‘The minute we saw ChatGPT, we knew this was like a “before and after” moment for the technology’

Tech firm Intercom was quick off the mark when the fuss around ChatGPT, the artificial intelligence chatbot, started gathering steam a couple of months ago.

The company, which makes customer communication software, integrated the AI technology developed by OpenAI into its own chatbots for its customers.

Since its launch, the ChatGPT tech has demonstrated its use in answering difficult questions and writing scripts and news stories. This made it a good fit for customer support bots.

For Paul Adams, chief product officer of Intercom, this marks a new chapter in AI development and one that could radically change the interactions between companies and their customers.

Intercom, a privately owned company which is valued at over €1bn, develops chatbots for company websites that can answer various support questions posed by customers. Adding OpenAI’s technology to the mix could potentially ramp up the complexity and volume of tasks that Intercom’s tech can handle.

“The minute we saw ChatGPT, we knew this was like a ‘before and after’ moment for the technology,” Adams told the Sunday Independent.

Fin, its AI chatbot, is currently in beta but is in use by dozens of companies ahead of a wider launch.

It uses OpenAI’s GPT-4, the artificial intelligence model that underpins ChatGPT, and trains itself on the data generated by companies’ operations to automate responses to customer queries.

It’s just one example of the many uses for generative AI – this new form of the tech that has caused much excitement and consternation across the tech industry.

Proponents laud the tech’s ability to improve productivity while critics fear the upending of entire jobs markets.

​Customer support appears to be the sector most ripe for AI’s impact – and customer support affects countless other industries in turn, from telecoms to banking, airlines to healthcare.

“I think it’s definitely going to be a significant disruption. I think that’s really clear to us,” says Adams. “I don’t think it will be necessarily what people think it will be.”

It will not be a case of a switch being flicked and all customer support representatives being replaced by AI chatbots, he says. The tech may have made its most advanced steps of late – but it is not perfect.

“Some customers are seeing amazing results. They’re seeing like 30pc resolution rate, which is huge for a support team. That’s a quarter or more of their queries that otherwise would have had to be fielded by humans,” he explains. “But not everyone sees those numbers.”

Fin is being refined to answer increasingly more complex or nuanced questions that early adopters of the technology are seeing.

Other companies are approaching AI with more caution.

Professional services firm EY opened its AI Lab in Dublin late last year just as the hype and buzz around generative AI was kicking off. Since then, clients have had more and more questions, according to Paul Pierotti, data and analytics partner at the firm.

“Where we are just now is that we have lots of clients that are curious. We have lots of clients that are nervous, and we’ve lots of clients coming up with really interesting ideas,” he says.

“They’re actually seeing their family use it in real life. They’re seeing their teenage kids using ChatGPT for particular things.”

Pierotti says that this stage of AI is comparable to the internet in 1997. It is still not widely used by businesses, but is on the cusp of something much larger.

Within the walls of EY’s AI Lab and its own services, partners are catching an early glimpse of the jobs that will be impacted. One of those is tax services.

“There’s a lot of complexity and understanding there, so how can you improve and support the tax expertise – particularly for geographies you don’t know as well?” Pierotti says.

He points to one customer that services 60 countries. Having the in-house expertise on all of those jurisdictions is impractical.

“How can you supplement the expertise in your team with some sort of a generative AI solution that can actually learn from the tax, the payroll or whatever rules that are out there for that country?”

Legal services, another admin-intensive area, could be aided by AI in the gathering and compiling of documents and first drafts of contracts alongside the work of paralegals.

Pierotti describes this budding relationship between the human and AI like a “co-pilot”.

“We in Ireland are navigators of lots of very complex supply chains around the world. If you look at life sciences and agri foods, the schedules and the planners for a lot of supply chains are here. There are lots of little decisions there that can be automated.”

​One does not need to look far to see headlines about AI’s potential toppling of the jobs market.

It is one facet of the ongoing Hollywood writers’ strike, with writers concerned that AI could be used to conceive scripts and concepts that would otherwise be done by a team of writers.

Last week BT announced plans to axe up to 55,000 jobs over the next several years, and replace around a fifth of those jobs with AI tools.

Job cuts have dogged the tech industry over the last 12 months, with large firms cutting thousands from their headcounts as they adjust to an economic downturn. Amid that upheaval, in recent weeks tech giants IBM and Dropbox have indicated that AI could at least partially fill the roles vacated in these cuts.

It is stories like these that set off the alarm bells around AI and the future of jobs.

“I wouldn’t be of the view that entire industries are simply going to be wiped out,” says Robert Ross, a senior lecturer in computer science in TU Dublin.

Ross, who is also an investigator at the ADAPT centre, says many industries are still going to need some human oversight on many tasks.

“The boundary line is going to move a bit. We’re going to see that some tasks where you used need a human in the loop can be made more automated – but it certainly isn’t going to clear everything out,” he explains.

“There’s a difference between something being disrupted and something being devastated.”

The dynamics will shift for the consumer experience, he says.

In the case of customer support, Ross envisions that AI may come to dominate those first lines of interaction, with the option to speak to a human agent becoming a paid option.

“It might be something associated with the premium product, a little bit extra in terms of cost – because it will be more costly.”

It remains to be seen, he adds, just how comfortable consumers will be with giving up control of decision making.

People are comfortable using voice assistants like Alexa because “it isn’t going to give much in the way of lip, so to speak” but how much control are you willing to hand over to the machines?

Last week a study published by researchers at Trinity College Dublin showed that more than 50pc of respondents were “worried” or “very worried” about AI-powered autonomous cars, but could still see the benefits.

For Ross, industries racing to pick up the baton with AI still need to balance the human impact rather than take a gung-ho approach to the technology.

“I think we need to be careful with how conversational AI is rolled out with children, and with anybody who might be considered in a vulnerable group. There isn’t necessarily enough research there yet.”

There has already been some backlash in the tech industry to this effect. It came after Snapchat, the social media app, integrated an AI chatbot into the service, which is popular with teenagers.

These are exactly the types of use cases that give cause for pause, Ross says.

“Don’t go too quickly too soon.”

Get ahead of the day with the morning headlines at 7.30am and Fionnán Sheahan’s exclusive take on the day’s news every afternoon, with our free daily newsletter.