Author

Louis Landry

Published

17 July 2026

Estimated Read time

5Mins

Louis Landry, Chief Technology Officer at Teradata explains why it’s an exciting time for enterprises who are leveraging agents to transform their processes.

Technological advances such as agentic AI have completely changed the way enterprises work for the better, but they have also posed several challenges. Businesses are not struggling with AI because the technology isn’t ready, but because the infrastructure beneath it isn’t.

Several demonstrations will have you believe that AI implementation is just another straightforward process, but this is not the case at all. Such presentations are impressive because the data is already polished, the queries are direct, and the answers are predictable, while the overall environment is controlled. But this is not the enterprise reality.

The typical enterprise runs multiple different SaaS applications, each with its own platform, data models and methods for showcasing the same business concepts. Not only that but they’re layered on top of millions of relationships and contextual rules that no demo was ever designed to handle. As a result of the complex state some enterprises’ systems are in, most of their AI adoption projects end up stalling or failing. This was also confirmed by last year’s MIT report which stated that 95% of businesses don’t see a return on their AI investment despite spending $30 to $40 billion.

So what are so many enterprises doing wrong and how can they revert it?

What’s hindering enterprises’ AI journey

Most AI models are impressive in isolation, but an enterprise environment doesn’t operate in siloes. Businesses have layered systems, interdependencies and nuanced rules that AI models will need to be able to support, otherwise they’ll fall short. Without the ability to act autonomously, they will not be able to provide meaningful change to the business.

With this in mind, a successful AI deployment needs proper orchestration. The 5% of enterprises that are winning this race coordinate their AI agents to own different areas within the business and come together through an integrated foundation instead of depending on a single model for everything. Successful enterprises see AI as a strategic differentiator, grounded in integrated, proprietary datasets that competitors can’t duplicate. They also ensure their systems are following the business objectives and operational limitations, as well as compliance requirements and competitive pressures.

AI autonomy in four levels

In terms of what successful enterprises have in common, they are all in what I refer to as the “AI autonomy journey” where they usually go through four different levels to build a robust foundation that transforms agentic AI deployments to capable and trusted advisors. Each business will achieve different levels of autonomy depending on how technologically advanced they are, their complexity and business value.

Looking into these levels in more detail, the first stage is where most businesses sit today. Their AI systems are taught to respond to basic interactions in natural language and collect data, though humans are still needed to validate the results.

The next level sees AI tools taking a more trusted analyst position. They start to comprehend context and intricacies while understanding connections between datasets. Such systems are able to recognise business terminologies and manage scenarios independently.

Here’s an example to help you visualise the two levels. An enterprise in the banking sector wants to understand its loan repayment rates across different customer segments and product types. A level one system will be able to retrieve and share the repayment figures with the exact criteria, while a level two system will not only gather the numbers, but will also flag which customer profiles are showing signs of risk and recommend next steps for the analyst. This stage goes beyond simply reporting figures, as it is able to translate complex financial patterns into clear and actionable intelligence.

Then we move on to level three, which is a significant turning point for an enterprise’s AI autonomy level. At this stage, the technology is able to make connections across your business and recommend actions with insights, including retention strategies, adjusting contract terms or raising potential risks.

And finally, at level four, we see AI stop advising and instead start acting. For example, if you prompt it to optimise customer retention while keeping margins at a certain point the system will autonomously identify at-risk customers, deploy the right strategies and track the results while adjusting its tactics in real-time. With that said, employees are still in control of defining the goals and setting the risk tolerance levels and compliance requirements.

The above levels reveal a clear progression. If level one is “prompt and prove” and level two is “prompt and trust,” then level three is “prompt, understand and suggest” and level four is “define, deploy, adapt”. Through this autonomous journey, the system gradually becomes an intelligent business analyst that actively improves how decisions are made, enabling the business to operate at a scale that simply wasn’t possible before.

The road to competitive advantage

As enterprises start exploring what their AI autonomy journey looks like, we will also see a gap forming between businesses at different stages. Companies moving on to higher levels of AI autonomy will gain real competitive advantage much faster than others. Having said that, achieving this isn’t as simple as a matter of deploying more powerful AI models. It requires a robust knowledge base that AI systems can actually depend on, otherwise they become unreliable and decisions get misaligned. This is where context engineering comes into play as a critical tool helping autonomous agents become more accurate and useful, and shaping how AI systems interact with the broader business. Context engineering also enables agents to perform complex tasks without worrying about errors coming up in the process.

It’s an exciting time for enterprises who are leveraging agents to transform their processes. The ones who are successful in their adoption will be those who understand early on that AI is only as powerful as the foundation it sits on. This means that it’s crucial to get the datasets, context and orchestration right from the beginning, to be able to progress with your AI autonomy journey and set yourself apart from competition.

Louis Landry, Chief Technology Officer, Teradata

  • Data & AI

Author

Sascha Giese

Published

Estimated Read time

5Mins

Sascha Giese, Tech Evangelist at SolarWinds talks to us about how to manage the transition from AI experimentation to delivery.

When Johnson & Johnson revealed a shift in its AI strategy last year, it signalled the end of a period of experimentation for the pharmaceutical giant and the beginning of a more disciplined, results-driven era.

At its peak, J&J employees were pursuing nearly 900 individual AI-related use cases. But as the WSJ reported at the time, this was a position that was simply unsustainable. Instead, the company decided to change tack and home in on high-value generative AI use cases in areas such as drug discovery and supply chains.

For many in the tech sector, this was a high-profile example of a business deciding to call time on pilots and trials in favour of a more structured approach where AI finally has to earn its keep.

Fast forward to the beginning of 2026, and the State of AI in the Enterprise report by the Deloitte AI Institute suggests that many companies are also following a similar path, with business leaders becoming impatient and wanting to see a return on their investment.

AI is shifting from a one-off investment to an ongoing operational expense

And it’s easy to see why. Amid the current AI boom, a recent report by Mavvrik, a US-based IT financial management platform, found that 80% of enterprises missed their AI infrastructure cost forecasts by more than 25%. It claimed that AI costs are “crushing margins” and that AI-related overheads are harder to manage than cloud costs because AI introduces new variables where “minor changes in usage can spike spend by 100x”.

