Headless Software: Headlines vs. Reality

Benjamin Encz
Benjamin Encz
Co-founder & CEO

8 minute read

TL;DR: There's a lot of excitement about headless software as AI agents become more capable. Some of it is justified, but the idea that traditional software interfaces are going away doesn't make much sense to me. For occasional users, cross-system workflows, and automation, headless can be a huge improvement. For someone spending eight hours a day in an application, a well-designed UI is often much faster. At Ashby, we're investing in both: better interfaces for recruiters, alongside Custom Agents in Slack, Scheduled Agents, MCP, and an expanding API. I don't think the future is entirely headless. I think it's a combination of different ways to interact with the same software - more below.


Headless software is having a moment.

With AI agents getting more capable, there's growing excitement about a future where we no longer need to open individual applications to get work done. Instead, we ask an AI assistant to do something, and it figures out which systems to interact with.

This is already how a lot of our users interact with Ashby today, and it can be an incredibly valuable way to use the product for certain use cases.

But I also think the conversation is getting ahead of itself.

A lot of companies are pushing their teams to adopt AI, often through fairly broad mandates. That creates pressure to find new ways to use the technology, even when the problem we're solving isn't entirely clear.

One result is a tendency to talk about headless software as though it's the logical next step for every application.

I'm skeptical of that framing.

There's a much more practical way to think about headless: Which users and workflows are actually better served without a dedicated interface?

That's the question I think we should be asking.

Headless isn't really a new idea

The simplest way to understand headless software is that you can interact with a system without using its own interface. That might happen through an API, an AI assistant, an integration with another application, or an automated workflow.

AI makes this much more interesting, but the underlying idea has been around for a long time.

Think about business intelligence tools.

Companies have been pulling data from different applications into consolidated dashboards for decades. An executive who wants to understand how the business is doing probably doesn't want to log into their CRM, ATS, finance system, and five other applications to piece together an answer.

They want the information in one place, organized around what matters to them.

Meanwhile, the sales team still spends much of its day in the CRM. The finance team still works in its dedicated applications.

Nobody concluded that BI tools meant we should get rid of those interfaces.

I think recruiting software is a good example of why this distinction matters.

There are people who effectively live in an ATS. Recruiters, recruiting coordinators, and some hiring managers spend hours every day reviewing candidates, managing interviews, and moving hiring processes forward.

Then there are people who might interact with recruiting software a handful of times a month. Executives checking on hiring progress. Interviewers preparing for conversations. Hiring managers who only occasionally need a piece of information.

It's hard to imagine why these two groups should have exactly the same experience.

Where headless makes a lot of sense

You don't want to make everyone learn another application

If you're an executive who wants to know how a few important searches are progressing, opening an ATS, finding the right report, and navigating through the data might feel like unnecessary work.

Why not just ask the question in Slack?

The same is true for an interviewer who needs some information before a meeting, or a hiring manager who wants a quick update on their open roles.

These people don't necessarily need to become proficient users of recruiting software. They just need to get something done.

This is one of the areas where conversational interfaces are genuinely useful. AI makes it possible to retrieve information and take actions without requiring people to know where everything lives in an application.

And unlike traditional integrations, these experiences can be much more flexible. You don't necessarily need to anticipate every possible question someone might ask.

Sometimes the work doesn't fit neatly into one application

Another compelling use case is work that involves multiple systems.

A company might want to combine recruiting information with workforce planning data, internal HR systems, or other business applications. Historically, that meant building custom reports and integrations, or asking someone to manually pull everything together.

Now we're starting to see customers build agents that can work across these systems.

That's exciting because the workflow doesn't have to follow the boundaries of the individual applications anymore.

An agent can retrieve information from several places, reason across it, and potentially take action in multiple systems.

I expect we'll see a lot more of this, particularly in larger organizations with complex internal systems.

Sometimes nobody needs to interact with the software at all

This is probably the most overlooked part of the conversation.

We tend to think of headless as replacing a graphical interface with a conversational one. But why should every workflow require a conversation?

If a recruiting leader wants a weekly summary of hiring progress, they shouldn't necessarily have to remember to ask for it.

An agent can run on a schedule, analyze what's happened, and send the relevant information to Slack or email.

The best experience might be the one where you don't need to do anything.

And this goes beyond reporting. There are plenty of recurring operational workflows where removing the need for someone to initiate the work is more valuable than changing the way they initiate it.

Where I think the headless hype goes too far

I don't buy the idea that chat will replace the interfaces of most business applications.

One reason is that people often underestimate just how efficient a good graphical interface can be, especially for someone who uses it all day.

Take interview scheduling.

If you're a recruiting coordinator scheduling a large number of interviews, you need to consider availability, constraints, conflicts, and different possible arrangements.

Often, you want to see a lot of information at once. You might compare several options, move things around, or quickly explore what happens if you change one part of a schedule.

Doing all of that through a text conversation would be ineffective.

Imagine asking an AI to move an interview, then asking to see three alternatives, then asking about the availability of another interviewer, then changing your mind and going back to the original option.

A well-designed scheduling interface can make that entire process much faster.

