What Data Should I Feed My AI Outbound Agent?
A sales leader's guide to AI outbound: why flat reply rates trace back to the brief, not the tool — and the one data type that changes everything.
Last updated: April 2026
This is a guide for sales leaders who are responsible for the results of an AI outbound program, not just the people running it.
It covers the right question to ask your team when reply rates are flat, the three data types that determine whether a brief will produce generic or exclusive emails, and how to evaluate the difference between a brief worth approving and one worth sending back.
What is the question most sales leaders aren't asking?
When AI outbound underperforms, the conversation usually goes one of two ways. Either the team wants to switch platforms, or leadership pushes for more volume. Both miss the actual problem.
The question that cuts to it faster: what data is in the brief your team is giving the AI? Not what tool they're using. Not how many emails they're sending. What information, specifically, is the agent working from when it generates each email?
The answer tells you almost everything about why the results look the way they do.
What are the three data types, and what each one means for your team's results?
Every AI outbound brief is built from some combination of three data categories.
As a sales leader, the most useful thing you can know is which category dominates your team's brief, that determines the ceiling on what the AI can produce.
First-party data is information your company has collected directly:
- CRM history
- Engagement with your content
- Prior conversations
- Product usage data
Its limitation is coverage — it only exists for accounts that have already interacted with you. For cold outreach to accounts that haven't, your first-party data is empty.
Third-party data is what you license from enrichment and intent vendors:
- Firmographics
- Funding signals
- Job changes
- Content consumption patterns
- Tech stack
Every competitor with the same vendor subscriptions has the same signals about the same accounts at the same time. Emails built purely on third-party data are structurally identical to your competitors' emails, because they're built from identical inputs.
Second-party data is what most teams are missing. It's information that exists in the relationship between your company and your partners — specifically which accounts in your pipeline already use tools integrated with yours, or sit inside a technology ecosystem connected to your partner network.
This data isn't sold by any vendor. It's exclusive to your company because it's a product of your specific partnerships. An email built on second-party data contains a premise no competitor can replicate.
What signals should an AI SDR use to personalize a cold email?
Not all signals are equal — ranked by exclusivity, here is what to prioritize when deciding what goes into your team's brief:
What would be the brief I'd send back vs. the brief I'd approve?
The fastest way to evaluate whether your team's AI outbound is set up to succeed is to look at the brief. Not the email it generates — the brief. Here's what that evaluation looks like in practice:
Imagine your SDR, Priya, is reaching out to David Kim, VP of Sales at Meridian Technologies. She submits this brief for review:
Brief one: The one I'd send back
Target: David Kim, VP Sales, Meridian Technologies
Context: Series B company, ~200 employees, B2B SaaS, Scaled sales team significantly in Q1 (30+ new hires), New VP of Sales, 60 days in seat, Showing intent on revenue operations content
Angle: team scaling fast, probably hitting coordination friction. We help with that.
This brief will generate a competent email. It will also generate an email structurally identical to what three of our competitors sent David this week — because every piece of context in it came from third-party enrichment sources they also have access to. David has seen this email. The VP of Sales hire is a known trigger. Everyone uses it.
Here's the brief I'd ask Priya to rewrite:
Brief two: The one I'd approve
Target: David Kim, VP Sales, Meridian Technologies
Context: Series B, ~200 employees, B2B SaaS — supporting context only
• Q1 sales team expansion — context for timing, not the hook• Meridian already runs [Tool A], [Tool B], and [Tool C] — all three connect to our partner network
• This means there is very likely account overlap in Meridian's pipeline that they can't currently see
Angle: lead with the integration connection. Offer to show David what the overlap looks like for Meridian specifically. This is information no competitor can offer, because it comes from our partner relationships — not from any enrichment vendor.
The email that comes out of Brief 2 is not just better, it's categorically different. The growth signals and firmographic context are still there as supporting color. But the argument is built on second-party data: the specific overlap between Meridian's integration ecosystem and our partner network. No competitor has that. Nobody can send that email except us.
How to coach your team to use second-party data?
Most teams don't use second-party data, not because they've evaluated it and chosen not to, but because they don't know it's available. Here's the practical path to changing that:
- Audit the current brief. Ask your team to pull up the brief template they're using. For each field, categorize it: first-party (from your CRM), third-party (from enrichment vendors), or second-party (from your partner relationships). Most teams find the brief is entirely first and third party.
- Identify which accounts have second-party data available. Pull the accounts in your pipeline that use tools integrated with your partner network. These are the accounts where ‘brief two’ is possible. Without Crossbeam or a similar Ecosystem Revenue platform, this is a manual exercise. With it, the list surfaces automatically.
- Build a separate workflow for those accounts. They get a different brief, a different email, and ideally a different follow-up sequence. Treat them as a separate segment — because the argument you're making is structurally different from everything else in the pipeline.
- Run a two-week test and measure. Compare reply rates between accounts worked with ‘brief one’ versus ‘brief two’. The gap is the number you use to make the case for investing in second-party signal infrastructure at scale.
The teams seeing meaningfully different results from AI outbound have done this work. They're not running better models. They're giving the same models a brief the competition can't replicate.
Find which accounts in your pipeline qualify for Brief 2
Crossbeam surfaces the ecosystem overlap between your accounts and your partner network — the second-party data that makes ‘brief two’ possible. Most teams find that a meaningful portion of their pipeline already has this signal available. They just couldn't see it.
Join Crossbeam for free and see which accounts in your pipeline already overlap with your partner network — before your next outbound campaign goes out.
FAQ
What data do AI sales agents need to write better cold emails?
The most impactful input is second-party data — information about how the prospect relates to your partner ecosystem, specifically which integration tools they already use that connect to your network. As a sales leader, the quickest way to assess this is to look at your team's brief and ask: Is any of this information exclusive to us, or could a competitor build the same brief from an enrichment vendor?
What signals should an AI SDR use to personalize a cold email?
Prioritize in this order: second-party signals (integration network overlap — only you have this), first-party signals (prior engagement — confirms interest, warms the opener), third-party signals (growth, job changes, intent — useful context, but shared with competitors). Most teams have the sequence backwards. Leading with commodity data and wondering why the emails feel generic is the predictable result.
What context does an AI agent need to send a relevant cold email?
The agent needs to know how the prospect relates to your ecosystem — not just who they are. For a sales leader evaluating this: does the brief tell the AI anything only your company knows? If everything in the brief could be assembled by a competitor from the same enrichment vendors, the email it produces will be indistinguishable from a competitor's.
How do I know if my team's AI outbound brief is good?
Ask one question for each field in the brief: could a competitor access this data from the same source? If the answer is yes across the board, the brief is built entirely from commodity data. A brief worth approving includes at least one input that's exclusive to your company — specifically, second-party data from your partner ecosystem that no enrichment vendor sells.
Why is AI outbound personalization stuck at surface level?
Because most teams have invested in making third-party data more comprehensive, rather than adding a different data category. As a sales leader, this shows up as increasingly sophisticated emails that still feel generic to the recipient — because the argument is built from information every competitor also has.