It’s a good point. Unlike traditional software, which tends to follow predictable instructions, AI systems explore multiple possible paths to reach a solution. Sometimes that could mean resolving an issue in five minutes. At other times, it might mean AI systems retrying, looping and testing alternatives for hours, leading to additional expense.

But there are other unforeseen costs as well. The dash to embed AI can often expose shortcomings in existing IT infrastructure. Legacy systems and on-prem environments, for example, that might have been perfectly suited to a pre-AI world may now lack the necessary processing capacity required.

What rightsizing AI adoption really looks like

As a result, this may force organisations to bring forward plans to upgrade their technology. But at what cost? It’s not simply a case of writing a cheque and flicking a switch. In practice, it means being clear about the task AI is meant to improve, and defining measurable indicators and a baseline against which to judge impact.

But rolling out AI isn’t just about the technology. One of the lessons we’ve all learned over the last couple of years is that if workflows are unclear or the quality of data is poor, then AI will expose those shortcomings. Without a proper assessment of whether the company is ready for AI, such a project could quickly turn out to be unsuccessful, becoming a time and money sink.

Equally, if AI is embedded into structured environments – with defined responsibilities and feedback loops – then we also know that AI can genuinely improve outcomes.

For instance, a recent report by SolarWinds sought to understand how generative AI (GenAI) has been incorporated into ITSM (IT Service Management) workflows. It focused on those areas within helpdesks designed to reduce manual effort, such as automatically suggesting ticket responses, sourcing relevant knowledge base articles and generating incident summaries.

The report wanted to discover whether AI was genuinely making life easier for IT teams or merely adding another layer of complexity. The findings showed helpdesks saw a significant decrease in average incident resolution time after enabling the GenAI features.

On its own, that saving of just under five hours per ticket – or just under 20% of the time spent per support request – is impressive. But it’s only when you multiply that efficiency gain across the hundreds, if not thousands, of tickets generated each year that the true scale of AI’s impact becomes clear.

While the savings clearly point to GenAI’s benefits, the report also highlighted a common thread around that all-important pivot from pilot to operations. Instead of viewing GenAI as a test or side project, those teams that integrated AI tools into their daily service desk workflows appeared to have better outcomes.

“These results show what’s achievable when AI adoption is combined with effective change management and a focus on process improvements,” said the report. “They also act as a benchmark for other organisations evaluating the potential impact of similar tools,” it said.

Proper safeguards are key

But perhaps the biggest lesson learned from the last couple of years is that if we’re to truly maximise ROI, then we need to look at the bigger picture and develop AI systems in a much more methodical way. In effect, we need to develop AI by design. What does that mean? Well, in terms of privacy and security, it means establishing clear rules around not just the use of data but also how AI tools act and behave.

We also need to look beyond bias checks during model training to ensure that fairness is woven into AI right at the start, while ensuring that humans have the final say. And to build confidence in a new generation of tools, we need a proper paper trail to ensure that AI decisions can be traced and analysed so that, in the event something goes wrong, people can understand not simply what happened but why. It’s a good practice to ask an AI vendor which models are in use, where the training data came from, and what happens to the data users feed into the AI.

It’s too soon to say definitively whether Johnson & Johnson’s pivot signalled the end of AI’s pilot phase and the beginning of a more disciplined era. What we do know, though, is that the organisations most likely to succeed are the ones that are able to successfully manage the transition from AI experimentation to delivery while keeping a keen eye on outcomes and cost.

By Sascha Giese, Tech Evangelist at SolarWinds

  • AI in Procurement

Author

Asdrubal Picardo

Published

16 July 2026

Estimated Read time

5Mins

Asdrubal Picardo, CEO at Squalify, explains why cyber risk now sits alongside credit, liquidity and operational risk as a driver of enterprise value.

For most of its history, cyber risk was somebody else’s problem – at least, that’s how the boardroom saw it. It lived in IT, spoke in acronyms, and was measured in things boards neither understood nor needed to. Firewalls. Patch cycles. Vulnerability scores. Important, certainly, but not the kind of thing that kept a CFO awake.

That era is over. Cyber risk now sits alongside credit, liquidity and operational risk as a driver of enterprise value. It shapes valuations, influences M&A decisions, and moves shareholder confidence in ways that are very visible and very fast. When something goes seriously wrong, it isn’t logged as an IT incident; it lands in the boardroom as a business crisis with a price tag.

The problem is that most organisations are still trying to govern a 2025 risk with 2005 tools.

How we got here – and why the approach most organisations use isn’t fit for purpose

The discipline has evolved considerably over the past two decades. But that evolution has been uneven, and the methods many organisations still rely on were designed for an earlier set of problems.

Stage one: heat maps and gut feel

For the better part of two decades, cyber risk management was essentially a technical exercise dressed up as governance. Risks were mapped onto colour-coded grids based on control assessments and vulnerability scans. Teams asked whether the firewall was configured correctly, whether patches were current, whether access rights matched the policy.

There was nothing wrong with this, as far as it went. It gave security teams a shared framework. It helped prioritise remediation. But it told you almost nothing that a board could act on. What does “high” actually mean? High compared to what? What would it cost? The heat map couldn’t say. A risk rated red in one business unit and amber in another might represent wildly different financial exposures. Or identical ones. There was no way to tell, and no common unit of account to compare them.

The board got a dashboard. What it needed was an answer.

Stage two: financial quantification, built from the bottom up

The next wave addressed the most glaring gap: it put numbers on risk. Frameworks like FAIR gave analysts a structured way to estimate monetary losses, drawing on probability distributions and loss modelling. For the first time, cyber risk could be expressed in terms that the rest of the business understood. That matters, because when you can denominate risk in money, you can compare it to other risks, weigh it against the cost of controls, and start making defensible decisions.

But the bottom-up approach carried its own limitations, and they’re serious ones. It works at the asset level (individual systems, individual threats, individual control failures) and tries to work upward toward an enterprise view. In practice, this requires granular technical data that is often incomplete, unreliable, or simply unavailable. Building a single quantification can consume months. And the further you zoom out, the more the analysis starts to wobble: aggregating dozens of system-level assessments into a coherent company-level picture is an exercise in compounding assumptions.

