AI - Clari CEO says niche is good when it comes to enterprise GPT

The rush to adopt ChatGPT and other large-language or generative AI systems has been epochal and extraordinary. It’s on a par with the 80s desktop computing revolution, the 90s dotcom boom, and the generational shifts towards search, mobility, smartphones, and social networking that followed. Now the GPT engine and its competitors are fast becoming business platforms, with applications spreading invisibly into people’s lives. 

But the sudden popular uptake of generative AI feels different to previous disruptions. Often it seems faddy, hype-driven, lazy, misguided, or tactical. Many users are shouting “me too!” without strategic forethought or consideration of their business models – just a glance at today’s share price or tweet deck.

As explored in previous diginomica reports, this troubling behaviour implicitly regards human talent, expertise, and skill as assets that can now be generated for free via a ChatGPT prompt. Italy has banned the technology until better guardrails can be put in place, while even one of the world’s most impulsive and popularity-obsessed men, Elon Musk, has urged a pause in development to let the world take a more considered view. 

Yet away from the me-too herd, solid enterprise applications are emerging, notably some that help organizations unlock the value of their own data, rather than of other people’s historic insights. Just ask your database a natural-language question, say a new generation of business intelligence (BI) providers. An innovation with few downsides – unless you’re a data analyst, perhaps (time to update that CV!).

But are those companies jumping on the same bandwagon as the public? 

It’s all about the data and workflows

Andy Byrne is founder and CEO of revenue software specialist Clari. Earlier this month, the 10-year-old company launched RevGPT. The new GPT-powered tool queries data stored in Clari’s RevDB database, so users can ask critical revenue questions – in areas such as risk, sales goals, and forecasts – then act accordingly. 

So, unlike the ChatGPT instances that have been trained on, and refer to, data scraped from the pre-2021 Web, RevGPT and its enterprise ilk solely access companies’ own, trusted data. Over time, their employees will experience a “flywheel effect” of increasingly accurate answers, actions, and outcomes from it, according to Clari.

Byrne tells me:

What’s powerful for us is that we're bounded in our use cases. RevGPT is not this general-purpose ‘ask anything of anything in the world’ application. And if you combine that with the proprietary data we have, we're able to make powerful predictions and suggestions in every ‘revenue moment’ for every revenue-critical employee.

He continues:

I've been in and around the revenue process for over 30 years. And for the past decade, it's been my professional responsibility to help CEOs, CROs, and boards answer the most important question in business: will they meet beat or miss on revenue? 

What we saw was that larger companies were using the three-headed Hydra – CRM, Excel, and BI – to run this most important business process. Systems that are general purpose and not designed from the ground up to run those workflows.

Unlike Clari, of course. The company now has over 1,000 customers, with data about $1 trillion of revenue being processed through its systems. So, why did Clari see OpenAI’s GPT engine as something that could enhance its product roster? Byrne says:

We think of RevGPT as being like the cousin of ChatGPT, but with a quota. Or like introducing a Chief of Staff for every person in the revenue process. And that Chief of Staff is there to drive more efficiency and productivity at scale.

Take time to revenue. How long does it take for a company to get results from their revenue process? How long does it take a rep to update you on their deal, to put that data in the system and comment on whether that deal is going to close? And how long does it take an exec team to roll up forecasts from EMEA or from APAC? These are all time-to-revenue considerations, or revenue moments. 

But because of the three-headed Hydra, it is taking them way too long, so they get revenue leak, which is happening everywhere across the system. There are leads that never get touched, and targets that go stale because no one's following up on them. There are deals that slip, or companies lose them and have no idea why. Boston Consulting Group has said that, on average, there is $2 trillion of loss happening every year due to revenue leak.

Hence Clari – and now RevGPT. But at what point did AI come into play? According to Byrne, AI is in the company’s DNA, but GPT presented unique advantages for a provider that has a decade of niche expertise behind it. Byrne explains:

Large language models, they're going to be commoditized. What's important is the data and the workflows, and not really the algorithm. Because everyone's going to have access to GPT-3, 4, all the way up to 10. So, it's really about the data you have, your proprietary data. And we have a lot of interesting proprietary data in RevDB.

So, we're putting a [software] Chief of Staff right beside the frontline manager. And it's analyzing all the conversations that are happening through conversational intelligence. 

Imagine you're that frontline manager. You have eight reps, say. And they each have 10 deals, and they are having three to five conversations a day, right across every one of those. That’s hundreds of conversations. How, as frontline manager, do you know what's going on? 

So, we have Smart Summaries [in RevGPT], which take those verbose transcripts. It analyzes them and provides the top-three things that happened in each meeting. The top-three next steps. 

And it's profound how accurate it is. It’s not always accurate – there are things that don't make sense – but for the most part, it's amazing. And if you tie that back to the time it takes for a frontline manager to understand what’s going on [it’s transformative].

Connections and competition

However, Byrne reveals that some of the impetus for adopting GPT came from Clari’s connections with the US venture capital community. As ever in Silicon Valley, a lot of innovations are driven behind the scenes by wealthy investors and their own networks of connections. He says:

We've been thinking about it ever since Sam [Altman, CEO] started OpenAI. If you look at our board, and predominantly Sequoia Capital, and the connections we have through Bain Capital, Silver Lake, Blackstone, and so on, they all have their own data science departments. And Sequoia, in particular, is very connected to the AI ecosystem. 

So, we've been playing around with it for years, both through both our direct work with early versions of ChatGPT, and through the connections our board has, which have given us access to all the earlier large-language models.

On that point, how concerned is Byrne that other vendors are deploying AI in the BI space – companies like ThoughtSpot, Sisense, and others? Might any of those offer revenue insights alongside their other offerings? And might OpenAI’s key investor, Microsoft, itself move deeply into that space?

In short, might Clari’s focus be too narrow for long-term success? Or does Byrne believe that niche depth and expertise are the way to go in a world of general business functionality?

He says:

The competitive threat is always a worry. I'm your classic paranoid entrepreneur, always wondering where the next competitive threat is. 

That said, there are going to be innovative, AI-based capabilities at every level, and every enterprise use case is hard. Where I’m confident is that we are in a very deep domain, one that’s very specific. It's all about running revenue, arguably the most important process in any company. And we believe our workflow design, sitting on top of our proprietary data, puts us in a position of strength.

Does Byrne have any concerns about the rush to adopt generative AI? His answer is refreshingly candid, given the hype surrounding the technology. He says:

If I was in my 30s, I would say no. But now I’m in my 50s, I'm definitely ambivalent about it. I do worry about how it's used. Where are the guardrails? 

I do think this is a great moment for government policy. We need to have the right brains working in government, with the technology leaders. And OpenAI is doing that. But we need to see an acceleration of the guardrails that need to be built for privacy and security. The ability to have some oversight.

But where we are in our domain, I’m less concerned – since it's bounded with guardrails on our proprietary data. The AI is only applicable to the use case we provide, so I think there's less risk. But there’s no question that there needs to be an accelerated thinktank put together, given the speed at which AI is being deployed and offered out to the marketplace.

My take

Wise words in a world of hype and evangelism. Evidence that solid use cases are emerging that help professionals do their jobs better, rather than regard creative people as a problem in need of urgent solution.