With tantalising promises around efficiency and the appeal of automation to aid cost reduction, there is likely already pressure within your business to implement AI and seize the benefits. According to Forbes, 83% of executives believe AI is a strategic priority.
We’ve seen a pace of implementation unlike anything before, driven largely by technology companies trying to gain a first-mover advantage. On a classic hype cycle, it seems we’re not near the peak of inflated expectations yet. Every day brings new advancements and, in turn, new expectations.
However, the fear of being left behind doesn’t change the fundamentals of delivery. Companies looking to AI need to step back from the hype and look to implement with the same due diligence as any other IT project in the past. Perhaps even more so.
Identify your business needs before implementing AI
It’s a familiar tale told through IT project history: implementing a tool to solve a problem without truly understanding what you’re trying to solve. It often doesn’t end well.
“You need to know the problem in order to solve it,” said Kimberly Ramsey at TigerTMS at our recent Future of Communications in Hospitality event. “You could throw AI at it and say, ‘Okay, let’s buy this AI; it’ll do this and that, and it’ll collect all this information.’
“Oh, well, great—but is that actually usable data? What are you actually going to do with that data to affect your operations and enhance processes moving forward?”
With AI, nothing has changed fundamentally in terms of delivery. First, you need to decide what outcomes you wish to achieve, then decide on the appropriate technology for the job—whether that’s AI or something else entirely.
“AI is the Swiss Army Knife. It’s the tool you use to build the outcome, it’s not the solution,” warned Rob Wiles from Zoom at the same event. “You’ve got to be laser-focused on the outcome I want by using this part of the Swiss Army Knife.”
Corey Vilhauer at Blend Interactive also echoes the sentiment, saying AI is a tool, not a solution. “If your first thought is ‘I wonder how AI can help us,’ I’m here to say you are already walking down the wrong path,” he advises.
“The real question should be ‘How can I improve this process,’ with AI augmentation as one of many potential tools. Instead of wondering how AI will help you hire fewer people or create meaningful communication out of thin air, think more about how large language models will help you pair critical human analysis with the added benefit of more data points.”
As with any project, the key is a deep understanding of the function affected. Interview those doing the work every day, learn their pain points, and only then begin to see how AI may help improve processes. Only when focused on the right outcomes can you begin to contemplate the rewards, such as lower costs or increased productivity.
Garbage in, garbage out
Many systems, such as CRMs, live or die by the quality of their data. AI is arguably even more sensitive to data quality, and the data you train it on is vital.
Its accuracy will be directly related to what it can learn from. Too little data or out-of-date data, and the AI may confidently provide incorrect answers.
It’s also important to be aware of security and access controls and ensure the AI cannot provide information that would otherwise be inaccessible to the end user.
“A large percentage of organisations don’t use the data that they have to good effect. The data is there, but sometimes it’s questionable where it is as it is spread across four or five different systems. And nobody is actually doing anything with it.” Ant Morse, Futurist, AI Thought Leader, and Founder of Adventa, said to us in this ClevrChat episode, discussing all things AI.
Before you begin implementing AI, it’s vital to ensure you have the right data and that it’s correct so that the AI can provide reliable answers.
Ant suggests three key steps.
- Identify where your data is stored – Establish a clear policy and understand what you need to keep and why you’re storing it. This housekeeping exercise is essential to understanding whether the data you have is also needed to comply with regulatory guidelines.
- Clean and optimise your data – Eliminate duplicates, tidy up the data and ensure accuracy to create a solid foundation.
- Put your data to work – Figure out how to feed this to AI systems to enhance services, streamline processes, or solve existing problems.
Your reputation is on the line
In the rush to get AI to market, disclaimers are common. “ChatGPT can make mistakes. Check important info” appears on the page of any search at the OpenAI tool. Imagine an Excel spreadsheet that imagined figures and the chaos that would cause, yet for AI, people take hallucinations (where AI tools are prone to ‘make up’ answers) as part of the cost of embracing a new and powerful technology.
However, your appetite for risks or mistakes won’t be shared by your customers, who often view attempts to remove human interaction with suspicion.
“Virgin Money chatbot scolds customer who typed ‘virgin’” was one headline recently that demonstrates the damage that poorly implemented AI can cause to a brand. Meanwhile, Air Canada tried to distance itself from incorrect advice given by its chatbot by arguing that the chatbot was a “separate legal entity that is responsible for its own actions”. The British Columbia Civil Resolution Tribunal disagreed that the chatbot was somehow responsible for itself and found in the air passenger’s favour.
These are just two cautionary tales of how to create a bad reputation with AI interactions.
As such, it’s vital to get it right when using internally, but especially externally. In cases like this, your brand is on the line, not the chat provider’s.
Consider an AI steering group
AI is, in some ways, a great leveller on a corporate level. It’s unlikely that one team or company has vastly more domain knowledge than another. Yet.
But with the market evolving very fast and AI likely to touch every part of your business, an AI steering group may be worthwhile.
Users can share knowledge of what they’re learning in the market and the success stories of their industry peers. With small modifications, what may be useful for one team may be relevant to another, so sharing and learning about successes and failures is vital.
It also pays to bring in technology partners such as CloudClevr or, at the very least, ensure you ask them the right questions. We have teams of people focused on learning technologies who work directly with our technology partners to learn and share knowledge with our customers. With the pace of change in AI, it’s vital to use partners’ resources effectively.
Don’t let AI fever distract from successful delivery
The benefits of AI are already becoming apparent, and there’s no doubt they will positively impact most businesses. However, don’t let the industry hype or excitement within your own business distract from what remains the core principles of any project delivery.
- Identify what you wish to achieve – focus on the outcomes, not the tool.
- Don’t try to deploy AI everywhere all at once. Get it working well once, then use the learnings to deploy to other use cases and departments.
- Get the fundamentals right in understanding your processes and ensuring you have not just the right data but good quality data to provide positive outcomes.
- Pull in knowledge wherever you can. Have people in your business who can focus on AI, and make sure you draw on the knowledge from partners such as CloudClevr.
- Remember your reputation is built on the outcomes, and you can’t, like Air Canada tried, distance yourself from your own tools. Taking your time will pay off in the long run.
It may be an exciting new tool, but with AI, it’s really a case of going back to basics.