The result is analysis that can feel rigorous in its detail while being strategically useless at the level where decisions actually get made. You end up knowing a great deal about the risk inside individual rooms while remaining largely ignorant about the building.

Stage three: start with the business, not the systems

The most recent evolution flips the logic entirely, and it’s the one that finally produces something boards can use.

Instead of starting with IT assets and working up, a top-down approach starts with the business: how does this organisation make money, what would a serious cyber event actually disrupt, and how exposed are those critical functions right now? It calibrates against real-world loss data drawn from insurance markets and large-scale incident histories, rather than from internal workshops and expert estimates. That means it captures what analysts might miss: unknown vulnerabilities, systemic exposures, and the full chain of second-order consequences that bottom-up models routinely undercount.

The outputs look different too. Risk expressed as a potential financial loss at the company or group level, broken down by business unit, tracked over time, and stress-tested against different investment scenarios, is something a board can actually govern. It fits into the same mental model as every other material risk on the register. It enables comparison, accountability, and decisions that can be explained and defended after the fact.

For groups operating across subsidiaries or jurisdictions, this matters even more. A consistent, top-down model makes it possible to compare entities on the same basis, set improvement targets proportionate to actual exposure, and track whether the aggregate risk position is moving in the right direction. Bottom-up methods, stitched together from incompatible local assessments, simply can’t do that.

What this demands of leadership

The shift creates clear obligations, and they run in both directions.

CISOs and CIOs need to stop thinking of financial fluency as someone else’s job. The ability to explain how a specific control investment reduces a measurable financial exposure (not just improves a risk rating) is now a core part of the role. Boards are increasingly asking for it, and those who can’t provide it are increasingly losing the argument for budget. Boards, for their part, should be asking harder questions. Not about the technical detail – that’s management’s job – but about the quality of the evidence behind the numbers. Does the reporting show how exposure has moved over time? Does it flag concentrations of risk? Are the assumptions visible and defensible? A report that can’t answer those questions isn’t risk governance; it’s risk theatre.

The standard has shifted

The principle that risk has to be measurable to be manageable isn’t new. It underpins how serious organisations have governed credit, liquidity and operational risk for decades. Entire disciplines were built on it: stress testing, capital modelling, scenario analysis. The assumption was always that if you can’t quantify an exposure, you can’t really manage it; you’re just hoping.

Cyber is now being held to that same standard. Regulators are demanding it. Insurers are pricing for it. And boards that have watched enough crises unfold in public are no longer willing to accept dashboards as a substitute for answers.

Cyber risk has earned its place at the table. The question is whether the people presenting it are ready to speak the language of everyone else in the room.

  • Cybersecurity

Author

Sean Evers

Published

15 July 2026

Estimated Read time

5Mins

Sean Evers, VP of Sales & Partner at Pipedrive explains why using AI effectively involves redesigning workflows around business problems, whilst maintaining a focus on the human experience.

AI appears to be everywhere, but clear results and a quantifiable ROI are not so common.

Many businesses are learning that simply adopting the latest tools doesn’t guarantee efficiency, innovation or growth. It’s a lesson that must be re-learned from time to time as the latest new tech paradigm comes upon the business world faster than the wisdom to make best use of them. And to be fair, faster than new technology itself becomes mature.

Too often in the past two years AI initiatives have been slower than desired in showing value because they’ve been treated as backend tech upgrades rather than strategic, human-centred transformations. An organisation is not a server rack, able to work at double the output when a new blade is inserted. Working with the grain of the business will always lead to better outcomes as the people and processes mix better with the changing variables being introduced.

Businesses can move beyond reflexive actions brought on by AI hype by embedding new technologies into problem-shaped workflows, creating meaningful impact through strategy, culture and human-centred implementation. AI tools are designed to solve specific challenges, not every challenge. We’re not at AGI yet (artificial general intelligence). Use tools wisely in limited areas, train teams well, and ensure a clearly communicated strategy supports team efforts. Then, all of a sudden, those investments in new tech will really move the needle on what’s been carefully defined as critically important.

Redesigning workflows

There is no best-in-class AI ‘Swiss army knife’. Even the most cutting-edge tech can’t deliver value unless workflows are redesigned around real business problems. AI succeeds when aligned to clear strategy, not when applied to broken processes like a sticking plaster. Firms positioning AI purely as an IT project will often encounter resistance, poor uptake and disappointing outcomes. The tech cycle has been this way forever, likely since fire and the wheel. Success always depends on employee understanding, willingness and trust in both their leadership and the ability of the tech to deliver on vendors’ promises.

Pipedrive’s ‘The evolving role of AI in sales workload management report’, highlighted how artificial intelligence has been reshaping sales roles and optimising workload distribution. In some areas, like in the sales function, the findings reveal that AI has become a co-pilot for professionals, enhancing efficiency and helping them focus on high-value activities:

  • One takeaway from the report was that AI users spend more time on strategic activities compared to non-users.
  • AI usage was still higher among sales managers (41%) than among salespeople (31%), indicating a need for more accessible AI tools and tailored training programmes.

Focus on what matters

AI centred on the human experience and business problems is what delivers real results. Thus, so-called soft skills and processes like corporate transparency, workplace empathy and ethical AI governance in practice are not really ‘soft’ considerations. Only a hard-headed, machine-minded person would fail to consider them as the essential enablers of organisational uptake and progress.

When employees are informed participants in change, AI becomes empowering rather than disruptive. That shift does not happen by accident. It requires leaders to articulate what problem AI is solving, for whom, and how success will be measured. Now, in many SMEs, the temptation is to deploy AI features because competitors are doing so, or because vendors promise transformational gains. But transformation without direction is simply noise.

Take a typical sales team in a growing business. Introducing AI-driven forecasting or automated lead scoring will not improve performance if the underlying pipeline stages are unclear or data hygiene is poor. In that scenario, AI merely accelerates existing inefficiencies. However, when leadership first defines what a qualified lead actually looks like, standardises processes and sets clear accountability, AI can enhance decision-making and free up time for higher-value conversations and really make a difference.

