Businesses are often daunted by the prospect of an autonomous agent becoming part of their team. Almost four in ten UK business leaders (38%) admit they’re fearful of using AI, above the 25% European average.
These are the primary concerns we hear when speaking to businesses of all sizes about introducing agentic AI to their team.
We need to remove the fear and uncertainty from the process. AI’s role is to connect people and technology, not replace them, and agentic AI is no different. Agents are becoming the thread that runs through a business, linking its people and technology together.
This isn’t just an IT project
Unlike other technological revolutions, introducing AI agents to businesses is no longer just the IT team’s responsibility. It needs to be a cohesive project touching every area of the business, from finance to HR to leadership.
UK business leaders use AI at almost the same rate as European counterparts (71% vs 70%). But that adoption falls off a cliff further down the business, with just 7% of junior teams using it day-to-day, against 20% across Europe.
AI must be truly embedded in business philosophy, right from the high-level strategic decisions to every customer interaction. It is not an add-on or a behind-the-scenes activity, but a commitment to innovating at every step of operations.
True AI adoption means not trying to hide where AI is used, but openly communicating internally about how AI, and AI agents in particular, can be used safely and properly to make employees’ lives easier, turbocharge business growth and deliver best-in-class customer service.
Start with the workflow, not the technology
We’ve spoken often about the importance of workflow auditing before AI is brought onboard. But what happens next?
Once you have reviewed businesses processes and identified where AI can bring the most value, the experimentation process begins.
Start with two to three identified processes and champions, whether that be revitalising current processes or building new ones where AI makes them now financially viable. The right AI solution depends entirely on the process, whether that’s a tailored AI agent built purely for the application, or using an existing agentic AI solution from partners like Microsoft.
It won’t be the same for each process; even within the same business, distinct processes need unique solutions. Getting this match right depends on deep knowledge of the AI agents on the market, as well as experience across a wide range of B2B sectors.
From financial reconciliation tools to templates for business development, to automating customer queries, there is already a huge wealth of both generative and agentic AI tools available. However, without the right approach and knowledge, it can seem like an impossible choice.
Three questions to ask yourselves when bringing in an agentic solution are:
- Is this an essential process, and what is its desired outcome?
- Does it fit into an existing category, or does it need its own tailored solution?
- How sensitive is the data being used, so what guardrails are needed?
Responses to these questions will help guide you to the solution best suited to the use case.
Closed model or open model?
The last question, on sensitivity of data, is often the question that keeps business leaders up at night. And rightly so. Data governance has moved from a technical footnote to a board-level responsibility under the Government’s Cyber Governance Code of Practice, and any AI rollout needs to be judged against that same bar.
There are two options with AI: closed models, that are built in your own environment and taught using only information that you feed it from your own company. And open large language models, which are the names you are most familiar with, that learn from the information you input and all data is public.
Retrieval-Augmented Generation, or RAG, is a model in which you train the LLM using information you feed it from your business. This can be a cost-effective way of creating a knowledge base which can assist your team easily. It is important that this is done in closed models, to keep data safe and secure.
72.8% of UK business leaders say they’re not fully confident using AI tools, against a 50.6% European average. That’s often where the right guidance, understanding which model actually works and why, makes the biggest difference.
Treat this like onboarding a person, because it is
Businesses always dedicate time, resource and attention to onboarding a new member of your team, so why would it be any different with agentic AI?
This is not an isolated technology entering the workplace. Agents work most effectively when they are centred within a workflow, team and business. Therefore, the more time you dedicate to introducing the agent to your business, to training it, and establishing a clear way of working, the more successful the agent will be.
The onboarding process will of course look different to a human team onboarding, because agents should always be human enabled. So, the best way to start is to involve the whole team in the training of the agent, in addition to their own AI training on how best to work with the agent. That way, it is clear that all decisions remain in human hands, and the agent is embedded in the teams’ way of working from day one.
It’s the mindset, not the tech, that’s hard
It is undeniable that AI has completely revolutionised our way of working. But that shouldn’t invoke fear. Companies that take the leap of faith now will be rewarded in the long-term, by reaping the rewards of an extended team. UK businesses are moving more cautiously than the European average on this front, with only 48% saying they trust AI more than they did a year ago, against 75% across Europe.
The mentality of AI agents becoming part of your team must change at all levels of the business. However, companies that truly implement this, through training, awareness and targeted implementation, will link their employees with technology to deliver real business growth.
We’re past the point of AI agents being a novelty, and into the harder, more interesting work of getting them properly embedded. Done well, that gives employees more knowledge and tools than they’ve ever had access to.
