Agentic AI could yield creative futures if it augments human workers
Can artificial intelligence systems capable of autonomous action and decision-making become a reality that augments human workers? Or will they be agents of oppression, hell-bent on efficiency?
Anyone who has been held in a customer service phone loop trying to resolve issues relating to products, services, accounts and so on will know the feeling – customer service can be a frustrating, time-consuming experience.
It feels like technology has not helped that much, so when Gartner trumpeted the prospects of agentic artificial intelligence (AI) last year, claiming it was a top tech trend for 2025, it felt a little hollow, something of a ‘here we go again’ moment. Was this a technology that would not only enable automated decision making, but would also form the bedrock on which more coherent and capable agents (AI or human) can operate?
Gartner refers to agentic AI as “a goal-driven digital workforce that autonomously makes plans and takes actions”. It sees it as “an extension of the workforce that doesn’t need vacations or other benefits”.
AI agents sit on top of this, beneficiaries of improved automated infrastructures and frameworks, managing and coordinating operational data. That’s the theory, at least, and recently Gartner doubled down, claiming that “by 2029, agentic AI will autonomously resolve 80% of common customer service issues without human intervention, leading to a 30% reduction in operational costs”.
[Agentic AI’s] ability to adapt in real time positions it as a cornerstone of digital transformation, particularly in sectors demanding dynamic decision-making
Kiran Raj, GlobalData
Kiran Raj, head of disruptive tech at GlobalData, says agentic AI has the potential to reduce administrative burdens in healthcare, streamline financial transactions, or optimise energy systems. “Its ability to adapt in real time positions it as a cornerstone of digital transformation, particularly in sectors demanding dynamic decision-making.”
These are big predictions. The problem is that when it comes to AI, we’ve had a lot of big predictions over the past 24 months. It can get a little weary fighting through the superlatives and references to “efficiency” and “seamless automation”.
While the idea of AI agents autonomously handling complex tasks is appealing, the reality is that adoption is slow. It’s a landscape still dominated by experimentation rather than scaled implementation.
That hasn’t stopped Nvidia CEO Jensen Huang from weighing in. During his keynote at the Nvidia GTC event recently, Huang described agentic AI as “the next wave” of intelligent systems, ones that can reason, plan and act with far greater autonomy than today’s predictive models. These agents, he suggested, will not just respond, they will actually do things.
Secret Escapes soups up chatbot with GenAI for better CX
Eirik Pettersen, chief technology officer (CTO) at travel company Secret Escapes, is one of the few UK executives already putting that claim to the test. Secret Escapes was an early adopter of Salesforce’s Agentforce technology, an autonomous AI agent platform designed to be an always-on support for employees and customers across a range of departments – including, of course, customer support.
“As soon as we switched over to the GenAI version [of our chatbot], negative customer comments dropped off completely,” says Pettersen.
When you are fielding disgruntled communications from customers unhappy with the tool that is supposed to help them solve their problems, you can see why Pettersen believes that Agentforce is “a step change in experience”.
It’s the kind of operational change that Forrester highlights in its latest research, describing agentic AI as “a paradigm shift in how enterprises scale, compete and create value.”
In its report, Agentic AI is rising and will reforge businesses that embrace it, Forrester argues that early adopters that align their investments with business goals will unlock growth, efficiency (there’s that word again), and entirely new revenue streams. But success depends on more than enthusiasm.
As soon as we switched over to the GenAI version [of our chatbot], negative customer comments dropped off completely
Eirik Pettersen, Secret Escapes
“Agentic AI isn’t just another technological evolution,” says Leslie Joseph, principal analyst at Forrester. “Organisations must rethink their operating models, invest in resilient AI foundations, and rally teams and workforces around a shared vision. The time to act is now – companies that wait risk obsolescence.”
Unlike traditional automation tools or standalone large language models (LLMs), agentic AI is defined by its autonomy and adaptability. As the Forrester report says: “It can plan strategically, reason through complex scenarios, collaborate between different components and leverage external tools to achieve objectives with remarkable autonomy.”
Forrester outlines a phased evolution where complex-flow agentic AI, which is already in use today, handles multistep tasks based on contextual goals. This is the entry point. Forrester goes on to suggest that multi-flow agentic AI will be next, enabling systems to collaborate and negotiate in real time across departments. Finally, there will be any-flow agentic AI, where swarms of agents act independently across enterprise ecosystems.
A reality check on autonomy
Not every organisation is ready to hand over the keys to an AI workforce, though. At Secret Escapes, Pettersen is still balancing experimentation with control.