This is where cautiously redesigning workflows becomes critical. AI should be introduced at friction points like repetitive admin, inconsistent reporting, or slow handovers between teams. This is better than when added as a layer across every function. A targeted deployment creates quick wins, builds confidence and demonstrates measurable returns. Over time, those incremental gains compound and any cultural and technological lessons can be carefully applied in a virtuous cycle.

Managing the transition

The role of leadership is equally important. When AI is positioned as a cost-cutting mechanism employees will view it with suspicion. When it is brought in as a productivity partner to reduce manual tasks and allow individuals to focus on creative, relational or strategic work then adoption improves. Communication must be clear about what will change and what will not. Ambiguity will always breed a very understandable resistance.

Ethical considerations also need to move from policy into well-understood daily practice. For SMEs in particular, governance can feel like a burden reserved for larger enterprises who can ‘do it properly’. In reality, lightweight but explicit guardrails are often enough. Define what data AI tools can access. Establish human review points for high-impact decisions. Be transparent with customers about how AI is being used in interactions. These measures build trust internally and externally, which protects the long-term value of relationships and contracts.

There is also a mindset shift required. AI is not a one-off implementation. At the current rate of change it appears that it will be an evolving capability for many years. Teams will need space to repeatedly test, refine and learn. That may mean starting with a single department, gathering feedback and iterating before scaling. It may mean accepting that some pilots will fail. The objective is not perfection at launch, but steady alignment between technology and business requirements.

Being an SME can be advantageous

For SMEs, this disciplined approach can actually be a competitive advantage. Larger organisations often struggle with legacy systems and complex approval chains. Smaller businesses can be more agile, provided they resist the urge to chase every new feature release. Clarity of purpose becomes their differentiator.

Ultimately, fitting AI to the problem is about respecting the fundamentals of good management. Define the objective. Align people and process. Introduce technology where it adds measurable value. Review and refine. The companies seeing real returns from AI in 2026 are not those with the most tools, but those with the clearest strategy.

AI will continue to evolve rapidly. New capabilities will emerge, and expectations will rise. But the principle remains constant: technology works best when it amplifies well-designed human systems. Organisations that remember this will find AI soon stops being overwhelming when enabled as a controlled, practical tool for growth.

Sean Evers, VP of Sales & Partner, Pipedrive

  • Digital Strategy
  • People & Culture

Author

Paul Done

Published

14 July 2026

Estimated Read time

5Mins

Paul Done, Field CTO at MongoDB talks to us about why people who think ‘vibe coding’ is all just lazy, AI generated code, are missing the valuable contribution AI can make to high-quality coding.

Very few phrases in software have spread as quickly, or been as misunderstood, as “vibe coding”. For some, it signals the democratisation of software development, or the practise of causal programming for throw-away applications. But to its loudest critics, vibe coding represents a shortcut culture: developers tossing four‑line prompts into an AI model, then flooding production systems with fragile code. Many engineers also dislike the term because it implies coding without really understanding the code itself or blindly accepting AI output. In turn, some see vibe coding as trivialising the real effort and expertise required in software development.

I’ve been in the industry long enough to have heard similar complaints before: a new technology trend or tool that will make developers lazy, hollow out engineering skills, and ultimately leave organisations exposed. But that framing misses the point, because it has become a catch-all to describe two very different practises, with very different implications for developer teams and enterprises.

AI-driven coding is not vibe coding

Vibe coding often distracts and undermines the very real and positive impact of adding AI to the programming workflow. The leaps in the past few months alone have meant that AI agents have rapidly transformed how software is developed. Rather than manually writing code line by line, developers can now assign complex tasks to AI agents in plain English and have them independently research solutions, write code, debug errors, and deploy systems. These agents are capable of handling long, multi-step workflows – for example, setting up infrastructure, integrating tools, testing functionality, and documenting the results – often completing in minutes what previously took days.

As a result, programming is shifting from direct coding to orchestrating and supervising AI-driven processes, where developers focus more on defining problems, guiding agents, and reviewing outcomes, rather than implementing every technical detail themselves.

This style of development still requires a comprehensive understanding of engineering principles. As a result, it should not be confused or conflated with vibe coding, where code may also get generated rapidly, but lacks proper reviews, and is deployed with limited consideration for scalability, security or ownership. When critics warn about insecure “slop” code entering production, they are often pointing to failures of process and governance. That is a legitimate concern. But it is a concern about how teams operate, not about the inherent properties of AI-assisted development.

Even long-standing sceptics are beginning to acknowledge this distinction. Figures such as Linus Torvalds, Donald Knuth, and Robert C. Martin have recognised that AI can play a constructive role when it sits within disciplined engineering practice. The conversation is shifting away from whether AI should be used at all, and towards how it should be governed.

AI accelerates existing coding practises

Teams that have always kept strong review standards, testing rigour and capable leadership will not have those practises washed away if AI enters part of their workflow. For these teams, AI helps to increase the output without lowering standards. This is because it amplifies the culture already there.

The same principle applies to security. Secure software depends on threat modelling, dependency management, access controls and continuous monitoring. If those disciplines are weak, vulnerabilities will surface regardless of whether the first draft was written by a human or generated by a model. AI changes velocity, not responsibility.

For CIOs and technology leaders, this has practical implications. Prohibiting AI tools is unlikely to succeed and may drive their use underground – and given the advances in AI-driven coding, it may even amount to a serious competitive disadvantage. The more effective response is to strengthen the controls around them. Clear coding standards, automated testing, AI-driven code review, policy-driven code scanning, and runtime observability must become the default.

As AI accelerates the pace of software development, traditional human-centric code review workflows will struggle to keep up. When code can be generated much faster than it can be manually reviewed, human reviewers quickly become the bottleneck. The emerging model is one where AI systems enforce standards, review and validate code continuously, and flag risks at machine speed, while human engineers focus on defining requirements, architecture and higher-level oversight, rather than line-by-line review. AI belongs inside a structured software development lifecycle with guardrails, traceability and accountability.

