Five things you need to know before piloting AI agents

The field of AI is moving forward at a blistering pace - new functionality is being released every week. The latest frontier in AI implementation is Agentic AI. These are AI systems that complete sophisticated operational tasks autonomously, mimicking human-like levels of comprehension and deducing next steps through reasoning and training, much like a new employee. This approach has powerful implications for your company (and your competitors too).?

So, if you are thinking of piloting AI agents in your organisation, what do you need to be thinking about? Below are some of the softer, non-technical insights we have gleaned through our work piloting AI agents at Alix. We develop AI agents for customer service applications, however, the learnings apply to any AI pilot.?

Have a visible senior organisational champion for AI?

Like any change process, having staff know that a new initiative matters to the leaders of an organisation is key to gathering the necessary support required for the change to be a success. This means having a key person who visibly understands the opportunities related to AI agent deployment and champions these through effective communication to employees. Running an AI pilot is an amazing way to excite teams about the improvements and change that is possible - wise leaders will grasp the opportunity to inspire new ways of working. A good example of how to do this is through a monthly roundup of new AI capabilities and some thoughts on how it could impact the organisation.?

Pilot in areas that have technical and operational fit

Using a new AI tool for the first time can feel like magic but it is still just technology, so finding the right tasks to pilot is key. This means taking time to talk to internal and external experts about what is possible and carefully define the desired business outcomes. What happens before and after the automated process? What tolerance is there for error during the pilot phase and how will these be monitored? How will data flow between one system and another? What processes are required for the pilot and what can be defined later on?

Starting small is not necessarily a bad choice, as it allows pilots to be contained and impacts measured effectively. Some of the most successful examples of tasks have skilled teams that can assess performance on an ongoing basis, providing valuable feedback to inform agent training and improve results. Think of an AI agent as a new team member - they need training!?

Identify leaders who can drive detailed operational deployment and champion results with teams

Some of the most valuable applications of AI are in processing data in certain ways, often requiring an expert to fully capture the nuance of an ideal outcome. Deep operational experts are an ideal sponsor for this type of project - informing on requirements and, critically, managing the relationships with teams that need to be consulted and brought along for the ride.?

A pilot is an opportunity to offer technical expertise to a user who might not be well served by existing organisational software. It is hugely empowering for them and their overworked teams, so choose a leader who gets excited about playing with new shiny tools and can transmit their enthusiasm to others too.?

Selecting the right person helps create a sense of excitement about the possibilities and ensures that solutions are designed to meet meaningful business outcomes.

Usability is key

When a new technology is deployed, it is important that users know how to use it and feel comfortable doing so. Co-design approaches help enormously in ensuring usability, both in terms of gathering accurate requirements and, perhaps more importantly, building a culture of listening and collaboration to ensure that the resulting technology meets their needs.?

Alix carefully co-designs the interface to feel easy to use and works with teams to support adoption. After all, what’s the point in having an amazing new tool if staff are too scared to use it??

Develop a learning mindset

Every leader wants to make the most of new technologies and with everything changing at such a rapid pace, it’s critical to have a system or dedicated 1-2 hours a week to learn about what is going on in your industry and adjacent verticals. However, real organisational learning is gained through application - that’s at the heart of a pilot. In planning a pilot, it’s worth considering where the greatest scope for value capture from that ongoing learning exists. Some questions that you might ask are:

• Which information functions are still largely manual and a significant contributor to costs or organisational risk??
• Where are there lots of edge cases and complexities that still require careful review from experts but might be simplified with new tools??
• Where are the most valuable opportunities for evolving your existing work by automating it, piece by piece??

This incremental approach allows new technologies to be tested in small ways, building on learning and improving over time. Ask any finance graduate about the magic of compounding and they’ll get all misty-eyed about how we massively underestimate the long-term value of this.?

The future of work is human - machine hybrid. Piloting AI helps employees to understand this new way of working and is a powerful opportunity for learning by doing.? It’s an organisational skillset that is absolutely fundamental to future winners, so don’t be afraid to make a start.?