Last month, we attended the 2026 Gartner IT Symposium/Xpo™ on the Gold Coast, joining technology leaders and analysts to explore some of the big forces shaping the next few years of technology and business.

One session that particularly caught our attention was Signature Series: Top Strategic Predictions for 2027 and Beyond, presented by Daryl Plummer, Distinguished VP Analyst and Gartner Fellow. Drawing on Gartner’s research, Plummer took us through ten predictions spanning physical AI, autonomous agents, the economics of AI, governance, energy and the changing nature of software.

While many of the predictions look towards 2029 and 2030, the decisions behind them are already beginning to take shape. Across these predictions, we also noticed three broader themes emerge.

Physical AI will create a new human and machine relationship

Gartner predicts that by 2030, 80% of frontline workers employed by international companies will be assisted by physical AI systems. That means AI moving beyond software and into the physical environment through robots, drones, autonomous vehicles and other systems able to sense and interact with the world around them.

For workers, that could mean delegating repetitive or hazardous tasks to machines, using robots and drones to monitor safety, or applying physical AI to turn large volumes of operational data into something people can act on.

The emphasis, however, is on assistance rather than straightforward workforce replacement. Gartner recommends taking a safety-first approach and investing in scalable platforms, governance and the skills required to manage these systems. The session also highlighted modular control policies, edge processing and virtual simulation as areas organisations should begin considering as physical AI matures.

Agents will be built to pursue government benefits at massive scale

Gartner also predicts that 2030 will be the year we see more than 10 billion autonomous agents created by people, companies and governments clogging public services.

Why? Because agents will increasingly be able to identify opportunities, determine eligibility and submit applications or claims with very little human intervention. Processes designed around people completing forms or navigating administrative hurdles could suddenly be confronted with automated agents doing the same thing at enormous scale.

That presents a clear infrastructure challenge, but also raises questions around identity, authority and fairness.

Governments will need stronger ways to establish whether an agent is authorised to act, whether a claim is legitimate and how automated demand should be handled. Gartner recommends modernising digital infrastructure and verification capabilities, with automated response systems likely to become part of the answer.

Cost exhaustion attacks will become a critical threat to AI systems

AI introduces new costs. And Gartner believes those costs could become a new attack surface.

By 2030, they predict that 80% of organisations with public-facing AI will have experienced a cost exhaustion attack, where malicious activity deliberately drives excessive AI usage and creates significant operational expense.

That changes how organisations need to think about token consumption. Monitoring it is no longer simply a matter of keeping internal AI budgets under control. Unexpected consumption could also be a security signal.

Gartner recommends treating token costs as a cybersecurity indicator, supported by measures such as token-rate monitoring, rate limits and security monitoring that extends across AI technologies. As more AI services become customer-facing, organisations will need visibility into both who is using them and what that usage is costing.

Business value and the rise of token governance roles

By 2029, Gartner foresees that 60% of organisations deploying AI will establish a dedicated function responsible for mapping the total cost of AI to value or profit.

The challenge is that AI costs do not always behave like conventional technology costs. An agent can repeatedly retry a task. A growing context window may require a model to process more information every time it responds. Organisations can also end up using expensive frontier models for jobs that could have been handled by something simpler.

The result is a growing need for token governance that connects consumption to the value being created. Gartner recommends linking token use directly to business metrics, supported by quotas, rate limits and clearer governance. Cutting AI expenditure indiscriminately misses the point too. The objective is understanding which costs are creating value and which are simply accumulating.

Token consumption and AI FinOps control

The need for visibility leads into Gartner’s next prediction.

By 2028, 60% of Global 500 companies will embed AI FinOps controls at inference, moving cost governance away from retrospective reporting and towards real-time optimisation.

Instead of discovering at the end of the month how much a particular AI workload consumed, organisations will increasingly need to understand and control spend as inference happens. That could make measures such as cost per task and token efficiency much more prominent in the way AI platforms are managed.

Gartner recommends instrumenting the inference path now, with runtime controls and telemetry built into AI platforms and applications. This becomes increasingly important as agentic workloads grow, because autonomous systems can generate costs at a speed that makes after-the-fact reporting much less useful.

Greater accountability around AI

As AI takes on a greater role in business processes and decisions, accountability becomes harder to leave undefined.

By 2030, Gartner predicts that 80% of Global 500 companies will contractually make their CIO or CAIO the “Evidence Custodian” for AI accountability.

The concept reflects a growing need to demonstrate what an AI system did, why controls were in place and who was responsible for ensuring those controls were followed. For large organisations, the consequences of an AI failure can extend well beyond the technology team.

Gartner recommends assessing whether existing digital evidence management systems can support these requirements. The larger point is straightforward: as AI becomes embedded in important decisions, organisations will need clearer records, more transparent guardrails and named accountability for its actions.

AI liability insurance as a driver of governance investment

Regulation may not be the only force pushing organisations towards stronger AI governance.

Gartner predicts that by 2030, insurers rather than regulators will drive AI governance, with strict underwriting requirements for AI liability insurance influencing the controls organisations put in place.

The session drew a distinction between traditional governance, risk and compliance practices and the governance AI systems increasingly require. Having policies on paper will not necessarily demonstrate that controls are working when an AI model or agent is running.

Instead, Gartner expects greater emphasis on operational and runtime governance, with technical controls embedded into AI systems and workflows. Organisations may therefore find that investment in AI governance platforms and other controls is driven not only by compliance requirements, but by the practical cost of insuring AI-related risk.

Energy production and private power providers

The growth of AI is also placing new pressure on something much more tangible: energy.

Gartner predicts that by 2030, $10 trillion in enterprise-owned energy will make Global 2000 organisations unexpected power providers, selling energy to grids and AI data centres and changing the traditional relationship between enterprises and utilities.

Rising demand from AI and hyperscale data centres is one driver. At the same time, organisations are investing in their own renewable generation, storage and energy management capabilities to improve resilience, reduce costs and meet sustainability objectives.

Gartner expects energy to become a more strategic and software-defined asset as a result. That means integrating energy and operational data, using automation to make real-time decisions and treating energy management as more than a facilities concern.

Burner business and a growing competitive gap

The traditional fast-follower strategy assumes there is time to watch what works, learn from the market and move once somebody else has proven the opportunity. Gartner believes AI could make that harder.

By 2029, Gartner predicts that 25% of Global 500 companies will continuously innovate componentised, AI-powered offerings, creating a competitive position that makes conventional fast-follower strategies increasingly difficult.

AI-native competitors can potentially develop, adapt and personalise offerings much faster. For established organisations, that places greater importance on the underlying foundations that make continuous innovation possible, including data, governance and talent.

Gartner’s recommendation is to look beyond AI as an internal efficiency tool. Customer-facing and front-office opportunities matter too, particularly where AI can change the product or service itself rather than simply make an existing process cheaper.

Disposable apps for a new software lifecycle

Finally, Gartner sees the way we build and maintain software changing significantly.

By 2029, it predicts that 80% of new applications will be intentionally disposable and used for less than one year.

As AI makes software development faster and more accessible, employees may be able to create applications for a particular project, process or temporary need and discard them when that need disappears. In that environment, continually upgrading and maintaining every application for years may no longer make sense.

But disposable does not mean consequence-free. Short-lived applications can still access sensitive information, influence decisions and create security or compliance obligations.

Gartner recommends risk-based governance, automated ways to keep track of business-created applications and updated records policies. Organisations will also need to think about the cost and speed of safely retiring software, not simply the speed at which it can be created.