How Sharesies is embracing the AI spring

There’s no doubt about it; the ‘AI spring’ has well and truly sprung.

With promises of improved employee efficiency and enhanced customer experience too good to ignore, statistics show that nearly half of New Zealand’s businesses are dipping their toes into some form of artificial intelligence (AI). 

Sharesies is one such business. In October 2023, the online share trading platform announced the integration of AI into the company, rolling out a product, Intercom Fin AI, in the customer support department. 

Through AI language models, Fin resolves customer issues with answers based solely on a company’s provided support content. In the case of Sharesies, this content came from help centre articles produced by the marketing team. 

“The motivation really was how can we free up our people’s time to focus on higher value tasks?” says Carl Olsen, Sharesies’ general manager for operations. 

The chatbot, built by San Francisco-based software company Intercom, seems to have done the trick. 

When Sharesies launched the AI assistant, the resolution rate for customer queries climbed from 19 percent to 45 percent within a couple of days. In the months since, the upward trajectory has continued, hitting 60 percent in mid-January. Response times have dropped from hours to minutes. 

Customer satisfaction for Fin is sitting at 76.9 percent, which Olsen says is a good baseline for the Sharesies team to build on. 

“It’s been a massive shift in terms of our support side of things, and we’re still able to offer human support as well, which has been key,” says Olsen. 

He describes the cost of incorporating the platform as “fairly modest,” but provides no number. 

Though Sharesies has long believed in making ‘computers do computer things and people do people things,’ the company – responsible for managing the finances of over 600,000 investors across New Zealand and Australia – exists within a highly regulated environment. Accordingly, incorporating AI into the business was no overnight matter.

Here’s what Olsen and Richard Clark, technical director at Sharesies and one of the business’ six cofounders, found made the process that bit smoother. 

The crux of it all? Take the time necessary and be cautious. 

‍

Don’t skip the paperwork

Like most of the world, AI regulation in New Zealand is a grey area. 

AI aficionados predict the country will follow the footsteps of the European Parliament, which, in June 2023, approved a world-first’ AI Act,’ aiming to harmonise the regulatory framework for AI use. (Check out what AI startup Arcanum founder Asa Cox has to say on the matter here). 

With top down AI guidance lacking, Clark says it’s on industry users to make robust governance decisions for their startups to protect both the business and customers from risk: “We can’t just go cowboy.” 

Though this can seem overwhelming from the outside, he says it’s much easier than people first realise – especially as startups have much of the paperwork in place thanks to pre-existing privacy and security policies.

“You’ve probably got about 90 percent of the job done already,” says Clark. 

A comprehensive governance resource for AI adoption can be found here. 

‍

Make AI work for you

Robust governance models are a good starting point, but putting time into testing the technology is fit for purpose is just as important. 

A well-documented issue for AI language models is known as ‘hallucination’, where AI will conjure answers of its own, hence Open AI’s disclaimer ‘ChatGPT can make mistakes. Consider checking important information.

For the Sharesies team, which deals with finance questions daily, one of its biggest concerns was that the AI chatbot might provide a customer with wrong advice. 

Intercom won Sharesies over as Fin was programmed to direct the user to human assistants if an answer could not be found. “I spent ages bashing at it trying to get it off-topic as part of my own confidence building in the solution,” Clark says. 

It also helped that a working group featuring people from both Intercom and Sharesies was set up early on in the process, opening the dialogue from the start. Through that, Clark says they were able to get significant assurance that customers' data would be appropriately protected.

‍

‍

Put people first

When it comes to AI, one of the biggest questions centres around the impact on the workforce. 

Recent data shows that just over a third of business leaders are concerned that incorporating AI will create fewer job opportunities for New Zealanders. Conversely, though not New Zealand-specific, a 2023 study by Forrester Consulting found that over half of front-line employees are worried AI will put them out of a job. 

With similar concerns, Olsen says managing the change with people was “front of mind” for the Sharesies team. The trick proved to be transparency and ground level engagement from the start. 

The wider team was involved in developing the content the AI would use. That process helped everyone see the value themselves, particularly around how AI could ‘open up space’ for the team to deal with grittier tasks. 

Olsen says the new dynamic has created opportunities for career development that were previously unavailable, such as operational staff moving into development and data analytics roles as an example. At the moment, Sharesies is deploying that additional capacity to finetune the AI.

The early team involvement also helped improve overall AI literacy within the organisation. 

‍

Keep checking in

Both Clark and Olsen say “constant review” is central to Sharesies’ AI governance model and critical to making the AI behave. 

“Never underestimate how much time is required to ensure the content you’re producing is always watered, fed and maintained,” says Olsen. 

Customers are also part of the monitoring process, with Sharesies ‘inserting humans’ to provide feedback for review. Those communications help determine where change is needed and the timeline for making that happen, says Olsen.  

While AI reviews were to be expected, Olsen says onboarding Fin unearthed an unexpected human mistake.

“Fin actually informed us of a change that we needed to make, whereby a page on our website was providing the right answer but we actually weren’t ready to provide that type of onboarding service just yet,” he says. 

Once made aware, thanks to the AI, Sharesies disabled the page. 

While AI is working for the business so far, Olsen and Clark reiterate that it is a tool. Like all tools, if the reviews were to find it negatively impacts processes in any way, then ‘turning it off’ is always on the table.

‍

‍

No matter what, get familiar

Having had his eye on the sector for a while, Clark says he’s developed “a bit of a philosophy” around AI, particularly for those interested in getting it into their startup. 

“My belief is that all of us have now developed a little bit of our brain like a search engine model. When you go to Google, [your brain] tells you what keywords are useful and what will get you nothing but spam.” 

He believes that people need the opportunity to develop an ‘instinct’ around AI as they did with Google. Making the workplace a safe and encouraging learning environment is a big part of that, he says. 

The philosophy applies to both those with intent to hop on the AI bandwagon and those without. 

As AI is about improving processes, taking the time to familiarise oneself with what's on offer in a low stakes environment can help discern if AI has a place in a business or not. 

Having the groundwork to embrace any potential future opportunity is critical, says Olsen. 

For Sharesies, AI continues to hold promise. Internally, the business has AI deployed to help lift capability, triage complex issues and resolve problems faster. 

On the customer service front, Olsen says improving customer satisfaction is the key objective. While the team is constantly feeding the system with new content to enhance experience, they are also in the process of creating multi-channel AI support, in collaboration with a partner.