A higher baseline for good code

With these in place, AI can raise the bar for what “good” looks like in our industry. Junior developers can receive immediate feedback on idiomatic patterns, security pitfalls and performance trade-offs. Senior engineers can offload repetitive tasks and focus on architectural coherence and long-term design. Test generation, documentation updates, refactoring and coding standards enforcement can become continuous activities rather than deferred clean-up exercises.

In many organisations, technical debt accumulates because improvement work competes with delivery deadlines. AI is reducing that tension by lowering the cost of maintaining standards. It can flag inconsistencies, suggest improvements and reinforce agreed conventions across large codebases.

This is particularly relevant for modern, data-intensive applications where distributed architectures and complex data models leave little margin for error. AI can support good engineering principles by improving validation, surfacing edge cases and increasing visibility into how systems evolve.

The debate around vibe coding muddies the waters about AI in our industry – the question is not whether AI will be part of software development. It already is. The differentiator will be how seriously organisations treat governance, architecture and long-term maintainability in an era of accelerated output.

By Paul Done, Field CTO at MongoDB

  • Data & AI

Author

Markus Nispel

Published

13 July 2026

Estimated Read time

5Mins

Markus Nispel, Head of AI Engineering & EMEA CTO at Extreme Networks explains why your strategy and budget aren’t holding up your AI network initiatives – your execution is.

The pilot phase is over. A third of companies have now scaled AI into production. For everybody else, the holdup isn’t strategy or budget – it’s execution.

Many AI initiatives struggle to deliver meaningful ROI for familiar reasons: missed business objectives and clearly defined KPIs, limited AI expertise, processes that haven’t been redesigned for AI, gaps in AI and data governance, and persistent challenges around data availability and siloes, quality, and preparation.

At the same time, supporting AI at scale requires infrastructure that can keep pace with continuous data flows, shifting workloads, and real-time decision-making. Traditional networks were built for stable, predictable workloads – not the speed, volume, and variability that AI-driven systems demand.

As AI moves from experimentation into production, networks must handle continuous data flows, adapt instantly to shifting demand, and operate at a scale that manual intervention simply cannot match. Meeting these demands requires a shift in network design and operation. This is where AI-powered, autonomous networking becomes essential.

Autonomous networking in practice

Autonomous networking allows systems to detect disruptions early, adjust traffic flows and configuration dynamically, and continuously optimise performance in real time, rather than reacting after issues occur.

This is made possible through multi-agent systems. Individual agents manage narrow tasks such as bandwidth monitoring, anomaly detection, and routing optimisation. Others sit above them, orchestrating them, interpreting user and business needs, and pulling the right capabilities together.

The system also works within defined governance frameworks. Humans stay involved in critical decisions, with full visibility, transparency and explainability into system activity, planning, reasoning and decision-making. Every automated action is logged and can be traced, verified and audited if necessary.

Trust, as in real life, is built through repeated positive outcomes. When networks keep making the right decisions, teams gain confidence, reduce oversight and can focus on higher-value work rather than constant troubleshooting.

This transition is already underway. Recent data reveals that 88% of organisations already rely on multiple AI-powered tools for networking and security operations. For stretched-thin teams, autonomous networking has grown into a necessity. Meanwhile, the strain is evident, with 92% of leaders reporting that AI places heavier demands on bandwidth and computing resources.

In this environment, network performance becomes a defining factor in whether AI systems can operate reliably and deliver measurable value.

Security through intelligent boundaries

Scaling AI also changes the security landscape. Networks are no longer connecting only users and devices; they are supporting a growing number of non-human identities. Agents across finance, marketing, engineering and operations interact with systems and data, often without human involvement.

Traditional security models were not designed for this level of autonomy. Access can no longer be broad or static. It must be tightly scoped, context-aware, and continuously verified. Every action must be visible and traceable.

But confidence is climbing. In fact, 93% of executives now say AI-powered networking reduces security risk rather than increasing it. That shift matters. It shows organisations are moving past scepticism and starting to take control.

That doesn’t mean we should be giving agents free rein. It means tightening the boundaries around what they’re allowed to do and enforcing strict agent governance. Access can’t be broad or permanent. It has to be specific, context-aware, and continuously verified. Every action should be traceable through a clear audit trail as AI for networking is implemented.

Where fabric architecture makes the difference

Networking for AI will not work without visibility. Real-time traffic monitoring and system connection mapping make enforcement possible. From there, segmentation becomes essential to contain risk and prevent unintended access or lateral movement – especially in the agentic era we have entered.

Fabric architecture enables this level of control at scale. By creating a unified network environment, it ensures that policies are applied consistently across all users, devices, and workloads.

This allows organisations to automatically isolate sensitive systems and tightly control how entities interact. If an AI agent behaves unexpectedly or is compromised, its access can be contained immediately, preventing wider impact.

Organisations that get this right bring identity, access control, visibility and threat detection together into a single, cohesive system. Autonomy can then operate within clearly defined boundaries, giving IT teams the confidence to guide, verify and trust every action the network takes.

Real-world application

The impact of this shift is visible across industries. Take retail in 2026. AI allows hyper-personalised shopping through cameras, IoT sensors, mobile apps and edge devices. Inventory data, customer preferences and behavioural patterns flow nonstop. Shelf labels, RFID systems and automated checkouts all require connectivity.

Healthcare follows similar patterns. AI supports diagnostics and robotic-assisted procedures, with patient data moving continuously across EMR systems, monitoring devices and analysis platforms. Without well-designed enterprise networks and access controls, those same systems can become exposure points rather than protection.

Manufacturing uses AI to predict equipment failures, optimise production lines and coordinate autonomous robots as sensors and machinery continuously communicate.

Across industries, without unified enterprise networks, stringent access control and defined human checkpoints, a single misconfigured AI agent could compromise sensitive data, disrupt operations or even endanger lives.

From experimentation to production

The era of experimentation is over. Organisations that operationalise AI effectively move faster, operate more securely, and respond to change with greater agility. Achieving this requires networks that can anticipate demand, enforce boundaries, and adapt continuously without compromising performance.

When that foundation is in place, AI can deliver on its promise – reducing operational friction, minimising disruption, and enabling teams to focus on innovation rather than maintenance.

