Adobe’s AI evangelist says enterprises must move beyond AI experimentation to trusted agentic systems

Adobe’s AI evangelist for Asia-Pacific and Japan, Rochelle Tognetti, said enterprises need to move beyond isolated AI experiments and start building trusted AI systems that can work across business functions with the right governance, data and security in place.

Speaking at the Adobe Summit held in Mumbai, Tognetti said AI has rapidly become a capable assistant for individual productivity, but businesses have yet to fully scale its use across organisations.

Over the past year, she said, AI has transformed how people search for information, create content and make purchasing decisions. “More than 25% of consumers globally are going to AI chat agents to discover products or research information. That’s one in four. AI is very firmly sitting between your brands and your consumers.”

She said while customers have changed their behaviour quickly, enterprises are still largely experimenting with AI at the task level. “There is a tonne of individual experimentation. Lots of isolated wins. But there still is an enormous opportunity to scale across your organisations.”

According to Tognetti, most companies are still using AI for individual prompts and workflows instead of redesigning business processes around customers. “Many enterprises are trying to automate existing processes rather than reimagining them with your customers at the centre, not built around the way your organisation operates.”

She said businesses should begin treating AI as a teammate rather than simply another software tool, especially as AI agents become capable of coordinating work across multiple enterprise systems.

However, she cautioned that greater autonomy also increases business risk if adequate safeguards are not in place. “Imagine a scenario where an AI agent accidentally deletes one of your customer records because it simply misunderstood a prompt. The liability here is real.”

To make AI reliable for enterprise use, Tognetti said organisations need three foundational elements: trusted business context, governance and security.

She explained that AI agents should be grounded in enterprise data and workflows through a trusted semantic layer, while permissions, ethical guardrails and security controls ensure agents remain within approved boundaries. “When you put all three together, the right semantic layer, the right guardrails, the right security, then you’ll have something most enterprises don’t have right now. You’ll have an AI system your business can trust and rely on every single day.”

Tognetti also outlined Adobe’s four-stage framework for AI adoption. It begins with AI assisting employees, progresses to AI augmenting human decision-making, then to AI automating workflows with human oversight, and finally to AI operating independently alongside teams.

She said most organisations currently sit in the first two stages. “The opportunity for you then is to get to level three. Deploy with confidence, learn from it and prepare for level four.”

Explaining Adobe’s approach, Tognetti said the company’s agentic AI platform combines large language models with an AI “harness” that governs how AI behaves in production environments. She added that the platform is designed with governance, auditability and integration capabilities, allowing enterprises to adopt AI while working with existing technology investments and AI platforms.