From contact centres to financial services, there are very few – if any – workers that know every possible detail of their business. In many cases, employees are working with a vast array of facts and figures, many of which they will only need to access once or twice across their entire career. 

Heather Richards, CEO, Transversal

However, when a worker does need access to a given piece of information, they may not get it from the most reliable source. They may be working from outdated documents; trusting a co-worker to know the answer or assuming that knowledge gained several years ago remains in-date.

As a result, even very good workers – particular those under time pressure – can often end up providing inaccurate or outdated information to customers or colleagues, which can lead to customer complains, internal issues or in some cases a lack of compliance.

It’s a serious problem for many businesses, but it’s an area that artificial intelligence is increasingly solving. An emerging class of AI products – often dubbed augmented intelligence solutions – are unifying company knowledge by providing accurate, consistent information exactly when it’s needed, in a way that is effortless for workers to use.

One such company is Transversal, which offers what CEO Heather Richards describes as “cognitive knowledge solutions” that are tailored to each business’ needs.

“Our main audience is either web self-service, contact centres or enterprise HR,” she says, adding that users experience the product “in very different ways, depending on how it’s deployed”.

“The end result is they should be able to find answers to questions and retrieve information as efficiently as possible,” she says.

“So that could be that they're asking a question: typing in a question, getting an answer back; it could be in a contact centre: they're interacting with a CRM system, they're automatically pushed relevant information based on who they're speaking to or what the situation is in terms of a customer service perspective.

“Or it could be even on a website: depending on the context of your interaction you’re automatically served relevant additional information.”

Immediate accurate information: Using AI to boost contact centre knowledge

The benefits of such technology clearly vary by field, but in contact centres, which are one of the most common application spaces for Transversal, it is about providing consistent and reliable knowledge quickly.

“In a contact centre, what it allows is first that one version of the truth, as opposed to the risks that you have with either agents resourcing a stack of pdfs that may or may not have been updated, or – probably even worse – them just remembering 'well, this is what I said last time so that might still be accurate', or asking the person next to them,” says Richards. 

“The other benefit is looking at the way that we're able to proactively push knowledge based on additional inputs, so it reduces the time it takes to get accurate information, and in contact centres that's a huge thing, when you're looking at the efficiencies, that you can bring to those interactions. 

“If you can reduce the time by 30% that it takes agents to deal with a specific query, that adds up very, very quickly.”

“If you can reduce the time by 30% that it takes agents to deal with a specific query, that adds up very, very quickly”

As with many solutions-focused AI products, Transversal is designed to work across many different sectors, which is demonstrated by the company’s broad customer base.

“We have customers in automotive, financial services, retail, telco,” she says. “We do have quite a lot of financial services customers, I think that part is that we're ISO 20001-compliant, so from a security perspective that's very appealing to financial services. Also they tend to be offering quite complex services across a range, and so you may need something like what we provide to enable agents to be up to speed on all different products.”

This multi-sector approach does not just apply to companies, but also to their customers.
“One of our more recent customers is Wolseley, and they actually sell plumbing and building supplies, and what we've actually enabled them to do is set up shared service contact centres, where one agent has the ability to basically provide help across a huge range of products, whereas before that was very difficult,” she says. 

“So more and more we're seeing customers like Wolseley, or even BPOs like Capita or Capgemini, using our technology to allow their agents to more efficiently serve a huge range of their end customers, as opposed to one sector.”

Bringing intelligence to knowledge solutions

The reason the technology can produce such results lies in the built-in intelligence, which goes beyond the conventional query and return associated with traditional database searches.

“In a typical search, you type in your word and it brings back matches. What we're doing is actually looking at the meaning behind what you're typing, or beyond what's being searched, so for example, it knows an apple is a fruit that grows on a tree, that can be red and green, and also a computer, depending on the context,” explains Richards. 

“And so it's not just doing keyword matching.  It's doing something a lot more sophisticated than synonyms: it's actually looking at the contextual meaning that you're looking for, which makes it quite smart.”

“It's doing something a lot more sophisticated than synonyms: it's actually looking at the contextual meaning that you're looking for, which makes it quite smart”

The system also takes context into account, in order to anticipate what is likely to be asked of it next. 

