I smiled when Sayan Chakraborty, Co-President mentioned “AI Winter” in his presentation at the Workday AI and ML Innovation Summit last week. I smiled for two reasons:
He expressed eloquently in two words what took me much longer to describe in my 2016 book, Silicon Collar.
“Since the 1950s! That is when Alan Turing defined his famous test to measure a machine’s ability to exhibit intelligent behavior equivalent to that of a human. In 1959, we got excited when Allen Newell and his colleagues coded the General Problem Solver. In 1968, Stanley Kubrick sent our minds into overdrive with HAL in his movie, 2001: A Space Odyssey. We applauded when IBM’s Deep Blue supercomputer beat Grandmaster Garry Kasparov at chess in 1997. We were impressed in 2011 when IBM’s Watson beat human champions at Jeopardy! and again in 2016 when Google’s AiphaGo showed it had mastered Go, the ancient board game. Currently, we are so excited about Amazon’s Echo digital assistant/home automation hub and its ability to recognize the human voice, that we are saying a machine has finally passed the Turing Test. Almost.”
Also, a few days prior, Margaret and I had returned to the US via New York JFK. I have done so countless times especially in the 80s when I did multiple overseas assignments for PwC. Each entry required long lines, paperwork and interaction with human bureaucrats. Even my signing up for Global Entry a decade ago had not changed things much. Broken kiosks, clumsy fingerprint scans – the UX was not pleasant. This time, Margaret went to a manned line (she does not have Global Entry), while I walked into a completely empty room with my choice of twenty kiosks. While I fumbled for my passport, the associate guided me “Just take off your glasses”. A bit cynically I did and facial recognition took over and a few seconds later I was through. Wow!
It took Margaret another 20 minutes to go through her more analog process and while I waited for her, I reminisced about the hundred times (not kidding) our passports had been touched during our trip to India. With a ten-year visa which the Indian government would not transfer from expired US passports, we were each required to travel with two passports. Several times at each airport, at every hotel, every currency conversion, even entry to monuments like the Taj Mahal our 4 passports were eyeballed by a human and a scanner and likely printed and filed. Think of all the wasted time and paper, not to mention the constant worry of losing our passports. This when the Indian government already has digital records of our passports and visas. Talk about a stark contrast to my JFK experience – the term “AI winter” definitely struck a chord.
But I am an enterprise technology analyst. Most AI innovation we have seen in the last decade has come from the consumer world – speech to text, machine vision, etc. I have been excited about enterprise AI countless times and it always appears to deliver less than promised. Like when I interviewed Zach Nelson, then CEO of NetSuite in 2009 for my book, The New Polymath.
He had told me:
“We have some of the best leading indicators on the economy. We can aggregate order value, cash flow, and several other metrics instantly in our base of over 6,000 customers and watch trends.”
And I had written:
“Soon, those customers will be able to benchmark themselves against aggregated data of their peers. That would obviate the need for mailed–in surveys. That capability has been the domain of benchmarking firms like Hackett, not of the software industry, so that is another innovation NetSuite is working to deliver. Nelson explains: “Take those benchmarks and some of the creative BPO [business process outsourcing] partnerships we are exploring (such as one announced in April 2010 with Genpact, the spinoff from GE described in Chapter 3), and the industry could see SLAs [service - level agreements] that don’t just monitor technical metrics like systems availability but business process metrics that have been elusive to codify.”
The reality nearly 15 years later, is that much of enterprise data is still locked up in on-premise systems and private clouds/proprietary data centers that few vendors are allowed to access. Corporations love to see benchmark data from their peers – but have been (overly) cautious in sharing their own data. Bureaucrats around the world have complicated data gathering with their privacy and data residence regulations. The amount of enterprise data available to train computers is minuscule compared to the billions of images, voicemail messages and other data that Google, Microsoft, Amazon and others were able to collect without the data gathering restrictions they have used to train their machines.
Workday Co-Founder, Co-CEO, and Chair Aneel Bhusri had got me excited about enterprise vs. consumer AI when at an investor conference the week prior, he commented:
“With ChatGPT, you can ask good questions but you get some wacky answers at times, because the Internet is the Wild West... But our data is absolutely constrained, normalized, it's clean. And as a result, we can produce pretty powerful results.”
The point about caution about the consumer web as a data source for ML data was reinforced several times during the Workday event. Adobe’s event this week also focused on protection against the coming misuse of Generative AI and had coincided with deep-fake photos of President Trump being arrested.
Sayan had also elaborated in a blog post:
“Workday was an early adopter of large language models (LLMs), the technology that has enabled Generative AI, and we use them in production today. We have started adopting Generative AI at Workday to solve a host of additional customer challenges. A canonical case for LLMs is content creation, and we can see how drafting performance reviews, job descriptions, and a host of other documents will be transformed by this approach. We’re going to continue to identify key use cases where Generative AI can add value to our customers and develop unique models that leverage both Workday data and external data sets.”
Thankfully, the AI/ML Innovation Summit was not just about ChatGPT. Workday provided several optimistic cues about the broader evolution of Enterprise AI and their own role in shaping the journey.