That’s what 2026 demands. Are you ready?

By Markus Nispel, Head of AI Engineering & EMEA CTO, Extreme Networks

  • Data & AI

Author

Taran Rai

Published

10 July 2026

Estimated Read time

5Mins

Taran Rai, Corporate Sustainability Manager at Epson explains why the digital vs print sustainability debate isn’t as straight forward as it may seem.

The sustainability debate around digital vs print has become deeply polarised, with digital widely perceived as the environmentally responsible approach and print positioned as inherently and inescapably wasteful.

But is it as simple as that? The short answer to that question is no, particularly as organisations scale their use of digital technologies (some exponentially) and the environmental impact of these activities becomes apparent.

Arguably, the most striking and alarming example of this trend is the dramatic growth of AI and its associated digital infrastructure. If predictions about the sector’s resource demands are even remotely accurate, the world faces a significant and sustained increase in consumption.

According to the International Energy Agency (IEA), for example, driven largely by AI growth, “global electricity consumption for data centres is projected to double to reach around 945 TWh by 2030”. To put this in context, from 2024 to 2030, data centre electricity consumption will grow by around 15% per year, “more than four times faster than the growth of total electricity consumption from all other sectors.”

Increases in water consumption are also raising serious concerns. According to the UK government, “AI is predicted to lead to an increase in global water usage from 1.1bn to 6.6bn cubic metres by 2027. This is equivalent to more than half of the UK’s total water usage.”

As the study goes on to point out, “The water demand of AI technologies is likely to threaten global and national water security, especially in areas of existing water stress, which can in turn threaten the biodiversity of local areas and the needs of human populations.”

Digital good, print bad?

But what does this mean for print? After all, the prevailing assumption is that digital would enable organisations to phase out print and, by definition, improve their environmental performance. But given the enormous sustainability challenges now facing the global digitalisation movement, the comparison between digital and print needs to be reframed, not as a question of substitution, but as a question of relative impact in context.

Indeed, if we go back a couple of decades to the early days of digitalisation, reducing reliance on paper was a central part of the case for change. Many of those arguments were valid, particularly given the inefficiencies in how organisations used print at the time. Something had to change, and it did.

Fast forward to 2026, and the landscape looks very different. Print is no longer defined by the inefficiencies that once characterised it, as advances in production processes and technology have significantly improved its environmental performance.

For example, improved production processes have reduced energy consumption and streamlined workflows by eliminating stages that previously produced excess material, especially for limited print runs. At the same time, the ability to produce on demand enables organisations to align output more closely with actual requirements, helping to reduce overproduction and unnecessary inventory.

More specifically, digital and inkjet-based print processes can significantly reduce water consumption, with studies showing reductions of 50–90% compared to traditional techniques. And, in some industrial applications, such as textile printing, digital methods can reduce water use even further, with estimates of up to 95% savings due to the removal of washing and post-processing stages.

Striking a better balance

The underlying issue is not which format is “better”, but that digital and print have different environmental impact profiles. Print impact is typically more visible and often concentrated at the point of production, whereas digital impact is less visible but continues over time through ongoing energy use.

Moving from print to digital does not eliminate environmental impact; it shifts it to other parts of the value chain. As digital usage scales, particularly with always-on services and data-intensive applications, this ongoing impact becomes more significant. As a result, defaulting to digital-first strategies can lead to environmental costs that are not always fully understood.

Rather than focusing on format alone to determine sustainability strategy and benefits, organisations instead need to consider how different approaches perform in specific use cases.

In some scenarios, for example, the continuous energy demands of digital delivery may outweigh the one-off impact of print, particularly where information is accessed repeatedly or stored over long periods. In others, digital will clearly offer advantages, especially where distribution scale or accessibility is the primary requirement.

Whatever situation applies, the key to making good sustainability choices is to recognise that outcomes are shaped by how information is delivered and used over time, not by the medium itself. This shift in thinking also aligns with broader moves towards lifecycle-based assessment, where environmental impact is evaluated across the full span of use rather than at a single point in time.

The approach is already being formalised, with the EU’s Ecodesign for Sustainable Products Regulation (ESPR) establishing a framework law that sets strict rules on sustainability, durability, and repairability for goods sold in the EU to promote a circular economy. EU rules also require that large companies publish regular reports on the social and environmental risks they face, and on how their activities impact people and the environment.

The underlying point is that the core arguments that form the digital/print debate are now much more nuanced than they were 10 or 20 years ago. Organisations need to strike the right balance based on their operational needs and sustainability obligations, not just to protect the environment but also to identify the most business-efficient processes in which digital and/or print must play a role.

By Taran Rai, Corporate Sustainability Manager, Epson

  • Sustainable Procurement

Author

Rob Demain, CEO, e2e-assure

Published

9 July 2026

Estimated Read time

5Mins

Rob Demain, CEO of e2e‑assure, explores whether critical national infrastructure (CNI) providers are truly ready to face today’s cyber‑attack threats.

Many critical national infrastructure (CNI) operators lack the ability to protect their infrastructure despite the UK being subjected to daily sub-threshold cyber attacks, according to the Strategic Defence Review 2025. It’s a situation that the Network and Information Systems (NIS) regulations, introduced back in 2018, sought to prevent. But since its inception, just over half of the operators of essential services have updated or strengthened their existing policies and processes, leaving many woefully unprotected.

In desperate need of reform, NIS is set to be superseded by the Cyber Security and Resilience Bill (CSRB), which is expected to become law later this year, at which point a consultation on implementation proposals will commence, followed by secondary legislation and an adjustment period for stakeholders. The bill will broaden the scope to include other organisations deemed critical to the national economy i.e. data centres, Managed Service Providers (MSPs) and critical suppliers. Plus, the government reserves the right to extend those categories still further as part of its ‘future proofing’, which will enable changes to be made to the act to accommodate emerging threats and potential targets.

New demands

All of these new entities will need to comply with the Cyber Assessment Framework (CAF), which lays out expected cybersecurity and resilience outcomes. First published in 2018 to support the NIS, it has undergone a number of revisions since, with v4.0 released in August 2025. This version places a far greater emphasis on proactive security and decision making based on real threat intelligence. As well as adding new contributing outcomes on understanding threats and secure software development and support, it also expands the sections on security monitoring and response and recovery, while an entirely new category has been added on threat hunting.

