We’re nearing the end of 2023, a year that saw conversational AI explode into the mainstream. I recently sat down with Frank Schneider, AI Evangelist at contact center technology provider, Verint, to get his take on some of the year’s hot topics for enterprises as they start thinking about 2024 plans.
Firstly, for all the talk and hype surrounding generative AI in particular, this specific technology hasn’t made its way into that many production enterprise use cases for 2023. Someone told me the other week that generative AI in production is like High School sex: everyone is talking about it, and everyone thinks that everyone else is doing it, but no one is.
Frank mentioned the pace of innovation accelerating this year: “there’s something new every other week.” He’s encouraged that advances like generative AI and large language models (LLMs) have the potential to drive real business transformation, not just hype, but market adoption is lagging behind where the technology is.
We discussed a number of big themes during the podcast, including:
Regulation and Safety
Concerns are growing about potential misuse of AI, from election interference to fraud. Frank stressed the need for “scrutability” – a word I’m an instant fan of. This is the ability to understand how models are trained and to explain the reasoning behind outputs.
This is all well and good as an aspiration, but even the scientists behind the models can’t say for sure why a given model produced a given output. They’re just too complex. You can potentially regulate what goes in and put guardrails on what comes out, but as for specifically pointing to the reasons for specific model output, we’re miles away from that.
Democratisation
Frank highlighted how brands like JP Morgan are now investing in their own branded LLMs, trained on its internal knowledge base. He expects more “power users” across industries as capabilities that were once cutting-edge and highly code-dependent start to become commoditised and accessible to less technical people.
I debate the semantics regarding branded LLMs because fundamentally, no brand is going to actually build their own models. The best they’ll do is to leverage existing foundation models and either fine tune them on their own company data, or provide the models with the ability to access company data via retrieval augmented generation (RAG). However, what’s certainly the case is that more brands are now thinking about having an LLM trained on their own data and this is something I think we’ll continue to see in 2024.
Trust and Compliance
As adoption of conversational AI accelerates, trust and compliance remains paramount, especially when handling sensitive customer data. Frank explained how Verint’s legacy in securing government call recordings makes it well-positioned to deliver AI embedded across the enterprise while maintaining rigorous security standards.
It’s certainly the case that brands will and should question the models they’re using and where their data goes, where it’s being stored and who can use it for what purposes. This is another reason I see open source foundation models being preferred by some organisations that don’t want to put all of their eggs in the OpenAI basket in 2024.
The Year Ahead
As we enter 2024, it’s clear conversational AI will continue its hype around the potential to reshape businesses and customer engagement. I certainly hope that we see an increase in deployments and some great case studies early in the new year.
I know everyone is trying, and we are, too; as we have been for the last 6 years, to make sure that this technology delivers at scale. Key questions remain around topics like regulation, safety and security concerning foundation models, but I share Frank’s optimism that generative AI’s momentum is only just getting started.
