FastForward #70: What baseball teaches us about AI
Hi everyone. Thanks for reading. I appreciate each and everyone of you. Just a heads up, that I'm headed on vacation at the end of next week, so there will be no FastForward for the next couple of weeks. Just a reminder, if you need a professional moderator for your event, drop me a line at moderate@fastforward.blog. If you like my newsletter, please share this week’s edition with a friend, and encourage them to subscribe, it really helps.💌 Sign up here.
ForwardThinking 🤔
What baseball teaches us about AI
As many of you know, I'm a Boston sports fan, especially the Red Sox. This year they are among the worst teams in the league, and it seems their story is a case study in how difficult transformation around AI can be. Ownership brought in a hardcore stats geek in Craig Breslow in Fall of 2023 to reshape the ball club. He proceeded to put his analytical stamp on the organization, and so far let's say, the results have not been terribly impressive with the worst statistical start at home since 1932 and a historically bad start overall.
Any executive who has tried to transform an older organization like the Red Sox knows there is going to be grumbling. But unlike most corporate executives, who answer to a board and shareholders, a pro sports exec like Breslow has to answer to the press and fans too, who think they know more than he does about running a team — and to be fair, given the current state, maybe we do.
Not everything has been an abject failure. The changes around pitching development have been successful for the most part, but when it's June, and your team is 33-46, and 13-25 at Fenway Park, well, clearly something is very wrong. We've been hearing for 15 years that data should drive decision-making, and in an age where there's more data than ever with AI to help parse the numbers, success should be at everyone's finger tips, right?
At the end of April, less than a month into the season, Breslow decided the problem was the manager and coaching staff and one wild night he fired them all, except for the pitching coach (whom he had hired) and a handful of others, and replaced them with his own guys. Little has changed. The team is still losing.
The firings couldn't cover up the weak roster that Breslow and his army of analytics nerds built. While injuries to star outfielder Roman Anthony and ace pitcher Garrett Crochet didn't help, the roster is still dominated by players who don't belong on a big league team, problems all of the front office's own making.
The human factor
This could be a case of simply over indexing on the analytical side of the equation. At the end of the day, as I've written in the past, players are human beings under those caps, and they bring with them all the emotional baggage and imperfections that we all bring to the table.
In an enterprise setting (which to be fair the Red Sox are ultimately), there is a growing belief that AI agents can solve that problem. But they aren't nearly as reliable as you might think, at least in my experience. I have created agents that worked fine for a couple of weeks, only to break or change in undesirable ways. Sounds a lot like us, maybe not for the same reasons, but unreliable all the same.
Look, every problem can't be solved by crunching numbers and throwing AI at it. Sometimes it takes the art, taste, subtlety and creativity that only humans can bring to bear on a problem. And I think that's what Breslow may have forgotten. Unforgiving fans refer to him as BresBot in the comments section of the Boston sports pages. It's a clear sign that he missed the fact that he wasn't just pushing numbers around a page, he's dealing with real humans.
Even as Breslow seems to recognize this, he hasn't been able to change the outcome. “If you are blindly following a model and knowing that the model is imperfect, you are going to make mistakes. The job that I have is to synthesize all of the information sources that we have. And we want to constantly improve all of that information, including a bunch of our models," he told the Boston Globe's Tim Healey in an article earlier this month.
Businesses face similar decisions about data and AI all the time. You may recall that Klarna learned a harsh lesson in 2024 when it replaced 700 human customer service agents with AI and framed it as an efficiency move. By last year, the CEO admitted he had mistakenly emphasized cost over quality, and some customers let him know in no uncertain terms that they preferred talking to a human, especially when it came to financial issues. Klarna responded by rehiring people. While it didn't abandon the AI altogether, it realized it wasn't applicable in every situation.
What AI and modeling can do is help a baseball staff sharpen a good player’s talent. It can’t magically turn a fringe player into a difference maker for the team. In the enterprise world, the same idea applies. It won’t turn a mediocre employee into a star, but it can help a strong one work faster and maybe reach beyond their core expertise.
