FastForward #62: Get ready for headless everything
Hi everyone. As agents take center stage (at least in the hype cycle), this week I explore the notion of headless everything. If you like my newsletter, please share this week’s edition of FastForward with a friend, it really helps.💌 Sign up here.
ForwardThinking 🤔
Get ready for headless everything
Last month I attended the CRM Playaz IRL event in Atlanta, where I had a memorable conversation with Alan Berkson, who runs analyst relations at Contentful, a headless content management company. Headless is simply a term for separating the front end user interface from the back end code. It was popularized by content management vendors as a way to connect to the content in their repositories via APIs without having to futz with an interface.
That is precisely what agents do. They bypass the interface to do stuff. Humans need an interface, but agents, being digital, just need to speak code to code. Alan and I were discussing this at IRL and I told him I should write a post called 'Headless Everything.'
Last week, Salesforce announced Headless 360, which told me I'd sat on this idea a beat too long. So let's talk headless. Salesforce launched as one of the first popular SaaS companies with the tagline "The End of Software" There was software of course, but you ran it in the browser instead of locally. The latest tag line is "No Browser Required." Onward we march into the future.
It's worth noting that this isn't the first time that Salesforce dipped its toe into the headless waters. It was talking about headless commerce years ago, but this is clearly an answer to the expected switch to agents doing work alongside humans.
In a conversation with Aaron Levie earlier this year, we talked about the power of the headless idea as it relates to Box and content management in general in the age of agents. "This idea of having a headless content management system that an agent can read from and write to is incredibly powerful for an AI architecture," Levie told me.
Humans still need software too
But Levie doesn't see the end of the front end by any means. He sees a world where agents and humans will work on the software with each having distinctly different needs when it comes to interacting with the application. "Where we believe we have the killer app is while agents need to be in a headless fashion, humans need to be able to still go and browse the files and see them and see what the agent did," he said. And that should apply regardless of whether it's Box, Salesforce or any other application.
This is especially true, the higher the stakes involved in whatever the agent is doing. Twilio CEO Khozema Shipchandler, whose company is best known for its communications APIs, doesn’t see a world where agents work without human oversight.
"I don't think there'll be a human out of the loop ever. A lot of the things that we do are high value transactions that have to go well. I think giving that up entirely over to an agent is highly, highly unlikely," he said.
If they are right, that means most companies are going to maintain a dual development path, one for human users and one for the agents. But the agentic piece means that today, the headless idea is more relevant than ever as companies try to deploy them to take actions with far more intelligence than earlier generations of software. While we humans will continue to interact with software as before, if agents take off the way the hype suggests, this pattern will only grow more important, and we may indeed be heading into an era of headless everything.
~Ron
What's new on the blog 📰
How Google Cloud’s head of startups thinks about building companies in the AI era
Ahead of Google Cloud Next, I interviewed Darren Mowry who runs their global startup program. Mowry has a unique perspective having also worked at AWS and Microsoft in his career.
He says that AI coding has changed the speed at which companies can build, but it hasn't altered the fundamental requirements of building a business.
"There have been moments during the hype cycle where people were saying, let's throw the fundamentals out, but there's always a return to, do you have clarity of vision? Do you know who your customers are? Do you know what problem you're solving, and do you have a pathway to profitability? We always come back to that," he said.
An open source project turns 10 and finds itself tailor-made for the agentic AI era
It's not often that you get a full-circle moment in reporting, but today I published an article on the tenth anniversary of the open source Cloud Custodian, a decade after writing about its launch from inside Capital One for TechCrunch.
What's most interesting is that a project launched in 2016 is just as relevant today, and perhaps even more so as AI reshapes the infrastructure landscape.
PagerDuty ditches per-seat pricing as AI rewires its business
When I was at HumanX earlier this month in San Francisco, I sat down with PagerDuty CEO Jennifer Tejada to chat about the changes AI is bringing to her business, including the way the company prices its services.
"We have a customer who says every hour costs them a million dollars during a major incident. And if we can help them compress a major incident from lasting two hours to 20 minutes, that's a massive value. Well, that's a performance-driven use case," she said.
CrowdStrike CTO faces two-way challenge when it comes to AI
CrowdStrike CTO Elia Zaitsev was initially excited by the possibilities of large language models until he realized the attackers would have the same tools. I talked to him about the two-way challenge of building AI defenses while the adversaries use AI to rewrite the rules.
"Once the initial euphoria wore off, I started thinking, man, the adversaries are going to be using this stuff too."
