Since Salesforce acquired Heroku all the way back in 2010, developers and ecosystem analysts have often debated its role in the ecosystem.
Historically celebrated for its simplicity and ease of use, Heroku has at times felt disconnected, leaving users wondering when and how it might become more deeply integrated within the Salesforce platform. With the rapid rise of generative AI and Salesforce’s release of Agentforce, however, Heroku seems poised for a significant resurgence.
To take a closer look at Heroku’s recent developments and its evolving role within Salesforce’s broader AI and integration strategy, I sat down with Heroku’s CMO, Betty Junod, to discuss everything from the benefits and challenges of Heroku Managed Inference to customer experiences with Agentforce, recent platform innovations, and even the impact of a recent outage.
The Quiet Evolution of Heroku’s Role
Heroku is traditionally recognized as an approachable platform-as-a-service, favored by developers for deploying apps quickly without heavy infrastructure management. Yet, despite its advantages, the broader Salesforce community has often perceived it as peripheral – a tool sitting somewhat apart from core Salesforce services.
Recently, however, this perception has begun to shift. As Salesforce expands its AI-driven offerings like Agentforce, Heroku’s strengths, such as scalability, flexibility, and developer friendliness, are becoming more important for users in the AI age.
Companies now rely heavily on integrated AI capabilities to enhance customer experiences, drive automation, and streamline internal processes, which are areas where Heroku naturally excels.
“When you think about Heroku, we’re an AI platform-as-a-service within the Salesforce ecosystem”, Betty explained. “We already have great SaaS applications like Agentforce that work straight out of the box. Then we’ve got more infrastructure-oriented solutions like Data Cloud and MuleSoft. Heroku fits into that space.
“What we bring to this environment is the pro-developer platform. Inside Salesforce, you’ve got Apex, but not every developer works in Apex. With Heroku, we enable coders working in eight of the most popular programming languages to build services that plug into the Salesforce experience seamlessly.”
This shift isn’t just about expanding Heroku’s utility but also about rethinking where innovation in the Salesforce ecosystem will come from. As AI moves from a novelty to a necessity for the average Salesforce customer, the need for customizable, model-aware applications that live close to user data is growing fast.
But if Heroku is to shed its peripheral label and become a core enabler of Agentforce, it’ll need more than marketing alignment and prove it can handle the scale, reliability, and integration depth that modern AI workloads demand.
How Managed Inference Helps Developers
When speaking to Betty, she mentioned that one of Heroku’s strongest values is giving developers the platform to focus on building logic and not wrestling endlessly with infrastructure.
“The biggest challenge with developer experience has always been that we give developers all the tools, but most of them aren’t infrastructure experts”, Betty said. “They’re not security experts or networking experts either. What they are experts in is building the actual logic their app needs to work.”
This thinking is what underpins Managed Inference – one of Heroku’s newest features that lets developers easily add foundation models to their applications with simple CLI commands.
The Managed Inference tool makes it easy for developers to integrate powerful AI models into their applications without the complexity of managing infrastructure. It provides access to a carefully selected set of high-performing models, including language, embedding, and multi-modal models, that are optimized for the most common and valuable use cases.
With simple CLI commands, developers can spin up models, test prompts, and start building with AI in just a few steps.
On top of that, Heroku agents extend this functionality by allowing developers to build AI agents that can execute code, call tools, and interact with application logic inside Heroku Dynos. This means developers can create more dynamic, intelligent applications that interweave traditional code with AI-driven logic.
“By bringing Managed Inference into Heroku, we let developers stay in their CLI. It’s now as simple as running a command to pick from a list of foundation models or LLMs and drop that straight into their app, without needing to worry about choosing the right provider, setting up access policies, or managing infrastructure. It’s the same principle Heroku’s always followed for hosting apps and managing containers – we’re just extending that same simplicity to AI.” Betty Junod, CMO at Heroku
In essence, the Heroku Managed Inference continues Heroku’s hopes of simplifying infrastructure complexity and allowing AI deployment to be something developers can actually use.