You can have all the vision in the world, but if you are trying to bring about substantive change like the type we are seeing today with AI and agents, and you want to bring your people along, you have to stop looking at your transformation as a pure technology exercise and begin looking at it as a human one.
And that feels like something that Craig Breslow overlooked, or maybe never knew. Managers who make everything about the numbers and tech, while forgetting the employees who have to execute on their behalf, are going to find themselves like Breslow and the Sox: in last place with a lot of angry people wondering what went wrong.
~Ron
What's new on the blog 📰
Databricks continues to soar as private company, but margins starting to shrink
In this the latest collab between me and Alex Wilhelm, we dive deep into the success of Databricks, which continues to soar as a private company. Alex and I have been covering this company for a long time – my favorite part of the article is where I link to like a half dozen analysis pieces we wrote together over the years when we were at TechCrunch. I wasn't kidding when I said we've covered them a long time.
This will be cross-posted on Alex's Cautious Optimism blog.
Zscaler CEO Jay Chaudhry's latest bet focuses on protecting enterprises from AI-fueled threats
Speaking to Zscaler founder and CEO Jay Chaudhry, it's hard not to be impressed, even if as a journalist I'm supposed to be above such things.
From his back story growing up in a small village in India to his move to the U.S. with a one-way ticket and $200 in his pocket to starting 5 companies, four of which were acquired before launching Zscaler in 2007, he has had a long and successful run, and he shows no signs of slowing down.
I spoke to him across two conversations. One in April and one at the company conference in Las Vegas earlier this month where he talked about going after the threat AI adds to the security equation:
"While these models can generate code, they can also create exploits, and they can create a fix for the exploit," Chaudhry said. "Here's the difference: If a model created an exploit, and it only works 90% of the time, the attacker doesn't care. If your fix only works 90% of the time, you have a problem."
Note that I had a paid engagement with Zscaler when I attended their conference. My editorial standards and guidelines apply regardless. I have a full disclosure statement at the top of the piece.
FastForward on PPN podcast featuring Don Schuerman, CTO and head of marketing at Pega
In our latest episode of the FastForward on PPN, I spoke to Don Schuerman from Pega Systems. We talked about his longevity, spending 28 years at one company, the ever evolving market, the implications of content velocity, the company's recently announced outcome-based pricing and much more.
AWS's new FinOps agent may sound transformative, but really isn't
When AWS announced a new FinOps agent to help control cloud costs earlier this month, it would have been easy to think this was a revolutionary step in the right direction.
But according to cloud expert David Linthicum, it's not all that different from the cost control tools companies have been using previously, just in an AI wrapper.
"But let's be clear: this is a feature upgrade, not a paradigm shift. Enterprises should evaluate it as better automation within the AWS ecosystem."
~David Linthicum
Salesforce returns to its acquisitive ways to help fill in agentic gaps
I've been following Salesforce for a good number of years, and when they agreed to acquire Fin (formerly Intercom) this month, it got me thinking about a time when the company shut down M&A because of activist pressure.
News of the Week 📣
Salesforce introduces new Help agent with outcome-based pricing
Salesforce stock has been under a ton of pressure from investors because like all SaaS stock, Wall Street sees the per seat license fading to black in the agentic era. Salesforce has an answer to that with its new Help agent: outcome-based pricing where the customer only pays if the problem is resolved.
As Salesforce's EVP and GM for Agentforce Service (formerly Service Cloud) Kishan Chetan puts it, they have created "an opinionated out-of-the-box complete agent with the help agent." That involves three components including quick set-up, an enhanced portal experience and the move to outcome-based pricing.
Rebecca Wettemann, founder and principal analyst Valoir, says this release in response to a couple of problems that Salesforce has witnessed with customers. "Adoption of AI agents isn't happening as quickly as Salesforce and others would like because of AI FOMU (fear of messing up)," Wettemann told FastForward. "Pilots aren't moving into production because people are worried AI agents will either go off the rails and do something deleterious to their business or run up a big token bill overnight."
