risk executive, board adviser, crisis manager and business partner into one role.
His central claim is that business alignment is not primarily a communication problem but a leadership and organizational design problem.
He proposes a dedicated Chief Security Officer who is a business leader first, owning cybersecurity, data protection, business continuity and resilience, with authority to reconcile competing departmental interests.
The CISO would keep technical accountability, and separating the roles could make the CISO more effective by removing board politics from the job.
From AI Pilots to Production: Preparing the Enterprise
The article explores the challenges of transitioning AI from promising pilots to production systems that businesses can depend on. While technical capability matters, success requires addressing process ambiguity, change management, data quality, and security considerations.
Organizations must balance speed with risk management and recognize that the distance between an interesting use case and a production capability involves data, testing, integration, and adoption work.
The focus should shift from asking what organizations are doing with AI to whether they can reliably depend on it at scale.
Most organizations lack enforceable AI controls despite having policies, enabling shadow AI and security risks.
While 65% of enterprises report experiencing AI agent-related incidents and nearly half have tied data leaks to unauthorized generative AI use, the problem isn't missing policies but absent enforcement mechanisms.
Organizations struggle with visibility into shadow AI tools, fragmented governance ownership across departments, and insufficient decommissioning processes.
Effective AI governance requires prioritization of high-risk use cases, embedding controls into existing workflows, measurement through technical evaluation, and building on established processes rather than creating governance from scratch.
Vendors are moving from per-seat to outcome-based pricing, complicating renewals and contract negotiation.
Enterprise software vendors are shifting toward outcome-based pricing as agentic AI spreads, charging for completed tasks rather than seats. In principle this improves budget predictability, tying cost to measurable results instead of volatile token consumption.
In practice implementation is inconsistent, ranging from per-resolution charges to per-seat contracts carrying outcome guarantees.
Only 19% of service buyers currently use outcome-based arrangements, though analysts expect gradual growth through 2031 as vendors and CIOs work out contract terms and agree on performance metrics.
The cybersecurity pioneer discusses the evolution of the CISO role, AI's impact on careers, and why operational resilience is the profession's next frontier.
Charles Blauner, who served as CISO at JPMorgan, Citigroup, and Deutsche Bank after entering the field at the dawn of information security, is featured in Dark Reading's Heard It From a CISO video series. Blauner reflects on the influence of Steve Katz, regarded as the godfather of the CISO role, and explains how mentorship, collaboration, and a culture of paying it forward helped shape the profession.
He argues that today's security leaders must be more than technical experts, with resilience, communication, and business acumen increasingly defining successful CISOs.
A GPU cloud developer advocate argues AI teams are repeating cloud cost mistakes on hardware 10x more expensive per hour.
The author, who works in developer relations at a Silicon Valley GPU cloud provider, describes a call they receive weekly. It opens with praise: the AI feature shipped on time, users love it, engagement is up.
Then finance closes the quarter and the celebrated feature turns out to lose money on every request.
The article argues companies learned cloud cost discipline the hard way and AI teams now risk repeating those mistakes on hardware that can cost ten times more per hour. It sets out the unit-economics discipline needed before an AI feature reaches scale.
Cost per completed task, not headline token price, determines real AI spend - and most enterprises can't measure it.
Open-weight models have broken vendor pricing monopolies and shifted enterprise bargaining power, but simply switching to a cheaper model does not reduce actual cost.
The article stresses that cost per completed task diverges sharply from headline per-token pricing, because a chattier model consuming more tokens bills more at an identical rate.
Most organizations lack the foundational pieces - documented performance benchmarks on real production queries and compliance protocols - needed to make a defensible model choice. Genuine cost reduction requires internal evaluation harnesses and clear business-outcome metrics rather than price shopping.
The Rise of the AI Operating Executive
Marianne Johnson, EVP and chief product and technology officer at Cox Automotive, is pioneering an integrated operating model that unifies product, technology, data, AI, engineering, and cybersecurity under one leadership structure.
Consolidating these traditionally separate functions has accelerated the organization's ability to operationalize AI at scale and adapt to rapid technological shifts.
She emphasizes that future leaders need multidisciplinary skills, high-change management experience, and genuine AI fluency, while staying curious and hands-on.
As agents increasingly handle code generation, the bottleneck shifts from development to ideation and discovery, requiring critical thinkers who can identify high-impact opportunities.
Omnicom CIO Leif Maiorini on why five-year IT plans are obsolete and how agentic AI will flatten organizational structures.
Omnicom CIO Leif Maiorini argues traditional five-year IT planning cycles are obsolete, pointing out that transformative technologies like agentic AI emerged only two years ago.
He advocates maintaining forward-looking strategic planning groups that track emerging technologies from AI to life sciences and anticipate their business impact.
Maiorini expects agentic AI to reshape organizational structure, replacing hierarchical vertical communication with horizontal teams organized around specific applications.
The article also promotes the CIO 100 Leadership Live: New York conference where Maiorini will discuss the next 12 to 24 months of CIO challenges. Topics flagged include AI economics and organizational redesign.
Scaling AI beyond pilots requires redesigned operating models and embedded governance, not faster deployment.
Most organizations face a readiness gap between AI capability and their ability to implement it, with 83% citing data quality challenges and 74% struggling to demonstrate ROI.
Foundational work on data readiness, operating models, governance, and skills is routinely skipped in the rush to deploy.
Successful implementation requires close IT and business partnership with explicit clarity on objectives, since AI does not tolerate ambiguity about expected outcomes.
Organizations must tackle data debt, secure executive sponsorship, build AI literacy, and differentiate governance depending on whether AI supports desktop productivity or sits inside critical business processes.
Netskope's Mike Anderson explains why understanding agent use cases means hearing directly from the sales team.
IT Brew interviews Mike Anderson, chief digital and information officer at cybersecurity and networking company Netskope, who supports enterprise technology innovation for roughly 3,000 global employees.
His argument is that a CIO wanting to know how the sales team can use AI agents has to hear it from the sales team directly.
That is harder than it sounds at a company with thousands of employees. Back in May he met with the chief marketing officer, chief revenue officer and other function leaders to do exactly that. The interview covers what he has learned leading technology at companies of very different sizes.