Here are some:
Large enterprise databases to train machines and related customer permissions
Workday today services 60 million users. Skills Cloud was launched 5 years ago. Half of Workday’s HR customers have participated and shared their continually evolving employee skills with their peers. At its scale it would be impossible to maintain without machine support. And it gives Workday a template to build similar clouds around other enterprise data attributes.
Now, you may say, 60 million users is not that large. These users generate plenty of data - over 400 billion transactions a year. Very few enterprise vendors can show that scale – at least in their clouds.
Side note - CMO and Executive Vice President, Corporate Growth Pete Schlampp’s comments on Workday’s Rock Star commercial during this year’s Super Bowl brought home another significant impact of these 60 million users. Enterprise tech companies have had a patchy record advertising on that widely watched show, but, Workday’s brand is much stronger and their logo is more recognized in the consumer world because of this user base. Aneel’s comments about conversations with CEOs and politicians suggest we may also see Workday active on policy matters.
Analytics and ML
Since Workday acquired Platfora in 2016 and branded it as Prism, it has become a vital tool for its users to analyze cross-functional data especially that from vertical feeder applications. As a result, we heard plenty from Workday and from several customers on panels that Workday’s datasets are growing beyond its traditional Finance and HR focus. But the vertical perspectives are still skin-deep. From a process design perspective you want to ensure quality and accuracy at the source. Anomalies and variances should be presented to operational users like clinical professionals, not wait for a Financial or HR user to notice them. A big reason, I believe, why Workday will grow a solid core of vertical processes (in what it defines as service industries) and its base of operational users, employees and partners. I am a broken record when it comes to verticals but think their expanded ML focus will accelerate the demand for industry specific operational data. However, Workday is not likely to acquire a legacy vertical application. As Aneel said during the event “We are always looking for next generation players - cloud native using AI and ML for industry applications. We would never take something that is outdated technology because the first thing we have to do is throw it all away and rewrite it from scratch.”
Planning and ML
Since Workday acquired Adaptive Insights in 2018 it has become a planning powerhouse. The pandemic and the chaotic maneuvering at most enterprises the last few years have led to more continuous planning, and increased the need for more machine-driven planning. I also heard at the summit about an expanded range of planning in areas such as demand forecasting. All the more reason to mine external weather and vertical data. I wrote a case study in 2019 about Costco and demand forecasting at its Food Courts. Besides factoring weather and nearby events, their algorithms were also starting to factor gas prices and related impact on pump traffic. If a customer comes for gas, he/she is likely to also be a prospect for prepared food.
No scary job loss forecasts
Many vendors will use an event focused on AI and ML to talk about doom and gloom around job losses. Frankly, that complicates enterprise adoption even more. Sayan took a contrarian step. He presented a slide which shows that at least in developing nations, we face shrinking workforces. Machines will be needed to make up this shortfall.
In Silicon Collar, I had highlighted how even the smart folks at Oxford U. and Gartner had fallen prey to wild doomsday models of job losses from automation. I had presented several examples of slow adoption of automation technologies over the last century – UPC scanners, ATM machines, autonomous cars and others. I had summarized the examples in this Strategy+Business article. Job losses from automation tend to be gradual – and in many scenarios there are job gains. I also made the point that automation helps eliminate what are 3D - dull, dirty and dangerous - tasks. Automation reshapes jobs, not eliminates them.
Enterprise AI will do similar – think of the drudgery I described above that India will be able to reduce as tourism booms over the next decade. Air India just placed a massive order of 500 new planes in anticipation of this boom.
Of course, from the questions during the Summit, today’s smart folks appear to want to focus more on bias in AI, Federated ML and other contemporary topics that will keep Workday engaged from other angles.
Aneel’s comments
Aneel’s prepared and casual comments showed his belief we are at yet another inflection point in the industry. Having upended the client/server market with Workday’s cloud offering, he is acutely aware of architectural disruptions in our industry. The arrival of new Co-CEO Carl Eschenbach has clearly energized him and he wants to get back to more of a product and platform focus. He said he is looking for a condo in Boulder, Colorado to spend more time with Sayan and the ML team.
One of his comments summarized his sentiments “I am not afraid of Oracle or SAP. I am more afraid of a disruptive new enterprise player built on ML and AI.”
Bottom Line
I walked away feeling more optimistic that enterprise tech may finally be poised to exit its “AI winter” – and that Workday will play a key role in this transition.
Burning Platform: Raising the Bar for Analyst Summits with Truly Zoho
In the 83nd episode of Burning Platform, we host Sandy Lo and Pooja Mittal in the marketing team of Zoho and they talk about the elaborate planning and logistics it took to host the Truly Zoho analyst event in Chennai and Tenkasi in India.
You have likely read perspectives from several analysts including Brian Sommer, Jeremy Cox, Ankita Singh, Thomas Wieberneit and seen videos from Paul Greenberg and Brent Leary and others. It is not an overstatement to say it was one of the most talked about events in the industry, especially considering only 15 analysts participated.
However, the positive outcomes and experiences took months of planning, which as they say had backups to backup plans and extremely complicated logistics many of which changed on an hourly basis
Here they are summarizing many of the gory details in about half and hour and sharing some of the photos and social media from the event.
Raising the bar for analyst summits does not give it enough credit – they blew the roof off!!
March 27, 2023 in Burning Questions, Industry Commentary | Permalink | Comments (0)