AI solutions can give businesses and governments a 75 per cent or more productivity gain in certain tasks. And though it’s still important to keep humans in the loop when implementing and using these solutions, experts say they’re here to stay.
“AI is increasingly embedded in the way we do business. As the technology continues to evolve, I think it will become even more tailored for each use case and produce more bespoke results for the end consumer,” Trellis Data’s Tim McLaren says.
“It’ll be important to ensure users are combining AI’s data-driven insights with human judgement… AI is always most powerful when enhancing human decision-making, not replacing it.”
The move towards tailored and in-house solutions comes down to one thing, according to Tim. He says many of our customers want to leverage AI but are worried their data will be leaked or end up overseas.
“They’re looking for trusted, local, secure AI solutions that integrate seamlessly into their existing processes,” he says.
One of the more powerful tools offered at the moment is Trellis Secure Chat – a new application from Trellis Data that combines the power of a large language model with an organisation’s entire data set. It is designed to help complete many time-consuming standard business tasks, such as building reporting, analysing large amounts of information and creating a range of new content.
This new application can analyse and leverage an organisation’s entire knowledge library in just seconds. It also links back to the source document, so clients can better trust the results by knowing where they came from – a key concern for many AI users.
Critical to Trellis’ clients – many of whom are government, defence, not-for-profits or service organisations such as legal or financial companies handling sensitive or personal data – the technology is an in-house alternative to other products such as ChatGPT.
“We make commitments to our customers that their data will never be used to train models unless that’s something they want. They can be sure their information won’t be leaked or, importantly, sent overseas,” Tim says.
By leveraging the company’s extensive knowledge base, the AI in Trellis Secure Chat learns from past successes and failures, applying that experience to generate new, tailored content. This technology means companies no longer need to worry about the loss of corporate memory that comes with personnel changes or be hamstrung by human limitations.
“Projects that happened in the past might be forgotten, the lessons learned evaporate. However, with Trellis Secure Chat you can leverage AI to bring together the information that exists across your databases and documents. All of a sudden you have the ability to create new content, new proposals, new plans and new reports for new uses and projects,” Tim says.
“We call this ‘Knowledge Mastery’ – the ability to leverage institutional knowledge to create innovative solutions for today’s challenges.”
Combining this large language model with corporate databases can yield powerful tools that can search more than a million documents with a sub-three-second latency and produce new content in moments.
“Imagine you’re working for a legal firm, and you know a colleague wrote something in a document or there’s relevant case law on a topic. You could manually comb through all the existing documents or search a phrase in Trellis Secure Chat, which would then look across your entire organisation and produce a result for you in under three seconds,” Tim says.
“It’s a bit like using Windows Explorer to search for documents, but on steroids … now imagine using that kind of technology across various industries.”
Another primary function of Trellis Secure Chat is generating new content. For example, a procurement document for an upcoming project based on a company’s previous tender proposals. Trellis Secure Chat can parse all the relevant data and extrapolate a new tender document, adhering to set parameters based on past successes and failures.
“You’re not going to get the final version right away every time, but at the very least, you’ll have a very good head start because you’ve short-circuited the time needed to review past proposals,” Tim says. “We also help train clients to use this technology, because it does require some skill and knowledge to get the best results.”
Tim says one risk to bear in mind with some generative AI is biases that can cause the technology to “hallucinate or create something you’re not after”.
Trellis Data works on solutions to suit its clients – organisations looking for “true to form answers”.
“Our internal research team once ran a bunch of documents submitted as part of an audit process for a company looking to secure its ISO 27001 certification to see if they’d meet the required information security standards,” Tim says.
“Third-party commercial models are typically geared towards being so ‘helpful’ that they stopped short of saying ‘This document doesn’t meet the standard’. When we ran again in Trellis Secure Chat, it told the truth – it wasn’t up to scratch.
“That’s the kind of straight-down-the-line answer our clients want and need.”
For more information contact Trellis Data.