Tourism Bookings NZ introduces
Ninety-six percent of marketers believe that there is a significant gap between their organisation's vision for generative AI and how it's being implemented in practice.
This was a key finding from recent research by Econsultancy, 'Bridging the Gap Between GenAI Ambition and Execution', which surveyed more than 500 marketers to explore the state of generative AI within organisations and where there might be a disconnect between aspiration and execution.
The research, launched at Econsultancy's latest Capability Leaders Forum, found that although 85% of marketers are using GenAI either daily (50%) or weekly (35%), only 37% of organisations have rolled out working GenAI solutions. Of these, just 14% are at the stage of measuring ROI from their implementations.
During the forum's panel discussion, leaders from E.on Next and Canon EMEA explained how their respective organisations have moved from generative AI pilot to scaled-up usage, the results they've seen, and how to ensure that employees have a positive investment in the change taking place.
Alex Thurgood, Strategic Planning and Digital Operations Director at Canon EMEA, described the organisation as being "slap bang in the middle" of scaling its AI maturity. He noted that going from pilot to scaled usage "feels like quite a big jump, if I'm completely honest … [it's] quite challenging."
Thurgood described Canon EMEA's story less as one of being 'supercharged' by AI implementation and more as one of "provable incremental improvement in some backend processes".
With that said, the act of investigating pain points that could be alleviated by AI identified numerous areas for improvement, including some that didn't even need to be solved with AI. "[The team found places] where we might have missed something already - in fact, I would say about 80% of the things we identified could be improved by better communication," said Thurgood.
"We already had the answer - it's just that half the people didn't know about it."
In their exploration of AI applications, Canon EMEA's team initially went after low-hanging fruit by looking at free tools they could employ, such as an AI module within the company's existing social media management platform.
While some of the hoped-for use cases, such as reports and keyword listening, didn't prove effective, the tool proved to be a good fit for English-language captioning, saving the social media team time and improving their output - and some additional use cases were found, including paid social captioning and alt text generation, upping the accessibility of Canon EMEA's social media.
Further experiments with Microsoft's Copilot followed, and this proved useful for process-oriented tasks like meeting summarisation, action point creation, and document retrieval.
Canon's approach to piloting generative AI has been very structured, with trials conducted for a minimum of three months that include clear KPIs from the outset, and the team has been careful to quantify whether and how each application has saved time and improved quality. Altogether, "around 70% of people were finding significant improvements in what they did and saving around two hours per person per week - which is not to be sniffed at," Thurgood said.
He added that the method of beginning with no-cost options and then moving up to a more moderate investment as the benefits are proven has worked well: "Proving the case, and [then] small investments of a few thousand Euros each time has been the way to build it up."
E.on Next, which was launched as the innovative, agile and B2C arm of energy company E.on in 2020, has similarly taken an internal-first approach to scaling generative AI. As Mo Nuur, Head of Customer Product at E.on Next, explained: "You can't afford to get it wrong when you put something in front of customers."
Some of the guardrails that the company puts in place to ensure that generative AI applications are ready to be scaled include always conducting an internal pilot with "clear, locked-down KPIs", and also partnering with external experts whenever necessary.
E.on Next has also set up an internal 'AI forum' that anyone in the organisation can join to discuss AI policies, tools or training, "so that we can leverage and scale AI in a safe way. We are in a regulated industry - we want to do right by our customers."
One customer-facing use case that is currently being trialled with 15,000 customers is 'energy usage disaggregation', which employs AI to break down exactly how energy is being used by customers and what that usage is costing them. "We're seeing really, really great results in our current PoC [proof of concept]," Nuur told attendees.
One of the KPIs being used to gauge the success of the pilot is customer happiness, tracked via the Customer Happiness Index (CHI) metric: so far, the group with access to energy usage disaggregation is 20% happier than the control group. Customer happiness has been linked to reduced churn and increased customer lifetime value, meaning that there are multiple knock-on benefits for Eon.Next.
Generative AI is also being used within E.on Next to make life easier for customer service agents. Agents have some 700 FAQs available to them that offer information to help customers; but navigating these quickly while on a call with a customer can be difficult.
The company has therefore rolled out a tool known as 'AI Whisperer' that is trained on the same data and can guide customer service agents through their conversations by offering the most relevant information.
E.on Next is aiming to scale this technology into a customer-facing sales chatbot that can converse with prospective customers who are at risk of churning away.
"When we see that customers are about to drop out of the sales funnel, and insights are telling us that customers are probably getting confused with the number of different propositions … we can inject an AI sales bot … to help with guided conversations," said Nuur.
The best strategy in the world can still fail if there isn't buy-in across the organisation - or as Rose Keen, Content and Insight Director at Econsultancy, put it during the panel discussion, "Change really lives and dies on the people."
Nuur stressed that the way to create real momentum around the adoption of a new technology like generative AI is to ensure people genuinely want to use it.
"The technology and people need to go hand in hand - we all sometimes assume that at the flick of a button you introduce a new tool and it's going to deliver all the results that we want; but at the end of the day … there's a key distinction between wanting to use a tool [and] using it because [they're] being told to use it.
"We're working really hard to win the minds and the hearts of people," he added. "…We don't want to force AI on anyone, but we want people to use it."
He noted that when people feel genuinely motivated to make use of a tool, "they will find use cases themselves for it". This doesn't have to be something huge and flashy - E.on Next's position is that even a 1% improvement in efficiency is desirable because it will only add up over time.
In addition to the internal AI forum that has been established, Nuur said that the company encourages protected time in employees' diaries every week that they can use for self-learning and experimenting with the available AI tools. "Just having that open … entrepreneurial spirit, regardless of your seniority, really helps."
Chiming with this, Alex Thurgood described Canon EMEA's approach to GenAI adoption as "bottom-up", with the organisation encouraging people to find individual or department-level uses for GenAI to be improved on a case-by-case basis - rather than following a top-down directive to implement AI.
"That felt really onerous at the time," he admitted. "But actually, it's been really positive - because it has made people think about, 'Is this the right solution for the problem that I've got?' as opposed to this magic bullet … that's going to solve all our problems.
"In reality, it's just a tool."
Canon EMEA has taken a learning-forward approach to adopting generative AI, first covering the core skills (dubbed 'AI core') and then expanding on that learning (dubbed 'AI more'). This has involved trainees honing hard skills like effective prompting as well as softer skills like critical thinking, identifying AI falsehoods, and how to get the most out of the technology.
"It's been a really positive experience for the people that have been on [the training], and I think it'll stand them in good stead moving forward for utilising these tools," said Thurgood.
Nuur compared generative AI to other technological innovations that enhanced people's capabilities without doing away with the need for humans operating them. "If you utilise [GenAI] properly, it's making us all more 'T-shaped'," he said.
"In a tough market, it's good to be a T-shaped person - to have those superpowers … and we've seen those colleagues who've been learning how to use AI raise their own profile within E.on Next and move on to do better things.
"Personally, I'm very optimistic," Nuur concluded.
