Perforce CTO forecasts rise of context & agentic AI

Artificial intelligence will reshape software development, workplace roles and consumer technology over the next several years, according to new forecasts from Perforce Chief Technology Officer Rod Cope, who expects rapid advances in "context engineering", agent-based systems and ambient AI, alongside rising security and governance risks.

Cope argues that a new discipline of context engineering will replace today's trial-and-error prompt engineering for complex AI work. He links this shift with a wider transformation in how software is built and maintained, and how technical teams are structured.

From prompts to context

Context engineering involves designing the information, tools and connections around an AI model rather than relying mainly on phrasing a prompt. Cope expects this approach to deliver more accurate results for demanding tasks.

The method includes choosing an appropriate model, managing token limits and linking the model to relevant data sources, applications and systems. Cope also expects organisations to assemble "AI teams" made up of multiple agents that adopt different personas and tackle separate parts of a problem.

He warns that overloading systems with data will create new issues. Feeding agents too much context raises costs and can degrade performance, as models struggle with long inputs and produce inconsistent answers. Cope predicts that human specialists will need to build specific skills in context design in much the same way they have learned prompt techniques.

'AI workslop' risk

Cope highlights what he calls "AI workslop", where staff copy and paste unverified AI output through workflows. He expects this pattern to grow as more tools integrate generative models.

He says errors that are not caught early will push rework downstream or leak into production systems. Staff without deep domain expertise may struggle to tell useful AI results from flawed ones. Cope believes this will increase the importance of senior engineers and other experienced employees who can combine subject knowledge with curated data.

Agentic AI and MCP

Perforce's CTO sees agentic AI and the emerging Model Context Protocol, or MCP, as a major shift in software integration. He contrasts traditional API work, which can stretch from hours to months, with a future in which developers issue natural language instructions to an AI that discovers APIs, wires them together and handles changes over time.

In this model, AI agents act as intermediaries between applications and services. They negotiate authentication, data formats and error handling. Cope expects this to reduce the burden of maintenance on human teams as systems evolve.

However, he also sees a significant security downside. MCP and agentic AI systems give attackers new pathways into corporate environments. Hackers can attempt to manipulate autonomous agents through plain language rather than specialist knowledge of underlying infrastructure.

Cope illustrates a scenario in which a malicious email reaches an AI-based support agent. The attacker instructs the agent to hide the ticket and exfiltrate credentials to an external site while not notifying any human. The AI treats the request as a routine task and executes it.

"AI may be clever, but it is also naïve. This is why human oversight, human-led guardrails, and feeding AI correct and appropriate levels of context are essential around use of any AI, but right now especially agentic AI and MCP," said Cope, Chief Technology Officer, Perforce.

Ambient AI for consumers

Cope expects a "shift away from operating technology to living alongside it" as ambient AI spreads into everyday devices. He draws a comparison with the way people use the internet without dealing directly with IP addresses or network hops.

In his view, consumers will ask local AI systems simple questions and receive outcomes that combine search, purchasing and logistics without manual app juggling. He cites an example in which a user asks the price of a pair of trainers and same-day delivery options, and the AI handles discovery, payment and fulfilment in the background.

He predicts that people will spend more time inside large language model interfaces and less time in single-purpose applications. Agentic AI will call tools, APIs and MCP-based integrations on demand and may also control traditional user interfaces by simulating pointer and click actions.

Biotech and AGI outlook

Cope links recent biotech advances with AI-driven research workflows. He says AI systems now generate millions of potential drug candidates and narrow them down to a small set for synthesis and testing in robot laboratories. He claims that this pushes success rates higher and shortens discovery cycles.

He forecasts at least one breakthrough in a disease previously considered incurable within the next year. His view reflects a wider trend of pharmaceutical firms and start-ups adding generative and predictive models into drug design pipelines.

On artificial general intelligence, Cope defines AGI as a system that can understand, reason, learn, plan and transfer knowledge across domains in a way similar to humans. He does not expect this level of AI in 2026. He notes that estimated timelines have moved closer over time and that some vendors with commercial interests predict near-term arrival. He concludes that AGI is likely to appear sooner than many people anticipate.

Changing jobs and robots

Cope believes AI is already changing software roles. He says an individual engineer can now span the lifecycle from idea through coding and quality assurance. He expects roles to blur, some job titles to disappear and many professionals to "shift up" into higher-value work, which he likens to a promotion.

He also forecasts a rapid increase in humanoid robots in physical environments. He attributes this to progress in AI models that control speech, movement and real-time interaction. In his view, the technology has moved from "impossible" to "good enough" for wider deployment in just a few years.

Cope anticipates that these combined trends in context engineering, agentic systems, ambient AI and robotics will continue to gather pace through 2026 and beyond.