All of this points to a far greater emphasis on being able to demonstrate assurance and proactively monitor all aspects of CNI infrastructure and that means more scrutiny of both IT and Operational Technology (OT). Until recently, securing OT wasn’t seen as a priority. These systems were chiefly concerned with maintaining system availability and minimising downtime. But their increased integration with IT systems to connect with the industrial Internet of Things (IoT) and deliver real-time monitoring, for example, are exposing these systems to attack, with threat actors able to move laterally from one environment to the other.

The threat posed by IT/OT convergence is well known, but it continues to be the Achilles heel of CNI, as revealed by the Volt Typhoon attack. This saw Chinese nation state actors maintain persistence across CNI in the USA since at least 2021, through the use of Living off the Land techniques, illustrating just how insidious and sustained these attacks can be.

Securing IT/OT systems

It’s these types of threat the CAF addresses through its risk and asset management requirements. Organisations must risk assess systems with respect to their dependencies and interactions with other systems such as IT/OT, and document and understand those dependencies. But other complementary frameworks can also be used to map IT/OT systems security, such as the ISMS within ISO27001 from an IT perspective and IEC 62443 from the OT side, in addition to ISO/IEC 27019 for process control systems.

Being able to follow these frameworks will require organisations to increase their security monitoring of both IT and OT and the transparency of their processes. They will need to transition from being reactive to proactive, and become resilient and risk informed, which will mean many will have to change their approach. These are really the only options available to them in this respect if they are to move the resilience needle.

The first is to decentralise and harden OT systems while keeping them segregated from IT. However, hardening alone can’t keep pace with digital transformation. Many OT assets cannot support multi-factor authentication (MFA) or accommodate rapid patching because they are downtime sensitive. So, surface hardening alone won’t confer the resilience needed long term.

The second option is to manage IT and OT together by giving everything an identity in a converged environment, but to do that you need to move the monitoring of OT into the Security Operations Centre (SOC). Centralised monitoring allows threats to be detected across both IT and OT networks, for teams to monitor east-west traffic, and to correlate alerts that might otherwise appear unrelated. And it’s this centralised management that will provide the visibility and control needed to improve IT/OT resilience.

Converged security

Such a converged SOC doesn’t just offer continuous visibility over IT, industrial control systems (ICS), OT and cloud environs, but also the real-time triage of critical alerts. These might include unauthorised PLC logic changes, unsafe set-point writes, abnormal OT protocol behaviour, lateral movement in ICS DMZs, OT malware, or unauthorised remote access into OT environments. These alerts are then grouped by operational impact, such as whether they present a safety critical risk or could lead to service degradation, so that they can be prioritised. Weekly threat hunts and detection surface validation over distributed environments provide the threat hunting capabilities needed to meet the CAF requirements and the SOC evidences, and provides that all important audit-ready compliance mapping to meet the demands of other frameworks too, such as IEC 62443, and ISO/IEC 27019.

Whether standing up a converged SOC internally or outsourcing, this capability is the most efficient way to adapt to the tightening regulations, particularly as we can expect ‘future proofing’ to lead to yet more demands. The emphasis is now firmly focused on the proactive monitoring of both IT and OT systems together, given their growing dependencies, so it makes sense for those organisations in scope – as well as those who could soon be – to begin to move their OT monitoring from the plant and into the all-seeing, all-knowing enclave of the SOC.

By Rob Demain, CEO, e2e-assure

  • Cybersecurity
  • Infrastructure & Cloud

Author

Satish Thiagarajan

Published

8 July 2026

Estimated Read time

4Mins

Satish Thiagarajan, founder of Brysa, a Salesforce and data consultancy based in the UK, explains why high quality data is the foundation for using AI successfully.

AI is everywhere in business right now, and for good reason. It offers the potential for better visibility across operations, fewer surprises, and more efficient use of resources. Most businesses have run a pilot to test that. Fewer have made it past one.

The stall point is consistent. Projects start well, attract investment, generate interest, and then stop delivering. The technology rarely gets the blame internally, but it tends to take it publicly. The real problem is almost always the data behind it.

Why data decides whether AI works

Business data systems weren’t designed to tell the whole story. They were each added over time to solve specific problems. One for finance, one for sales, another for HR, and a few more for marketing, support, and operations. Individually, they do the job for which they were intended, but they don’t really connect. And that’s fine, until you try to apply AI.

AI doesn’t think in terms of individual systems. It looks for patterns across everything, from customers and revenue to pipeline activity and service history. When that information is split across systems that don’t line up, those patterns break down, and what you’re left with is data that doesn’t quite agree with itself: records that don’t match, account data that conflicts with billing history, customer activity that looks different depending on where you check.

At that point, the problem isn’t a lack of data. It’s that the data doesn’t hold together. And if the data doesn’t hold together, the output won’t either.

What “AI-ready” actually looks like

Before AI can do anything useful, the data has to make sense on its own. Without that, AI is, effectively, guessing.

Context matters just as much. Data only becomes useful when it carries the relationships behind it. A customer record on its own doesn’t say much. Connect it to purchases, support history, and engagement activity, and you start to see why things happened, and that’s what AI needs to work with.

Timing forms another pressure point. Business moves quickly, but the data often doesn’t. If updates are delayed, or stitched together after the fact, AI will always be working from an outdated version of events. When data flows in real time, it reflects what’s actually happening, not what happened last week.

Then there’s continuity. When past performance connects directly to current activity, AI has something to learn from. Without that link, every decision starts from scratch. To get to that point, data needs to be consolidated in one place, and that’s where a CRM comes in.

Bringing it together in one place

Used properly, a CRM becomes more than a system for managing contacts. It acts as a central hub for accounts, opportunities, cases, campaigns, and service activity. AI doesn’t need perfectly clean data, but it does need to understand how work actually happens, and a CRM gives it that. Business data is constantly changing, and it needs to scale across teams, regions, and functions. Role-based access, audit trails, and clear permissions aren’t just nice-to-haves, they keep data usable as it grows.

