05.09.2025

ForwardThinking 🤔

The age of AI-first mandates is here

Shopify CEO Toby Lütke turned heads last month with his company-wide, AI-first mandate. But it didn't take long before other companies like Fiverr, Box, Duolingo, HubSpot, Meta and Google quickly followed suit with similar pronouncements. Although these directives vary somewhat, they all share the core goal of encouraging employees to use AI to improve their work process and efficiency.

I didn’t get a mandate from my employer, but as the sole person responsible for producing this publication every week, I’m looking for any edge I can get, and I figured AI should be able to help, right? And, yes, it has the capacity to help me as I perform my tasks.

I’ve used AI for research, to query across multiple sources and as my editor. It’s worth noting that I still send my commentaries and profiles to a human editor for final review and she always finds something that the AI didn’t, the human difference, I guess.

Then there’s the matter of AI accuracy. Let’s just say that I’ve had mixed results with AI. When AI works, it works great. As I was preparing to write this commentary, I wanted to collect that list of companies I referenced in my opening. I went into Perplexity, and after some back and forth with the model, I asked it to compile the results in a table format with the company, the name of the CEO, the source and the result if available. It produced this:

Chart generated by Perplexity with names of the companies that have issued AI-first mandates.

Pretty great, I have to admit, but that’s not always the case. The other day, I dropped a pdf of an interview transcript into ChatGPT and asked for a summarization of a particular part of the transcript with quotes and timestamps to back up what it found, and it did something entirely bonkers: it made up several quotes. With an actual transcript to work from, it made up something that wasn’t even in it.

The trust factor

But even if AI worked flawlessly, there are deeper issues at play like trust and fear. As I wrote in this space several weeks ago, people are much more willing to forgive human mistakes than they are bot errors. When AI gives us bad information, it makes us wonder about the accuracy of all of the information we get from these models.

But it’s more than that. While I agree that we all should be exploring how these tools work, as Sol Rashidi, a former AWS exec, speaking at an event at RSA last week, noted, there is a fundamental resistance right now among some employees when it comes to AI, mandates be damned. And why wouldn’t there be in her view.

Stone bench with the word 'TRUST' etched in the side.

Photo by Dave Lowe on Unsplash

“How are we going to market with artificial intelligence right now,” she asked. “It’s the new digital workforce. We're going to increase productivity and capacity, which means, what are we going to do? Reduce headcount, decrease teams, be more lean – and you're starting to see a lot of that happen,” she said.

It’s a blunt truth. In her view, workers may see AI, not as a helper, but as a threat. “Well, why on earth would any knowledge worker document what they know into a PDF and push it into a large language model? You're threatening their longevity and their future. Why would they?” Why indeed. Self-preservation might suggest resisting AI, but if AI is the future, then survival may depend on understanding it. And that brings us to the training problem.

The training problem

Even folks who want to learn AI with the best of intentions may have trouble mastering these tools. I consider myself tech savvy. I’m willing to try new technologies and experiment and see how this all works for me, but not everybody is like that. For a lot of people, they know what they know when it comes to tech, and they get flustered when they get outside of their comfort zones, and that’s perfectly normal.

Yet these companies are asking, and in some cases, requiring, their workers to embrace AI fully. Except that most companies are not preparing their employees for this new reality. It requires a new way of working. People, who are not comfortable with technology, are not going to just suddenly “figure it out.” They’ll need support like peer mentors who are actively exploring the tools, formal help desks to answer questions and structured training programs. Unfortunately, in many cases, that support just isn’t there.

Group of people sitting around an oval conference table. Each has a laptop and a notebook. A facilitator is at the front of the room teaching them about something.

Photo by Campaign Creators on Unsplash

Consider that a recent survey conducted by AWS (which I wrote about the other day) found that just 56% of respondents had a training plan in place. That leaves a lot of companies kicking the training idea down the road (along with security). Yet we are seeing companies include AI skills in job postings, while requiring their current employees to start using it. There’s the AI. Have at it, folks.

