Take a look at any software vendor's marketing materials and you're sure to see some variation of the word "intelligence" splattered everywhere.
It's part of a tectonic shift happening within enterprise technology. Companies spent the last several years moving their systems to the internet and, along the way, rapidly adopting new applications.
Now they're scrambling to try to compile information from those various systems to deploy analytics in an effort to predict the future. So instead of asking their Salesforce software to figure out the day's sales tally, executives want to combine that sales data with data from Gong to understand whether certain customers may be at risk of leaving, or with Marketo data to figure out how best to market a product.
That demand is forcing a major change among large software providers like SAP, Oracle, Salesforce and Workday that, for the past two decades, fought to be the home on the web for an enterprise's most valuable information. In the process, those providers eagerly created walled gardens around their systems, a design choice that may have helped customers get up and running on their platforms but one that is increasingly out of favor as more organizations seek to diversify their tech stack.
There's a new battle brewing as those so-called "systems of records'' and "systems of engagement" become commoditized and baked into IT infrastructure. And it's one where upstarts may have the advantage over the market leaders.
Legacy vendors "were counting on a lock-in of their data … and they aren't able to do that because customers are demanding that Salesforce and Microsoft be open. It's ushering in that next generation of company that can be more nimble," Emergence Capital Founder and General Partner Gordon Ritter, an early Salesforce investor, told Protocol. "It's a classic disrupter opportunity because the data that those incumbents built their systems on doesn't matter as much anymore," he added.
Startups like DataRobot, Databricks, Dataiku and Scuba Analytics all promise to break down those data silos and combine information from various sources to try to answer more forward-looking questions — or what some have labeled the "systems of intelligence."
It's clear there's investor enthusiasm for this concept: Databricks just reached a valuation of $38 billion. DataRobot sits at $6.3 billion, while Dataiku's recent $400 million funding round propelled it up to $4.6 billion. Snowflake commands a market cap of $94 billion.
The applications built and maintained by those companies allow businesses to make use of all the data stored within those various programs to begin to churn out predictions about the future of the business. So instead of using the information to populate real-time corporate dashboards, companies can begin to query the data to answer questions like: What products are customers going to want in five months?
And as the tech behemoths focus attention on expanding their already sprawling product suites, it's giving those upstarts a big advantage. So much so that the larger competitors are actually infusing tech from the upstarts into their platforms. Vendors like Adobe, Salesforce and Atlassian, for example, all use Databricks' tech to support their own intelligence features.
The trend is posing a major challenge for software giants that want to avoid becoming the next Oracle — a company that built an enormous business in an earlier era of technology and has struggled to grow against newer options. Still, given how entrenched companies like Salesforce and Workday are within organizations, it's still very much an uphill battle for startups.
"You get small, agile startups that are able to do things very well and, eventually, the big guys catch up and nick away that success," said Valoir analyst and CEO Rebecca Wettemann.
'They haven't thought about that openness'
The market leaders have been working for years to build prediction features into their existing products with varying amounts of success.
Oracle began to popularize the category back in 2007 with the launch of its "BI Applications" suite and it became a multibillion-dollar business. On the other hand, IBM failed spectacularly to turn Watson into an all-purpose AI engine.
More recently, Microsoft's Satya Nadella touted the concept of "systems of intelligence." And Google Cloud recently announced a new product that integrates information from various sources — like an ERP platform from SAP — to provide deeper insights into corporate supply chains.
"Intelligence needs to be leveraged much better," said Hans Thalbauer, managing director at Google Cloud. "We are focused on the aspect of data and making data available and accessible … not trying to create applications for the market" like an ERP app, he added.
Salesforce, however, is probably the most prominent example of a big enterprise software vendor embracing intelligent systems.
The company released its own AI engine called Einstein in 2016, but the product has struggled to find its footing. Salesforce claims it's running 100 billion "predictions" every day, and the company has lofty ambitions to merge Einstein more deeply into its data analytics tool Tableau, suggesting that it sees a bigger role for the tool despite one of the product's leaders previously conceding that customers aren't "going to do everything" with it.
"Tableau is a general-purpose data-analytics platform," said Gartner analyst Jason Wong. It's "not quite machine learning yet, but the Einstein analytics team is part of Tableau now. So there is a vision for Salesforce via Tableau to get involved in this area."
Still, unless Salesforce can significantly improve Einstein's abilities, it's unlikely to keep pace with the innovation happening outside CEO Marc Benioff's fiefdom. While Einstein might be great at analyzing Salesforce data, for example, companies rarely stay within one software ecosystem and instead are increasingly using a variety of applications from multiple vendors. In other words, Einstein is great when you want to generate a real-time sales forecast but becomes less useful when you want to know whether that forecast is within the typical range for that week of that quarter.