Of course, AI can help with scheduling. It can automate repetitive steps, suggest solutions, and handle simpler cases without human involvement. We're investing in that ourselves.

But those improvements don't imply that the visual interface should disappear.

I think we're sometimes conflating two different ideas: using AI to make a workflow better, and using AI as the interface for that workflow.

They're not the same thing.

The broader point applies to most software used by power users.

People who spend hours every day in an application benefit from being able to scan information, compare things side by side, navigate quickly, and develop familiarity with the interface.

A conversation is great when you know what you want to ask. It's not always great when you need to explore information, maintain context, or perform a long sequence of related actions.

That's why I expect dedicated interfaces to remain incredibly important, even as AI becomes a much bigger part of how those interfaces work.

What we're actually seeing customers do

At Ashby, we're seeing both approaches emerge, often within the same organization.

Some of our largest enterprise customers are building their own internal agents on top of our APIs.

These companies have workflows that are specific to their business. They might have internal tools, custom processes, or requirements that wouldn't make sense to build into a general-purpose recruiting application.

We're working with design partners here, including leading AI companies and large publicly traded software companies.

One of the big priorities for us has been expanding the API surface so these customers can do more. That includes making more recruiting workflows accessible programmatically, such as creating jobs and sending sequences.

But building your own agents isn't the right answer for everyone.

Most companies don't want to build and maintain a collection of internal agents just to make recruiting information more accessible.

They'd rather have those capabilities built into the software they're already using.

That's where we're seeing a lot of interest in Ashby's own agents. Customers can create agents for their specific needs without having to build the underlying infrastructure. They can use those agents directly from Slack, or schedule them to run automatically and send information to the right people.

We've seen particularly encouraging adoption following the launch of Scheduled Agents.

I don't think one of these approaches is going to replace the other.

Some enterprises will continue to build deeply customized solutions. Many customers will prefer what we provide out of the box. And their recruiting teams will still spend much of their day in the core application.

That's a perfectly sensible combination.

Don't start by asking whether you should go headless

If I were evaluating headless recruiting workflows, I'd approach it much like a product manager would.

Start with the users.

Who are they? How often do they interact with recruiting software? What are they trying to accomplish? What makes that difficult today?

If you're trying to make hiring information more accessible to executives, a conversational interface could be a great solution.

If your team needs to connect recruiting data with several internal systems, an agent built on top of APIs might be exactly what you need.

If you're trying to automate a recurring report, perhaps nobody needs an interface at all (Ashby offer scheduled reports & dashboards).

But if you're trying to make recruiting coordinators more productive, you might get a much bigger improvement from a better scheduling UI.

I worry that when we start with a broad goal like "move to headless," we risk overlooking those distinctions.

A headless experience isn't inherently better because it's newer or uses AI.

It's better when it actually makes someone's work easier.

How we're approaching this at Ashby

We've been investing in several different ways to interact with Ashby, and I think that reflects where the market is going.

For people who want to work from Slack or through natural language, we've launched Ashby Assistant and Custom Agents.

Instead of building their own AI integrations, customers can create agents in Ashby that understand their recruiting workflows. Through our Slack integration, employees can ask questions and take action without having to open the ATS.

We've also introduced Scheduled Agents, which let customers automate recurring tasks and proactively send information through Slack or email. I particularly like this model because it moves beyond the assumption that every AI interaction needs to start with a prompt.

For teams building their own agents and integrations, we've launched our MCP Server and are continuing to expand it.

We're also investing heavily in our Developer API. We've documented 66 changelog updates over the past 90 days, covering new capabilities, improvements, fixes, and documentation. We're working closely with enterprise design partners to prioritize the additional functionality they need.

The ambition here is to make more of Ashby's underlying capabilities available to customers who want to build their own experiences around them.

And for people who spend most of their day in Ashby, we're continuing to invest just as heavily in the core product.

We want recruiters and recruiting coordinators to have excellent interfaces for managing complex workflows. We're incorporating AI into those experiences where it can save time or reduce manual work, without assuming that everything should turn into a chat conversation.

I see these as complementary investments.

A recruiting coordinator might spend their day in Ashby's scheduling interface. An executive might ask an agent in Slack how hiring is progressing. A larger enterprise might have internal agents performing specialized operations through our APIs. And a hiring manager might get an automated weekly update without having to ask for it.

All of those experiences can run on the same underlying platform.

That's ultimately what we're trying to enable.

I expect headless interactions to become a much bigger part of how people use business software. But I don't expect most power users to abandon dedicated applications, and I don't think we should want them to.

The goal isn't to get rid of the interface.

It's to make sure people don't have to use one when there's a better way to get their work done.

Share this post

Subscribe to our Community and Event Updates

Join our community of recruiting teams at fast growing companies.

ShopifyVantaSnowflakeLemonadeIroncladRampHackerOneFullStoryGorgiasZapierModern Treasury
ShopifyVantaSnowflakeLemonadeIroncladRampHackerOneFullStoryGorgiasZapierModern Treasury
VerkadaRetoolMarqetaRedditMercury
VerkadaRetoolMarqetaRedditMercury