As we already know, AI is changing everything, and with that comes a lot of confusion for users, as we’ve already seen with Agentforce. Heroku is working to keep the path as simple and accessible as possible, even as the pace of AI innovation accelerates.
“AI is moving so fast – what was hot three months ago might already be outdated,” Betty explained. “With every big shift like this, the entire ecosystem explodes with new options.
“What I often see is customers saying, ‘I’m looking at everything, but I’m not sure what to do.’ What’s great about our approach is that we’re offering very specific, constrained use cases. For example, start with your support agents, or in sales, it might be something like lead follow-up.
“The key is to keep it focused. If you give people a completely blank slate, they can fall into analysis paralysis and never get anything live. So the best path is to start small, learn and adapt, and then ask, ‘Where else can this save time or improve efficiency?’ That’s how real progress with AI starts to happen.”
The Power Outage: What Went Wrong?
Of course, Heroku’s big “elephant in the room” over the last month has been their recent outage. On June 10, a platform-wide disruption left many users locked out of their dashboards and CLI tools, while apps across the globe lost outbound connectivity. The incident stemmed from an automated system update that unintentionally broke routing rules for dynos, causing a chain reaction that even took down Heroku’s own status page.
The outage lasted nearly 24 hours and hit teams of all sizes hard, from solo developers to enterprise platforms.
When I prompted Betty on this – only a day following the outage – she kept her response brief: “Ultimately, customer trust is our number one priority. Keeping all of our services reliable and maintaining uptime is the most important thing.”
Many shared their annoyance at this outage online, with some suggesting they start implementing their disaster recovery plan and potentially migrate to alternative services.
To their credit, Heroku followed up with a detailed and transparent postmortem, outlining the root cause, the steps taken to resolve the issue, and clear plans to prevent a repeat. Their honesty and accountability went some way in easing tensions, especially for long-time users invested in the platform.
Still, the scale and nature of the outage raised serious questions. For many, it wasn’t just about the downtime, but how deeply it exposed weaknesses in Heroku’s internal tooling and communication infrastructure.
Rebuilding that trust won’t happen overnight. Developers and businesses alike will be watching closely to see if Heroku can back its words in the months ahead.
Looking Ahead for Heroku
At its core, Heroku has always been about helping people turn ideas into apps. Looking ahead, that mission hasn’t changed – but what it now means to be a “developer” definitely has.
In our conversation, Betty pointed to the rapid evolution of low-code tools as a signal of what’s coming next. “What’s most exciting right now is how the definition of low-code is evolving,” she said. “Tools like Cursor let people use natural language or diagrams to generate workflows, which eventually translate into code and run on a platform.”
This shift opens the door to a much broader set of builders – people who don’t write Java, Python, or .NET, but who can still create software using AI. And with millions of professional developers already in the world, what happens when millions more can now build, too?
For Heroku, that future could be a huge opportunity. If apps created through English prompts or flowcharts still need a place to run, Heroku could be the default engine powering them, regardless of whether the author is a software engineer or a product manager with ChatGPT.
Betty summed it up clearly: “I’m excited to see what kinds of applications start emerging, and wherever those apps are built, they’ll all need to run somewhere. I’m hoping Heroku can be the place that makes that seamless.”
Whether it’s AI-native agents or low-code user-generated apps, Heroku’s next chapter is likely to be defined by who’s building and how easy the platform makes it to bring those ideas to life.
Final Thoughts
The disconnected sentiment that the ecosystem has felt about Heroku has been long-standing, but the introduction of Agentforce and Heroku’s longstanding approach to new problems may be just the thing to realign the platform.
Heroku is going beyond being an optional extra and is looking like an essential piece of Salesforce’s AI puzzle. With the introduction of Managed Inference and tighter integration into the wider platform, the peripheral label attached to Heroku may be changing.
Of course, vision and reality don’t always move in sync. The recent outage proved that reliability will be just as important as innovation if Heroku wants to earn long-term trust.
But if Salesforce can deliver on the promise – giving developers a flexible, AI-ready platform that just works – then Heroku might finally find itself back in the spotlight as a launchpad for what’s next.
