Agentic vs copilots: what’s the future of GenAI?

It has been two years since the launch of ChatGPT, but with generative AI still high among tech priorities, the nature of artificial intelligence for businesses is changing.

Whilst generative AI differs from previous types of AI by offering the ability to create new content and ideas, such as text, images, videos, music, and more, from pre-defined data, it still requires a significant amount of human input.

But for enterprises looking to transform the way they work, there is a new AI kid on the block—agentic AI. Almost a direct successor to chatbots, agentic AI refers to artificial intelligence systems that possess a degree of autonomy and can act on their own to achieve pre-defined goals.

This means, rather than responding to prompts a la ChatGPT, agents can make decisions, plan actions, and learn from their experiences, to achieve these goals. Siri or Alexa, but with autonomy and the ability to learn.

Just last week, analyst firm Gartner named Agentic AI as its top technology trend for 2025 at the Gartner IT Symposium/Xpo 2024. It predicts that by 2028, at least 15% of day-to-day work decisions will be taken autonomously through agentic AI, up from 0% in 2024.

“Agentic AI systems autonomously plan and take actions to meet user-defined goals,” said Gene Alvarez, distinguished VP analyst at Gartner.

It should come as no surprise, then,  to see major vendors launching their own agentic AI solutions. Last month, Salesforce took wraps off its “Agentforce” AI at its massive Dreamforce showcase, with Microsoft reportedly set to follow suit next month.

Celonis, Lenovo, Bud Financial, and CrewAI are among some of the names also launching agentic hybrid AI offerings.

So is agentic AI the next big thing, or just a stepping stone on the generative path?

Agents vs Copilots

As he strode through aisles at his Dreamforce keynote last month, Salesforce CEO Mark Benioff revealed that the likes of Saks, Wiley and Wyndham Hotels had all used the software giant’s GenAI offering to create “next level” agents.

Agentforce is a suite of autonomous AI agents that Salesforce claims will “augment employees” by handling tasks in service, sales, marketing, and commerce.

He set out bold ambitions for the platform, saying Salesforce aims to have a billion users interacting with AI agents by next year’s event and empower one billion agents with Agentforce by the end of 2025.

He took aim at rivals for using GenAI in a “bolt-on, DIY” way that offered little business value for enterprises, adding that the tech was better off embedded into applications through virtual agents.

“Agentforce represents the Third Wave of AI—advancing beyond copilots to a new era of highly accurate, low-hallucination intelligent agents that actively drive customer success,” he said.

“Unlike other platforms, Agentforce is a revolutionary and trusted solution that seamlessly integrates AI across every workflow, embedding itself deeply into the heart of the customer journey. This means anticipating needs, strengthening relationships, driving growth, and taking proactive action at every touchpoint.”

He took a swipe at rivals offering copilot models, including the likes of Microsoft. But the Windows giant has already announced plans to dip into the agentic game.

At Microsoft’s AI Tour event in London, the software firm revealed it will add the ability to build autonomous AI agents within its Copilot Studio starting next month.

Nadella also said Microsoft will launch ten new autonomous agents in Dynamics 364, its suite of enterprise resource planning and customer relationship management apps.

Commenting on the transition from copilots to agentic AI, Summeet Arora, chief development officer at ThoughtSpot, said that while copilots act primarily as augmentation tools that boost productivity, agentic AI will democratise the analytical skills of data analysts, “bringing their capabilities to the fingertips of every business user, engineer, and application builder.”

He added: “We’ve moved from an era of descriptive-predictive analytics to assistive copilots, and now, agentic implementations are at the forefront of the analytics revolution. As the tectonic plates shift, analytics itself must be autonomous for the age of the autonomous business. These models not only support the future of AI and analytics, but the future of human creative potential itself.”

Call my agent

Generative AI is moving out of the “hype” phase with enterprises and vendor partners now looking for scalable use cases and a return on investment, claims SAP EMEA AI officer Jesper Schleimann, who adds that copilots will continue to play a key role for many businesses.

“Lots of organisations can claim to be ‘AI-powered’ but if they truly want to redefine how they operate, be agile to market demand and pressures, and succeed in today’s fast-moving business landscape, then they need an AI copilot like Joule.”

Joule is SAP’s AI assistant which can be integrated directly into existing business applications.

“Copilots, like Joule, optimise how businesses run with unprecedented intelligence, efficiency, and decision-making capabilities – not to mention the capability to push boundaries around creativity and innovation. For businesses, this means faster processes, lower costs, and better scalability, all while freeing employees to focus on strategic initiatives.”

But the industry is again on the move, from an era of descriptive-predictive analytics to assistive copilots, and now, agentic implementations are at the forefront of the analytics revolution.

That is the claim of Sumeet Arora, chief development officer at ThoughtSpot, who adds: “Agentic AI will democratise the analytical skills of data analysts, bringing their capabilities to the fingertips of every business user, engineer, and application builder. The models will reduce bottlenecks in time-to-insights, driving unprecedented levels of productivity.”

Crash landing for copilots?

With the rise of agents seemingly inevitable, does that mean the end of copilots? Or can models that leverage human interaction and development with AI exist alongside models deliberately designed to cut out or reduce the human element?

For Eleanor Watson, IEEE member, AI ethics engineer and AI Faculty at Singularity University, there is a downside to switching to agentic AI models.

This, she says, comes with the “potential risk of over-reliance and reduced human oversight, as these systems operate at arm’s length.”

She adds: “Agentic systems require very careful value and goal alignment to help ensure that systems do what we want of them, not simply what we tell them. Otherwise, systems may ‘work to rule’, take dangerous shortcuts, or railroad others and violate their boundaries for the sake of expediency.”

Contrasting this with copilots, which focus on collaboration, she explains these retain “human agency” while boosting productivity through “intelligent augmentation.”

This is “particularly useful in creative, knowledge-based roles where human oversight remains necessary. Copilot AI will soon come to wireless headphones, listening and commenting on our daily lives, for example, “Close the deal!”

She concludes “The choice between these models depends on a company’s priorities. Businesses that prioritise full automation may lean toward agentic models, while those seeking augmented intelligence may prefer co-pilots. Both have potential, but their success will hinge on how well they align with the specific needs and risk tolerance of the enterprise.”

Dom Couldwell, head of field engineering EMEA at DataStax agrees that enterprise decision makers need to be aware of the differences between the different AI models, and says it is vital to determine the situations where agentic or copilots may be a better fit.

He explains: “Agentic AI is better suited for more complex processes that require support across multiple tasks to achieve that goal. The reason for this is that LLMs and generative AI applications can be trained to support copilot use cases by leveraging data for specific tasks using Retrieval Augmented Generation, or RAG, but they may not be able to switch and perform as well on multiple different tasks even where you have multiple RAG approaches and options in place.”

However, there is a risk that in cutting out human input, companies could “use a sledgehammer to crack a walnut”, bringing additional costs and less autonomy.

“In short, use a copilot to assist a person where the task is specific. Use Agentic AI where the goal is more complex and can involve multiple tasks being completed. Use RAG in both to provide the context supplied to the LLM.”