The Return of the AI SDR

The first wave of AI SDR companies arrived with a fairly simple thesis: much of the work involved in outbound sales is repetitive, so why not automate it?

The opportunity seemed obvious. SDRs spend a significant portion of their time researching accounts, identifying prospects, writing emails, managing follow-ups, and updating CRMs. These are activities that, at least on the surface, look well suited to automation.

But the first generation of products ran into a more fundamental problem. They underestimated how much judgment is embedded in seemingly routine sales work.

Knowing which prospect to contact, what angle to lead with, how to position a product against a particular competitor, or when to abandon an approach requires considerably more than the ability to generate a convincing email. It requires an understanding of the company, its product, its customers, and the market it operates in.

Early AI SDRs largely operated without this context. They could automate the mechanics of outbound, but struggled with the decisions that determine whether outbound is effective in the first place.

The result was predictable: a proliferation of tools that made it easier to send more messages, without necessarily making those messages more relevant or effective.

The category’s first attempt was ultimately constrained by the capabilities of the underlying models. They were good at generating outputs, but not yet reliable enough at absorbing context, exercising judgment, and carrying complex work through to completion.

That distinction matters. Because the opportunity was never simply to automate email outreach. It was to automate a meaningful portion of the knowledge work involved in generating revenue.

And for the first time, the technology is beginning to make that possible.

Why now?

There are three developments that make this moment fundamentally different.

1. Models are getting good enough to do knowledge work.

The biggest shift in AI isn’t that models can write better emails. It’s that they are becoming increasingly capable of understanding complex information and acting on it.

Modern models can work across large amounts of context, synthesize information from different sources, use external tools, and reason through multi-step tasks. They can take an objective, break it into smaller pieces of work, and carry those tasks through with increasingly less human intervention.

This is particularly relevant to sales.

A good salesperson needs to understand a company’s product, its customers, its competitive positioning, and the market it operates in. They need to research accounts, recognize relevant signals, identify opportunities, and decide how best to approach each prospect.

Much of this is knowledge work. It requires judgment, but it is also informed by information that can be collected, structured, and made available to software.

As models improve, more of this work can be delegated to agents.

We are moving beyond software that helps employees complete individual tasks toward software that can take responsibility for entire workflows.

2. Companies can encode their best sales judgment.

Every great sales organization has developed its own way of selling.

Its best people understand which customers are most likely to buy, which product capabilities resonate with different audiences, how to position against competitors, and which approaches consistently generate results.

This knowledge is one of a company’s most valuable assets. Yet much of it remains difficult to scale.

It lives in the heads of experienced employees, scattered across call recordings, CRM notes, internal documents, and informal processes. New hires spend months learning what the best people already know.

AI creates an opportunity to change this.

Companies can begin encoding their sales knowledge directly into software: their positioning, qualification criteria, customer profiles, messaging strategies, successful sales plays, and lessons from previous campaigns.

An agent equipped with this context can do more than follow a predefined sequence. It can apply the company’s judgment to new situations, make decisions consistent with how its best people sell, and execute those decisions at scale.

It can also learn from the results.

As agents run more campaigns, they accumulate information about which segments respond, which messages resonate, and which approaches generate opportunities. That feedback can be used to refine the underlying playbook.

In effect, companies can begin turning their best sales practices into something that is repeatable, deployable, and continuously improving.

3. A new economy is emerging around agents.

There is another shift happening that will fundamentally change how companies generate revenue.

Most of today’s software, distribution channels, and sales infrastructure were built for human users. But as agents become increasingly capable of taking action on behalf of people and businesses, a new ecosystem will emerge around them.

Agents will need tools, data, services, and infrastructure. Companies will build products specifically for agents to discover and use. New channels for reaching customers will develop, including ways to influence the recommendations and purchasing decisions agents make on behalf of their users.

We expect this to create an entirely new market for distribution and commerce.

The companies that build for this shift will need more than traditional sales software. They will need infrastructure capable of operating in an environment where both the seller and, increasingly, the buyer may be an agent.

The next generation of sales teams will be built around AI teammates.

We think the mistake is to view AI SDRs as cheaper replacements for human SDRs.

The more interesting possibility is to give every company an entirely new set of teammates.

Consider what a sales team could look like when it is no longer constrained by headcount.

A company could have agents responsible for researching specific markets, identifying buying signals, finding relevant prospects, developing account strategies, executing outbound campaigns, and managing follow-ups.

Each agent would have its own responsibilities, tools, and context. They would operate as part of a coordinated sales organization, rather than as isolated automations.

And unlike traditional software, these teammates could take on increasingly complex work as their capabilities improve.

A human salesperson might spend hours researching an account, identifying the right stakeholders, and developing an approach. An agent could do much of that work independently, then execute the resulting strategy across the relevant channels.

The point isn’t simply to do the same work with fewer people. It’s to make previously uneconomical sales activities viable.

Companies could pursue smaller market segments, run more experiments, personalize at a much greater scale, and respond to opportunities that would otherwise go untouched.

The best sales organizations will be those that can deploy the most effective combination of human judgment and machine execution.

This is where Cardinal starts.

We’re building Cardinal around a simple belief: every company should be able to build and deploy its own AI sales team.

Today, that starts with AI teammates that can research markets, find prospects, develop sales strategies, and execute outbound.

But we think the opportunity extends well beyond the traditional sales stack.

As agents become a larger part of how businesses operate, the infrastructure around selling will evolve alongside them. Agents will need ways to discover products, evaluate services, access information, and make purchasing decisions. Companies will need new ways to distribute their products and reach customers in an increasingly agent-driven economy.

This will create opportunities for entirely new kinds of tools and infrastructure: from helping companies sell directly to agents, to making products discoverable and accessible to autonomous systems.

We’re interested in building for that future.

For now, we’re starting with the sales team. It’s a very exciting time to build!