The outcome-based pricing means customers don't have to worry about that last part, and if it gets escalated to a human, well, they don't have to pay at all. As Wettemann points out though, figuring out what constitutes a satisfied customer is a problem.
People usually contact vendors because they have a big problem. Most of us know how to check the whereabouts of our package, or what terminal to go to at the airport. You call when a part is missing or broken, or your flight got unceremoniously rerouted and you're suddenly in a middle seat on a flight that takes off at 5 am.
That isn't usually the type of call a bot can handle, no matter how high quality the bot voice may be, and that's Salesforce's challenge here. Resolving issues that may be tricky without going to a human so they get paid is not going to be easy.
OpenAI makes move to control the stack with new custom chip
I've said it before, and it's worth repeating that OpenAI is still technically a startup, yet it's trying to be a full stack vendor like Google, AWS and Microsoft. This week, it took another leap at that goal with the announcement of a new custom chip.
The company that wants to do it all announced a partnership with Broadcom to build its very own custom inference chip called Jalapeño. They follow in the footsteps of Google's custom TPUs, Amazon's Trainium and Inferentia and Microsoft's Maia chips. They want to play with the big boys, who build custom hardware to power their solutions.
Here's how the company described it (in glowing terms, of course): "OpenAI’s first Intelligence Processor: an accelerator architected around OpenAI’s vision for the future of LLM inference, and the first AI accelerator in a multi-generation compute platform the companies are building together to make advanced AI faster, more reliable, and more accessible to more people."
That sounds pretty good if they can pull it off, but OpenAI is still technically a startup and it's attempting to put its fingers in multiple pies with a broad ambition instead of simply focusing on a couple of things and leaving the rest for down the road when they are more stable.
Sam Altman and company obviously don't see it that way, even if it's something that every VC knows when it comes to developing companies. You have to pick your battles, but like a precocious child, OpenAI wants it all and it wants it all now. We'll see how well that goes.
Accenture makes a hard move into cybersecurity with a flurry of acquisitions
Accenture is widely known as a consulting organization with hundreds of thousands of employees worldwide, but what if AI agents could start to take over some of the tasks that Accenture's highly paid consultants used to do? That would clearly threaten their business model.
But Accenture isn't sitting still waiting to be disrupted by outside forces. It's moving fast to take over adjacent markets and it sees a big opportunity in cybersecurity, a business that generated $10 billion for the company last year, out of a total of nearly $70 billion.
Ten billion isn't anything to sneeze at , so the company has to decided to expand that market with three recent acquisitions including a majority stake in Dragos and all of runZero and NetRise, three startups that raised $438 million, $20 million and $26 million respectively, per Crunchbase.
The company paid over $4 billion for the three companies with the goal of expanding its cybersecurity footprint. It's going to take more than these three companies to substantially move the needle though with the three generating revenue of just $208 million in ARR as of this month.
What I'm reading 📚
Tech stock slump could be a reality-check moment
~By Emily Peck, Axios
There should have been an op-ed here but you filed AI slop
~By Anna Moloney, CityAM
Boffin claims Microsoft's supposed quantum leap does not compute due to 'basic Python errors'
~By Thomas Claburn, The Register
What I'm watching 📺
The CEO of AWS on why Amazon is hiring 11,000 interns and junior employees
~Casey Newton, Platformer News
Look who's talking 👄
"Going around and slapping a bunch of bots on broken processes can maybe feel like we're doing stuff, but I guarantee you go through one winter and that patch is gone, and the pothole is probably going to be worse than it was before. If you want to make things better for people, you need to think about where the road really needs to go, and whether you're actually paving the right road for people to run on."
~Don Schuerman, CTO/Head of Marketing, Pega as told to the FastForward on PPN podcast.