Vibe coding won't kill SaaS, but complacency could
Everyone's talking about vibe coding killing SaaS, but the enterprise leaders I spoke to from Zuora, ServiceNow, LaunchDarkly, Forrester and SAP say it's not that simple. The real risk isn't where most people think it is.
"There have been a lot of lessons learned, and those lessons can't be vibe-coded."
~Cameron Etezadi, CTO, LaunchDarkly
News of the Week 📣
Amazon Bedrock AgentCore moves from component toolkit to managed agent harness
Amazon released Amazon Bedrock AgentCore last year as a componentized way to run and manage agents, bundling a runtime, browser, observability, identity, memory, and a code interpreter into a single platform. This week, the company went a step further by turning those pieces into a managed agent harness, making it much easier and faster for developers to create and operate the agents they build.
The problem with the componentized approach was that it still put a pretty significant burden on developers to create the orchestration layer for the agent before it built the agent itself. This layer is often called "the harness" in industry lingo. By introducing a managed harness, it reduces the amount of work and lets developers get to the agent building much faster.
And you have flexibility because it's built on top of an open source platform. "The harness in AgentCore is powered by Strands Agents, AWS’s open source framework. "When you need custom orchestration logic, specialized routing or multi-agent coordination, you switch from config to code-defined harness, with the same platform, same microVM isolation, same deployment pipeline," the company wrote in a blog post announcing the new feature.
It's probably not a coincidence that they chose to do this the same week Google was holding Google Cloud Next in Las Vegas. The games these companies play, but it's still a useful tool for developers as the agentic management layer evolves. The new features are available in preview.
Adobe launches CX Enterprise, the agentic system to rule them all
Adobe held its annual Adobe Summit this week and introduced a new comprehensive agentic system called CX Enterprise. The idea is to have one end-to-end agentic stack to replace the old Experience Cloud (because clouds aren't cool anymore, so they needed to change the name).
The idea is to bring together a bunch of different components including AI agents, reusable agent skills (packaged instructions that define specific tasks or workflows) and MCP endpoints (connections to other software). They put a governance layer on top to bring some discipline to the process, and encourage customers to keep humans in the loop to ensure accuracy and control, but the end product is automated marketing and customer experience workflows.
Adobe stock has plummeted in the last year, and if the last five days are any indication, the new announcement hasn't impressed investors yet. It's worth noting that long-time CEO Shantanu Narayen announced he was stepping down from the CEO role last month as the company struggles to deal with increasing competition from AI tools. He will remain in place until a successor is found.
Google announces Gemini Enterprise Agent Platform to bring long-missing integration layer
This week, Google held its annual Google Cloud Next customer conference in Las Vegas. There were announcements aplenty, from new TPUs to new security and data products, but perhaps the biggest was Google’s expanded Gemini Enterprise Agent Platform, which the company is calling as a more complete orchestration and management layer for enterprise agents.
To be fair, Google is also repackaging and extending existing pieces, including Gemini Enterprise, Vertex AI, and Google Workspace capabilities, into a more unified platform, while adding new app-layer and management features including Agent Runtime, Memory Bank, Agent Registry, Agent Identity, Agent Gateway and Agent Studio. These area all pieces enterprise developers should welcome as they try to build and manage agents.
As with Amazon Bedrock AgentCore, this piece is essentially what the industry is now calling "the harness." Jim Rowan, head of AI at Deloitte, says this kind of platform is critical for enterprise users building agents, not only from a management standpoint, but also in a way that could reduce costs.
"By layering capabilities into the harness rather than relying on the core model for heavy lifting, enterprises theoretically reduce token consumption and compute waste," Rowan told FastForward.
I'll be taking a deeper dive into this announcement next, so watch for it.
What I'm reading 📚
Spotify shares its all-time most-streamed music, podcasts
~By Kate Perez, USA Today
Apple, Amazon Push Back on Stricter Emissions Reporting Rules
~By Olivia Raimonde, Bloomberg
Google deepens Thinking Machines Lab ties with new multibillion-dollar deal
~By Rebecca Bellan, TechCrunch
Anthropic is thinking about removing Claude Code from its cheapest plan
~By Frederic Lardinois, The New Stack
Look who's talking 👄
"If you've got something super highly sensitive, highly confidential, things like data sovereignty, extremely high levels of regulation, do you want to be using a model that also is indexing on Taylor Swift's Greatest Hits?"
~Darren Mowry, head of Google Startups in his interview for this week's FastForward profile.