Salesforce has tried to address its challenges in multiple ways, including with Hyperforce, its tool that enables the company's products to work with all the major public clouds. It also recently launched a big partnership with AWS that enables information to flow more freely across the two systems. And, of course, it has MuleSoft, the API provider it acquired for $6.5 billion in 2016.
"MuleSoft is a platform that you can use to integrate every system at your company: Those back office systems, those legacy systems you like to pretend don't exist but aren't going anywhere, your supply chain, your ERP," said Bret Taylor, Salesforce operating chief and apparent heir to Benioff, at Dreamforce earlier this week.
But managing those connections on your own could be a much more difficult endeavor than using a packaged solution from a third-party vendor. And with engineering resources in short supply, CIOs may be hesitant to add yet another burden to an overworked IT department.
It's ultimately why the shift Taylor is talking about is a challenging one. Linking together a hodgepodge of legacy systems and new, digitally-native applications in hopes of developing AI-based models to predict future behavior or outcomes is no easy task.
And while megavendors have armies of engineers they can put towards solving that problem, they're simultaneously trying to get customers to use more of their own tools. It also requires them to rework many of their core systems that weren't designed to look outside of their walled gardens, which gives startups another advantage.
The market leaders are at "a little bit of a disadvantage because they haven't thought about that openness," said Wong. "That is a key part [of the problem] for these megavendors. Their own core platform architecture, is it in a state where they can implement some system of intelligence to really maximize it?"
One increasingly attractive arrangement is to centralize information from a multitude of systems into a data lake, or basically a vast repository of data. That enables organizations to then layer on a third-party tool to conduct analysis.
"Most of the systems and most of the data now is within the data lake. I wouldn't have made the same statement four or five years ago, so it's a change," said Dataiku CEO Florian Douetteau.
Modern vendors like Fivetran are trying to help push that trend along by enabling customers to simply dump their information from any system into a data lake, warehouse or lakehouse — a variation on the historical method known as ETL, where the data is changed from its raw form before it's deposited into a repository. The trend has become so pronounced that there is even a nascent industry focused on "reverse ETL," or taking that information out of Databricks after analytical processing, for example, and putting it back into Salesforce.
Ultimately, industry insiders speculate that the larger vendors will seek to tackle this opportunity by buying one of the up-and-coming "intelligence" vendors: Salesforce's tried-and-true approach. The problem? Those mergers are getting very costly. On top of that, the Biden administration has signaled it will take closer scrutiny of so-called "killer acquisitions," or those purchases that seek to nullify competition from nascent rivals.
It's one reason why Salesforce's $29 billion purchase of Slack perplexed some who thought the company needed to take a bigger leap into the intelligence market.
"That was a good step, a relevant step. But Slack is closer to a system of record than it is a system of intelligence," said Emergence Capital's Ritter. "It's a step in the right direction, but not a major step."
The classic question: Build or buy?
Getting such a system operational requires developing a unified understanding of how individual data will be treated. Information on bill-to customers, for example, must be handled differently than ship-to. And connecting populated corporate dashboards into a common collaboration platform like Slack requires building that link and then maintaining it as usage expands. It's one reason why vendors are eagerly expanding their partner ecosystems to make it easier for end users to manage this problem.
Any lapse in the system could mean that executives are making decisions based on operational snapshots relying on out-of-date or wrong data. The lift is so heavy that for every $1 spent on licenses for analytics tools, $4 are spent on services for it, according to Digital.ai CTO Gaurav Rewari.
"The amount of tooling that you need to purchase and the amount of scarce skills that you need to bring to bear to be able to stand up a solution like this for your own needs is daunting, time consuming and expensive," he told Protocol. "That's where the opportunity is, to productize the work that would normally be done by hand into a prebuilt system of intelligence applications in specific domains."
This process also assumes that companies are using the same applications forever, when in reality they are increasingly being switched out at faster cycles. A new head of sales, for example, could want to use Microsoft's CRM offering instead of Salesforce. While it's rarely cost advantageous in the short term to pivot to a new vendor, such a situation is becoming more common as data is, increasingly, stored outside the applications themselves.
And as those integration capabilities become more robust, it's no longer as necessary for enterprises to stick with one IT vendor. Instead, they can pursue so-called "best of breed" solutions and connect those to a central intelligence hub.
That's the case at Scuba Analytics, a startup that has raised $92 million to date, according to PitchBook. The company has some customers that integrate hundreds of applications together, per CEO Tony Ayaz.
"The market is going to be about intelligence," he said. And companies like Salesforce "need to gather better intelligence to not only run their own business, but serve their customers better. Because everyone is looking for actionable intelligence."