I get it. AI has the potential to be a game changing technology. I’m not denying that, but for companies to simply expect that their employees can suddenly (magically) start using it to increase their productivity, well it’s not that simple. They still have their day jobs to do while they are supposed to be learning this new way of working. Will leaders demanding an AI-first approach tolerate a short-term dip in productivity while employees figure how out how to use AI tools in their jobs?

And as I have learned from personal experience, this is not easy. There are so many tools with new functionality coming at us almost every day. Mandates are easy. Transformation isn’t. Without training, support and time, these mandates will just be empty words.

~Ron

What's new on the blog 📰

Palo Alto Networks tries to stay agile by maintaining a startup mentality

I recently interviewed Palo Alto Networks chief product officer, Lee Klarich, who has worked at the company since its earliest days. I was particularly interested in the 22 acquisitions Palo Alto has made over the years, and how the company decides when to build and when to buy. Often it comes down to buying a startup when it has a unique expertise that the company lacks in-house.

"If you look across the different acquisitions, you'll see we've done more acquisitions in new areas like cloud security, than we have in, say, network security, which I think makes sense given our roots,” he said.

Coincidentally, I got a perfect buy example at the end of last month when his firm announced its intent to acquire boldstart portfolio company, Protect AI.

Read the full article >>

Palo Alto Networks chief product officer Lee Klarich, speaking at an event during RSA on April 28, 2025 in front of a sign bearing the company's name.

Photo by Ron Miller

Enterprises continue to struggle to find AI success

While the AI hype continues unabated, the data doesn't lie, and study after study is showing that enterprises continue to struggle to implement AI projects at scale successfully. The number has been stuck stubbornly at around 25% success rate for months. Two more studies, this time from AWS and IBM, continue to bear this out. I pulled out what I consider to be the key reasons.

Read the full article >>

How JPMorgan Chase partners with startups to drive innovation at massive scale

Larry Feinsmith is managing director and head of global tech strategy, innovation and partnerships at JPMorgan Chase. I talked to him about what it takes for startups to work with a bank with the scale of JPMC, which boasts a $17 billion IT budget with 44,000 developers on staff.

"When we go to market with the emerging technology ecosystem, it's not just to find the next trend. It's to look at these companies and how they can solve a material challenge or opportunity that we have."

Read the full article >>

JP Morgan Chase sign on Mission in San Francisco with George Rickey statue in the background.

Photo by Ron Miller

Security leaders know AI is inevitable, but struggle with trust issues

At a panel last month in San Francisco (where boldstart was a co-sponsor), CISOs talked about the inevitability of AI inside their organizations, and the real challenge of applying security policy in a rapidly changing environment.

Read the full article >>

News of the Week 📣

Another burst of open Ai news including a big acquisition

Another week, another flurry of OpenAI news. In a week full of meaningful happenings at the LLM firm, perhaps the biggest news was that it has agreed to buy Windsurf, the makers of the popular AI coding tool for a hefty $3 billion. Not too shabby for a 4-year old company that raised almost $245 million, per Crunchbase. Its last raise was $150 million Series C at a $1.25 billion valuation, just last August, so this appears to be sweet return on investment for the startup's investors -- and a fast one.

The company also announced it was bringing in Instacart CEO Fidji Simo to run the applications side of the house with the nifty title CEO of applications. She spent a decade at Meta before taking over at Instacart and brings some executive assistance for the other CEO, Sam Altman, who is reportedly having a hard time running everything as the company continues to grow quickly. She may have the CEO title, but she will still report to Altman.

Fidji Simo, former CEO at Instacart, who was announced as CEO of Applications at OpenAI on May 8, 2025.

Fidji Simo photo courtesy of Instacart

If that wasn't enough for you, the company announced some corporate governance changes. The firm was under pressure from none other than Elon Musk, an early investor, who was unhappy about OpenAI's plan to convert from not for profit to, well, profit. In an effort to split the difference, it converted to a public benefit corporation instead, while keeping part of the non-profit structure in place. It's a complicated maneuver that apparently didn't appease Musk anyway (nor a couple of state attorney generals).

