Deloitte is expanding its artificial intelligence controls and assurance capabilities as businesses increase their use of generative and agentic AI, putting greater focus on governance, model performance and the reliability of AI-generated outputs.
The professional services firm is positioning its AI Controls and Assurance offering to help organisations assess and manage risks associated with deploying AI across business operations. The services cover areas including AI governance, controls, data integrity, model performance and outputs, as companies prepare their AI systems for independent and evidence-based evaluation.
The offering is designed to address risks across the AI lifecycle rather than focusing only on models after deployment. Deloitte's wider Trustworthy AI approach includes AI risk management and governance, regulatory support, model risk management, engineering controls and AI audit and assurance.
As part of its AI assurance capabilities, Deloitte provides services around model validation, including quantitative testing, guardrails and evaluation techniques involving large language models. Its framework also covers continuous monitoring of AI systems and the design and implementation of controls.
The expansion comes as enterprises move beyond standalone generative AI tools towards systems with greater autonomy. Agentic AI can independently identify tasks, create plans and execute actions, introducing additional governance considerations for companies using the technology in business processes.
Deloitte has identified audit trails, human oversight, guardrails and goal alignment among the controls organisations may need when deploying autonomous AI agents. The company has also highlighted transparency, fairness, privacy and security, reliability and accountability as areas for organisations to consider when evaluating AI systems.
For enterprises, AI governance is increasingly intersecting with existing audit, risk and compliance processes. Businesses deploying AI in finance and accounting, for example, may need to assess whether AI-generated outputs are sufficiently accurate and reliable while maintaining records of how automated decisions and processes are produced.
Deloitte's AI assurance strategy also extends to its own audit technology.
In June, the company introduced a unified network of AI agents within Deloitte Omnia, its global audit and assurance platform. The system brings multiple AI agents together under a common framework, allowing them to coordinate and execute workflows while supporting risk analysis and other audit activities.
Omnia's agentic capabilities are available to Deloitte's nearly 85,000 Audit & Assurance professionals globally. The company said the technology was developed internally and follows its Trustworthy AI framework, with governance, controls and compliance incorporated throughout the lifecycle.
Deloitte is simultaneously investing in AI training through programmes such as the Deloitte AI Academy and AI-powered learning tools. Its approach maintains human supervision over AI-supported audit and assurance processes.
The development reflects a broader shift in enterprise AI adoption, where companies are moving from experimentation towards integrating AI into operational workflows.
As those systems take on more business-critical activities, controls around data, model behaviour, accountability and monitoring are becoming part of enterprise AI deployment. Deloitte's expanded assurance offering is aimed at helping organisations establish and evaluate those controls as AI moves deeper into corporate operations.