“One of the main goals of our project was to reduce escalation to live agents,” says Pettersen. As a business that outsources its customer support (and pays human agents by the hour), this reasoning is understandably embedded within cost management. Escalating calls to humans can be expensive.
“We’ve seen around a 5% improvement, but we’re now learning to relax some of our guardrails and let the system do more. We’ve shackled our AI a bit too tightly,” Pettersen admits. “It hasn’t really had free rein yet.”
One of the main goals of our project was to reduce escalation to live agents. We’ve seen around a 5% improvement, but we’re now learning to relax some of our guardrails and let the system do more. We’ve shackled our AI a bit too tightly. It hasn’t really had free rein yet
Eirik Pettersen, Secret Escapes
What’s equally important, he says, is the empowerment of his non-technical teams. “Our head of customer service has been the one building the prompts. What’s music to my ears as a CTO is where people can improve the solution without having to use my engineers or my capacity.”
Pettersen highlights five early use cases, ranging from managing baggage queries and occupancy rules to processing cancellations and special requests, all of which can vary dramatically in format and language across airlines and hotels. Trying to model this manually was brittle and complex. But the flexibility of agentic AI offers a more human-like approach to resolving nuance at scale, he says.
Secret Escapes is not alone here. GlobalData cites a number of companies now experimenting with or scaling agentic systems. BT is using ServiceNow’s Now platform to try to cut resolution times to under a minute and reduce paperwork by 55%, while OpenTable has deployed Salesforce Agentforce and Service Cloud “to shave minutes off support calls while freeing staff to handle complex issues”.
These examples highlight agentic AI’s flexibility across sectors, from customer support to supply chain orchestration and marketing optimisation. But, as DeepMind CEO Demis Hassabis recently warned, while the signs are promising, companies have to be wary of compounding errors in AI. It is far from error-proof, which means that if agentic AI is to achieve what it sets out to achieve, organisations have to revisit the raw data and data processes that fuel their AI.
From experimentation to reinvention
The idea of moving from task automation to business model transformation is exactly what consultants like Publicis Sapient are now urging clients to explore.
“Agentic AI will become the way AI is delivered,” says Simon James, head of data and AI at Publicis Sapient. “An evolution from replicating legacy processes to reimagining ones that can be improved upon by generative AI.”
He sees the term agentic fading over time, as it becomes embedded within everyday enterprise AI. But for now, businesses must make some tough calls.
“What companies need to decide is the level of autonomy they’re comfortable delegating to agents and what role human oversight should play,” he says.
Rather than betting on general-purpose super-agents, James notes that successful implementations are more modest in scope. “We’ve seen the definition of agentic gravitate toward a more deterministic model, breaking a process into small chunks, where each block is an agent with limited, discrete functions. It’s more practical, more transparent, and easier to manage,” he says.
What companies need to decide is the level of autonomy they’re comfortable delegating to agents and what role human oversight should play
Simon James, Publicis Sapient
But is it worth the cost? As agentic AI adoption increases, so too do the questions around cost, not just in terms of implementation, but in operational overhead.
“Businesses need to be mindful of the increase in cloud consumption costs driven by AI applications,” says James. “Any AI-focused programme should be self-funding, and proportionate to the opportunity and evidence of results.”
This aligns with Forrester’s view that agentic AI should not be treated as just another tech layer. It’s not about automation for automation’s sake, it’s about rethinking what’s possible.
“The winners of the AI age will not merely be the fastest adopters,” says the firm’s report. “They’ll be those who redesign their business models to harness agentic AI as a strategic differentiator.”
The momentum is also reflected in the investment landscape. According to GlobalData, agentic AI attracted $1.8bn in venture capital funding across 69 deals in 2024 alone. Notable raises include $220m for Paris-based startup H and $97.2m for New York’s Emergence AI, signalling confidence in this emerging infrastructure layer.
But we have seen AI investment spikes before, and the individual investment sums are not earth-shattering – at least, not yet. Perhaps the real indicator of agentic AI’s potential success lies in the reasoning applied by corporations in adopting agentic AI.
As Ethan Mollick, one of the authors of The cybernetic teammate: A field experiment on generative AI reshaping teamwork and expertise, recently wrote: “Companies that focus solely on efficiency gains from AI will not only find workers unwilling to share their AI discoveries for fear of making themselves redundant, but will also miss the opportunity to think bigger about the future of work.”
The point is that if organisations see agentic AI as purely a way of saving money, then they will undermine the true value of agentic AI in supplementing, or even enhancing, existing employee capabilities and productivity. This may be a harder sell to a financial director looking at head count, but it could well be the reality, at least for the next five years.
Perhaps the true test of agentic AI won’t be the tech itself but whether companies use it to reinvent or simply reinforce old habits.