IBM and Servicenow introduce new agent orchestration tools

We are starting to see the pieces of underlying AI agent infrastructure falling into place. A couple of weeks ago we talked about the rise of the communications protocols, MCP and Agent2Agent. While they might not be the final answer, they are a start. Last week we talked about the security piece, and today we are going address the orchestration layer.

If organizations adopt multiple agents from different vendors undertaking a variety of tasks, as the emerging vision suggests, they’ll need a way to manage and coordinate them. Enter the orchestration tool, which like a conductor guiding an orchestra, signals when each agent should join in or step back.

To that end ServiceNow announced AI Control Tower, while IBM announced IBM watsonx Orchestrator at their respective user conferences this week. It's a significant missing piece, that we're seeing fall into place, from two large firms used to working with enterprise-grade organizations.

Photo by Kazuo ota on Unsplash

Ray Wang, founder and principal analyst at Constellation Research sees an independent agent control tool as being absolutely crucial for customers. "Customers need a place to build, run and manage agents at scale. The future is cross-platform for Agentic AI and most customers don’t want to be locked in by their transactional vendor," he said.

Both of these efforts are clearly part of a continuing story to build the underlying infrastructure to make agentic AI possible inside large organizations. While we still have a ways to go before all of the issues will be resolved, this week's announcements were another step forward.

Microsoft adopts Google's agent2agent protocol

This week Microsoft announced it would be adopting Google's Agent2Agent protocol announced in April at Google Cloud Next.

The A2A protocol, as the name implies is designed to help agents from different vendors talk to one another, which is going to be absolutely essential if we hope to see agents working together inside organizations. Microsoft seems to recognize this, and without a lot of other options, decided to embrace this one, even if it's from a cloud rival.

Microsoft sign on building in Cologne, Germany.

Photo by Raimond Spekking. Used under Creative Commons Attribution-ShareAlike 4.0 on Wikimedia.

"As agents take on more sophisticated roles, they need access not only to diverse models and tools but also to one another. That is why we are committed to advancing open protocols like Agent2Agent (A2A), coming soon to Azure AI Foundry and Copilot Studio, which will enable agents to collaborate across clouds, platforms and organizational boundaries," Microsoft noted in a blog post announcing support for Google's protocol.

Google has indicated that the A2A protocol should complement the MCP communications, which enables agents to communicate with external services.

What we're reading 📚

Person reading book, but all you can see is their hands holding the book. They are sitting on a hardwood floor steeped in shadow, creating an interesting look.

LivPhoto by Blaz Photo on Unsplash

Decision Intelligence

Please cheat at your job

Cassie Kozyrkov

Platformer News

Stats from a dying web

Casey Newton

CNBC

Amazon says new warehouse robot can ‘feel’ items, but won’t replace workers

Kate Tarasov

Caller Calls Back

My bot will attend for me

Rafe Needleman

Look who's talking 👄

"To win, you are going to need to build special purpose models, much smaller, that are tailored for a particular use case, and that can ingest the enterprise data and then work."

- IBM CEO Arvind Krishna at the IBM Think 2025 Keynote this week.

Please subscribe to the newsletter and feel free to share.

3350 Virginia St, Suite 262, Miami, FL 33133

ForwardThinking 🤔

The age of AI-first mandates is here

The trust factor

The training problem

What's new on the blog 📰

Palo Alto Networks tries to stay agile by maintaining a startup mentality

Enterprises continue to struggle to find AI success

How JPMorgan Chase partners with startups to drive innovation at massive scale

Security leaders know AI is inevitable, but struggle with trust issues

News of the Week 📣

Another burst of open Ai news including a big acquisition

IBM and Servicenow introduce new agent orchestration tools

Microsoft adopts Google's agent2agent protocol

What we're reading 📚

Decision Intelligence

Please cheat at your job

Platformer News

Stats from a dying web

CNBC

Amazon says new warehouse robot can ‘feel’ items, but won’t replace workers

Caller Calls Back

My bot will attend for me

Look who's talking 👄

Subscribe

FastForward #21: The age of AI-first mandates is here

You may also like...