Building a usable data foundation

The goal when faced with fragmented systems isn’t to replace everything, but to connect what’s already there. Customer data, comprising account histories, contact records, and interaction activity, is standardised and can be used across sales, marketing, and service, reducing duplication and avoiding inconsistency.

Workflows matter as well. If data quality depends on someone fixing issues later, it won’t hold. When validation happens at the point of entry, and updates flow automatically across systems, accuracy is built in, rather than an afterthought. When the data is reliable and connected, AI can support things like demand forecasting, lead prioritisation, or churn prediction, and good governance ties it all together.

AI doesn’t fail because the models aren’t capable. When the underlying data is fragmented, AI reflects that fragmentation. When the data is coherent and connected, AI has something solid to work with. That’s the difference between systems that generate outputs and systems that actually support decisions.

Satish Thiagarajan is the founder of Brysa, a Salesforce and data consultancy based in the UK. His company advises media, industrial, and services clients on using Data Cloud and Agentforce to turn signals into action. His work focuses on closing the loop between insight and execution in sales, marketing, and service.

  • Data & AI

Author

Chris Derham

Published

7 July 2026

Estimated Read time

5Mins

Chris Derham, Business Development Director at Alcatel-Lucent Enterprise talks to us about the buildings of the future and how they can help us reach net zero.

Buildings and their operations sit among the biggest climate challenges of our era, but they also offer one of the most impactful levers for emissions reduction.

When we look at the UK, data shows that direct and indirect emissions from the buildings sector in the UK account for 27.7% and 9.1% of total energy-related CO2 emissions, respectively. Significantly, per capita emissions from the buildings sector in the UK are 1.2 times the G20 average, suggesting that there is an opportunity to reduce this impact.

If net zero is the destination, effective management of the built environment is one of the most important routes to get there.

From gadgets to digital ecosystems

For years, “smart building” meant a collection of clever devices, from automated blinds to motion sensors. Today, that model is outdated.

Modern smart buildings function more like digital organisms. Instead of isolated systems operating independently, a central orchestration layer brings them together, aggregating and coordinating multiple sub-systems through a unified digital framework. Lighting, HVAC, occupancy monitoring, and security systems now communicate through shared data frameworks, adjusting in real time to how a space is actually used.

Artificial intelligence sits at the core of this shift as it can analyse usage patterns, forecast demand, and make autonomous adjustments. For example, a meeting room that typically fills at 10 am can be air-conditioned just in time, or lights on an underused floor of the building can be powered down before energy is wasted.

None of this works without high-performance connectivity. Advanced networking technologies such as Wi-Fi 7 and private 5G are being deployed to ensure low latency, high capacity, and airtight security. These networks form the invisible nervous system of the building, carrying the data streams that enable continuous optimisation.

Across the Channel

In many EU countries, smart buildings are increasingly treated as essential infrastructure, rather than an optional innovation.

Countries such as France and Germany have embedded automation and monitoring requirements into climate legislation, and digital capability is becoming a matter of compliance, not just a competitive differentiator.

At the centre of this transformation is the Energy Performance of Buildings Directive. This legislation requires all new buildings to be zero-emission by 2030 and existing stock to follow by 2050. Achieving that at scale without digital systems that continuously measure and optimise performance would be nearly impossible, meaning that as member states translate the directive into national law from 2026 onwards, investment is expected to intensify.

France provides a concrete example through its Decree BACS, mandating building automation and control systems in larger non-residential properties. By 2027, any such building with output above 70kW must comply with the legislation.

Progress in the UK

The UK presents a more fragmented picture. Innovation is not the problem. Manchester’s Triangulum initiative is a standout example of smart, low-carbon urban development. By integrating energy-efficient technologies, IoT sensors, and renewable energy systems, it demonstrates how coordinated digital infrastructure can reduce emissions and create more sustainable, liveable city spaces.

The difference lies in consistency. The UK lacks a comprehensive, national framework that mandates or systematically incentivises smart building adoption at scale. Instead, progress tends to occur on a project-by-project basis.

One lesson from continental Europe is that well-designed regulation does not stifle innovation, it often accelerates it.

A strategic framework in the UK could:

• Define common standards for smart buildings and energy efficiency
• Support interoperability across regions and technologies
• Encourage investment in digital infrastructure upgrades

The next opportunity lies in coordination. A national standard for smart buildings and smart cities could provide clarity for investors and developers, while also allowing for local flexibility.

The UK’s move towards net-zero

The good news is that although the UK still lacks an overarching framework on smart buildings and cities, many existing regulations are being adapted and modernised to help reach the goal of net-zero by 2030.

For example, the Minimum Energy Efficiency Standards (MEES) currently require commercial and private rented domestic properties to have an Energy Performance Certificate (EPC) rating of at least E. The government is currently consulting on proposals to raise this standard, which would potentially require all rented commercial buildings to achieve EPC B by 2030 to remain legally lettable.

In practice, this means significant upgrades for the UK’s commercial building stock. Investment in building fabric materials, LED lighting and controls, HVAC replacement, building management systems, and smart energy monitoring will be instrumental in achieving this standard and reducing environmental impact.

Other regulatory changes, such as reform of the Energy Performance of Buildings (EPB) Regime, will also play a key role when it comes to improved data collection, new EPC metrics that better reflect operational performance, and quality assurance for EPC assessments.

Taken together, this legislation is likely to require the installation of smart technologies across UK buildings.

Smart buildings: An important opportunity

Smart buildings are a powerful tool for cutting carbon emissions. By using real-time monitoring, automation, and AI, they reduce energy waste and optimise heating, cooling, and lighting. Beyond efficiency, they provide the data and control needed to achieve large-scale emissions reductions.

Effective regulation is key to driving adoption and creating a low-carbon, sustainable built environment. If the UK is to accelerate toward net zero, smart buildings must move from isolated examples of best practice to the standard for how we design, operate, and upgrade our spaces.

Chris Derham, Business Development Director, Networks Division, Alcatel-Lucent Enterprise

  • Data & AI
  • Sustainability Technology