Main content CEO on using intelligence and teamwork to respond to a crisis

Gary Flood Profile picture for user gflood May 26, 2020
Sri Ambati, CEO, says that artificial intelligence is a useful tool - but organisations need to be willing to listen to the data.

Image of Sri Ambati, CEO
(Image sourced via

Today, the enterprise faces not just the immediate, overwhelming issue of COVID-19, but multiple challenges. If you're a bank, every day you need better answers to mortgage lending, credit risk scores, fraud/fraud detection, anti-money laundering; if a retailer, you're always having to think about how you'll find your next best customer, what your next best offer is, how to attract new customers? How do you understand the patterns of your customers, and so on.

The latest tool in the enterprise armory to do this is supposed to be Machine Learning (ML), which is really what 99.9% of what's meant when we say "Artificial Intelligence" right now. The promise is that Machine Learning is an additional set of capabilities that should give businesses in every industry the ability to garner better intelligence on trends with their data in their possession.

The problem: doing ML right is hard. You need deep mathematical capability, as well as lots of data to work over-and not every company has a bank of trained data scientists ready to be unleashed on a problem. This is where, a leading ML company, is trying to make a play by a Freemium model of ML access, effectively making it very easy for "any" company to get up to speed quickly with the approach. Simply put, it's a maths/stats package that gives you a quick on-ramp to do "automated" Machine Learning.

That means that even if you haven't heard of it, you may already be using it somewhere in your organisation; H2O is used by 20,000 separate entities and by hundreds of thousands of individual data scientists. Analysts think its Open Source route to market is valid, with Gartner telling clients it has the "Strongest Completeness of Vision" in its 2020 Data Science and Machine Learning Magic Quadrant.

H2O has impressive penetration in financial services/insurance, healthcare, telco, retail, pharmaceutical and marketing. H20-based ML is being used in managing claims, detecting fraud, improving clinical workflows and predicting hospital acquired infections. On the paid-for version, commercial customers of include brands like Wells Fargo, Capital One, Kaiser Permanente and Nationwide Insurance, do similar things; we're talking sifting large datasets to spot anomalies and patterns to do credit scoring and investing better, essentially, as well as supply chain optimisation and, right now, a lot of COVID-19 response planning.

Listening to the data

Getting to 20,0000 user organisations in only eight years since you started offering payroll is progress that's attracted the interest of VCs, with the company successfully raising north of $72m last year. The company claims that it's really the main play in automated Open Source Machine Learning platforms, but in the corporate space there are two main competitors Dataiku and DataRobot.

What's particularly interesting about H20 is that its users are investors, according to its CEO, Sri Ambati, who told us many of the financial firms who use the company's offering are also "very strong believers in what we do" in more tangible ways (including Goldman Sachs). As Ambati told diginomica,

What we're trying to do is democratice AI and make sure very high-grade Machine Learning and the best mathematical algorithms are there for you to build your own data science capability from scratch. 250,000 such data scientists use us every day, already.

Fine, but quite a few CIOs and CEOS remain sceptical about the "AI" bubble, which surely has to burst (again) soon. For Ambati:

There's a lot of AI hype, yes. But some sort of intelligence is super-important for any business leader trying to react to the crises that we seem to be experiencing one after another, from Brexit to COVID; they used to be like 9/11, and be once a decade, now they're every year.

We can help you spot the patterns before they become widely visible. But you still have to have the courage to act on that intelligence and take the decisive step, of course. But there are always three people that need to come together to make that happen: AI is a team sport.

The CIO can put AI in the enterprise and ensure the software fits within the specifications of the enterprise, can ingest the data, and the team has the data ready to go-but you need a data science person or a savvy business analyst to actually run it. All those people have to all be in agreement that this is the way forward.

The obvious place where such decisive steps need to be taken will be getting set for Recovery, post-COVID-19, of course. This is where H2O believes Machine Learning will really come into its own:

The big thing right now in retail is distribution, right? So as stores come back online, they will need to determine their supply chain, their inventory, but even so they're trying to figure out customer patterns. AI can detect those patterns better than humans. It can help you by saying ‘Serve this up to Sri,' for example, "or this next offer to Derek because he's likely to take that offer and buy something.'

So no, I don't think AI will go out of fashion. In fact, now more so than ever, companies are calling us because of Coronavirus and the impact on their business and how we can help them. What we see as the next logical step of a digital transformation is an AI transformation, in fact.

My Take

Ambati's clearly all about the quality of the work, and less so about the big IPO cash-out. After all, this is a guy who's teaching his young daughters Python at home on Saturday mornings and explaining logarithmic scales to them so they understand what's going on with Coronavirus; he's still a programmer at heart, and one still in love, as they say at the Stanford d.School, with "the problem and not the solution". And as he says, the people who come to work for him seem to be cut from the same cloth, too; he claims the "world's best physicists, fastest compiler writers and best data scientists" are rocking up at 2307 Leghorn St, and we have no reason to doubt him.

A grey colored placeholder image