Since the public release of OpenAI’s ChatGPT chatbot in November 2022, generative AI has captured the imagination of academics, businesses and the public, and reintroduced debates around the impact of artificial intelligence globally. Combined with natural language processing (NLP) technology, models such as GPT-3 and GPT-4 are able to create a more personalised chatbot experience for customers. While capable of boosting speed of creative tasks and content creation with the relatively simple use of text prompts, the technology has been scrutinised for its information flaws and potential biases.
Indeed, building a strong AI model that can widely benefit an organisation long-term relies on accurate datasets and rigid, continuous algorithm training carried out by a diverse group of professionals. It’s this mission that is being focused on at integrator Sprint Reply. Making use of Sprint Reply’s consultancy services is aiding organisations in getting the best value out of chatbot technology, powered by GPT models.
The benefits of chatbots in generating sales for start-ups — Here’s how chatbot technology can drive value for start-up sales teams.
Understanding the true capabilities
As tools such as ChatGPT continue to evolve, organisations need to clearly determine which business processes would truly benefit from generative AI deployments, and avoid jumping on the trend for the sake of it. This means keeping up with updates to models like GPT-3 and GPT-4, to gain an understanding of how the company can innovate long-term. This is where an integrator like Sprint Reply can help, satisfying the crucial need for a human-in-the-loop.
“It’s a constant trail of adoption, which calls for daily tweaking and tuning, but while it almost feels the same as traditional NLP, you get there a lot quicker due to the capability of large language models to better understand your intentions,” said Sprint Reply technical director and associate partner, Tim Shepheard-Walwyn.
“I think one of the main themes from our early conversations, is the idea of failing fast in the market. We’re keen to be transparent about not knowing all of the limitations, and neither does the customer. So we’re almost the trusted hands that help them to navigate that journey without having all of the answers.”
While still in its early stages of exploration, use cases identified by Shepheard-Walwyn as potentially valuable in the long-term include:
- customer email processing – helping to respond to queries for faster data-driven responses;
- document summarisation – especially for large documents that take time to read through;
- meeting recording and diarisation – helping to summarise the key points made in meetings, and identify who exactly made those points, for future reference.
“We’ve found email to be a channel that’s gone untouched,” explained Shepheard-Walwyn. “All emphasis is currently on NLP going into generative AI and extending it as a chatbot function across multiple channels.
“But we’ve been using gen AI to process customer complaints and inbound, as well as manage multi-threaded conversations and fulfilment.”
Retaining customer trust
Once company employees become accustomed to testing and deploying large language models, there is also the duty of ensuring that customers stay onboard with the innovation, maintaining a consistent level of service throughout. Fears have been raised around chatbot technology usage at scale being misleading, or even incorrect in its answers to customer queries due to reported data limitations. This situation calls for complete transparency – a key aspect of many discussions around regulation of the technology in recent times.
“We’ve seen some organisations display transparency as almost a watermark over their avatars, with it being made clear that these avatars are being generated by AI. This should be called out and explained to customers for any content, even use cases like email,” said Shepheard-Walwyn.
“Consumers need the option of stating that they don’t want an AI delivering service, with businesses in this case ensuring that humans are on hand next time – this awareness is really important.
“We also need to be transparent about the iteration, and capture any negative customer experiences quickly, because otherwise this can start to snowball, and affect thousands of people as well as brand reputation.”
AI-led business processes – getting the balance right between business impact and staff satisfaction — How businesses can properly weigh up innovation and employee purpose, for long-term success.
The benefits of a specialist integrator
Pivoting in line with customer needs and industry trends is a key aspect of Sprint Reply’s operations as a tech integrator. As a deliberately technology-agnostic brand – not purely utilising a hyperscaler environment like Microsoft, AWS or Google – more options are available going forward for properly aligning gen AI deployment strategies with the particular goals of the client business.
While it remains common to fail fast in an early stage innovation realm like generative AI, the long-term goal is to find the ideas that have legs, and allow for consistent growth as well as maintained customer experience.
Shepheard-Walwyn explained: “We want to compete by bringing the deployment cycle forwards, but while being light to touch and not going into too much depth and burning money in the process.
“I won’t say that traditional projects as we know them are completely dead yet, but we’re pulling the rug out and seeing what happens.”
This article was written as part of a paid content campaign with Reply.
Related:
Tech leaders profile: seeing AI as an opportunity — John, Mike and Kevin from N-able’s leadership team explain why you should be leaning into AI and how to embrace the opportunity mindset.
