Sector-specific, process-led AI set to reshape firms

Enterprise adoption of artificial intelligence is set to move away from generic tools and towards industry-specific, context-aware services, according to Joe Logan, Chief Information Officer at iManage.

Logan said organisations are starting to demand AI systems that reflect their sector, workflows and compliance obligations rather than broad, one-size-fits-all models. He links this shift with growing scrutiny of data residency, privacy rules and emerging AI regulation in major markets.

Process-led prompting

Logan argued that "process prompting" will become a key differentiator between advanced and casual AI users in business settings. Process prompting refers to the design of structured instructions that set out not only what the user wants but also how the AI should reach its conclusions.

He said early testing inside enterprises has shown that prompts which demand transparent reasoning improve the usefulness of outputs. Requesting that models "show their work" produces more complete and detailed answers than simple questions.

Logan highlighted the growing use of multi-step instructions that break down tasks into discrete actions for the AI system. Users are also starting to rely on fixed templates that define both the final format and the method the model should apply.

These more prescriptive prompts focus on the process behind the response rather than only on the final answer. Logan said users who gain skill in this style of interaction will extract more value from AI tools as they become embedded in daily workflows.

AI-as-a-service growth

Logan said he expects rapid expansion of AI-as-a-service offerings from next year, in a pattern similar to the rise of internet services and software-as-a-service in previous decades. He predicts that leading platforms will arise first in sectors such as legal, healthcare, pharmaceuticals and scientific research.

He described the current market as fragmented. Enterprises are experimenting with internally focused "small box" solutions and pilot deployments that often lack formal governance frameworks.

Agent-style tools inside workplace applications have started to show how this model might develop. Logan cited systems such as Microsoft Copilot that weave AI into established productivity software as early examples of this direction.

He anticipates that the most successful platforms will be those that combine workflow design with industry knowledge and regulatory alignment. These services will package predefined processes and compliance measures for specific sectors.

Compliance and trust

Logan expects data residency and geopolitical tensions to make vendor trust an explicit selection criterion for global enterprises. He said multinationals already face difficulty meeting the strictest rules in each jurisdiction without slowing their operations.

He believes customers will look for evidence that cloud providers can handle complex, multi-country data flows and legal requirements. They will prioritise vendors with audited controls and recognised frameworks for privacy, security and AI risk.

According to Logan, independent attestations will become central in this process. He highlighted standards and schemes such as ISO, CSA STAR, FedRAMP, IRAP, Cyber Essentials Plus, the EU-US Data Privacy Framework, emerging NIST AI risk guidelines and the EU AI Act as examples of frameworks enterprises will reference when assessing suppliers.

He said vendors that cannot meet these expectations for compliance infrastructure will lose ground in enterprise procurement cycles. Those that can provide documented assurances across multiple regimes will gain share.

Shift from build to buy

Logan also forecasts a change in how organisations provision AI models. Some enterprises have spent the past year experimenting with building their own large language models. He now expects a move towards purchasing industry-specific systems instead of maintaining in-house platforms based on generic tools.

He said generic assistants such as broad-purpose chatbots usually produce outputs that span many topics and use cases. These outputs often lack the specialised terminology, workflows and constraints that apply in individual sectors.

Logan expects more demand for contextually tailored AI products that reflect the norms and processes of areas such as professional services, regulated industries and knowledge-based work. Vendors will update these products in line with changing rules, technology developments and user behaviour.

This ongoing adaptation could extend the life and relevance of sector-specific AI services. Logan said regular refinement and tuning by specialist providers would respond more quickly to regulatory change than isolated internal builds.

"This ability to grasp process-led prompting will become the defining user skill for effective use of AI, and will be the difference between the sophisticated and casual user of AI," said Logan.