“As soon as you've clicked on something it starts clustering together other related information within the knowledge base, which means that it then proactively starts serving information based on your own behaviour at that time; how you're browsing, what you're looking at,” she adds.

“So really with only one or two clicks, you could be taken through an entire chain of thought.”

AI solutions for now, not tomorrow

What makes Transversal stand out from more conventional machine learning-driven AI solutions is that it doesn’t require a period of learning before it can be used. 

 “If you look at other solutions that are characterised as AI, say for example IBM Watson, that relies on a huge amount of data, that's where big data is useful, to actually analyse and then start finding patterns within those vast amounts of data,” says Richards. 

“The algorithms we use in our solution are finely tuned for small datasets, which means that it’s not having to do a lot of analysis across huge amounts of data to find an answer, it's actually doing matching in a very small dataset. So that's what makes our AI unique.”

Trained on the internet but built to draw exclusively from clients’ own editable databases, it can be implemented through an API, with a custom service provided by the company or even plugged into widely used CRM systems such as SAP or Salesforce. 

“It also makes it very useful for customer experience or customer service applications, because it means it’s up and running and useful from day one,” she adds.

“So after 8-12 weeks' deployment, away you go. Now it does learn over time, so as you add additional information into the knowledge base it automatically starts clustering that information into the right categories without any additional input, but again what makes us unique is the training required to get it working is very minimal.”

“Many people, in terms of having to have a demonstrable ROI, cannot turn around to the board and say 'well, in five years' time we might have this up and running'”

For businesses looking to make better use of their data through AI, this makes the product a very appealing offering. Previously many companies have looked to machine learning to add value to their data, but have often found they have too small datasets for the results to be meaningful, or that they have to wait for a long time before they see a return on their efforts.

“Say you take something that's working solely on machine learning, which does require vast amounts of data to start accurately spotting the appropriate trends or patterns, in many cases organisations either don't have the data, or it takes a couple of years to acquire it,” explains Richards.

“So it could be that maybe two or three years down the line you have a system that's working great, but when you take that back to the real world, many people, in terms of having to have a demonstrable ROI, cannot turn around to the board and say 'well, in five years' time we might have this up and running'. 

“We've seen that quite a bit, and it’s also one of the key reasons that our customers have selected us over other vendors is that you get the intelligence of the AI that we provide but without the requirement of the long training times or the requirement of huge datasets.”

A model for near-term enterprise AI

Offering an alternative vision of the application artificial intelligence to the workforce, this kind of augmented intelligence has the potential for a far more positive image that the job-replacing automation solutions that so often capture the public’s attention.

“I think with business in general there's a lot of scaremongering out there about robots are going to take all of our jobs,” says Richards. “For business in general, I think [the future is] bringing back the focus onto how computers and artificial intelligence, which is basically just machines that can mimic cognitive functions, can actually help improve and augment human behaviours and responsibilities.”

However, while products such as Transversal are out there, this is a market that at present remains under-served. 

“You say AI in a customer service context and people immediately think of bots, chatbots: 'Oh, I'm going to have a conversation with a robot and ask a question'. And yes, that's one aspect of it, if you want to have a conversation with a robot, great, if you want something that can pass the Turing test, happy days, but that's not what we're trying to do,” she says. 

“AI isn't and doesn't always have to be 'let's have a conversation with a bot'. There are more seamless, almost more invisible ways to provide answers to questions.”

“It's a different problem we're trying to solve: instead of a question/answer, question/answer, like a conversation, what we're able to do is actually provide you the information with as little input as possible – it could be a couple of keystrokes, not even a full question, and we're able to advise exactly what it is that you're looking for through a combination of sort of search-ahead technologies and contextual information retrieval, as opposed to requiring the pain of having a full conversation while you're trying to type on your iPhone.

“It is an underserved market, and I think a lot of it has to do with the sometimes not full appreciation that AI isn't and doesn't always have to be 'let's have a conversation with a bot'. There are more seamless, almost more invisible ways to provide answers to questions, and in many cases in a customer service experience that's the more desirable one, the one that's the least intrusive.”

Share this article