We added Salesforce Agentforce to our AI stack just over a month ago. We’d already added 10+ AI Agents at SaaStr and we didn’t know for sure how it would go. Not because we doubted the technology, but because we had some worries we’d need a Salesforce admin to make it work or see a similar set of challenges getting going at first. All AI Agents take work to train, set up, scale, onboarding, etc.
Turns out? No issues. While every AI Agent takes time to train, train AgentForce we did. And the early results are really, really impressive.
One month in to Agentforce, here’s what we’re seeing. And the results are really, really strong:
- 72% open rate (vs. 0% when humans never sent the email)
- Highest response rate of our AI platforms
- ~3,000 emails sent to previously ghosted leads
- Already closing deals from contacts that got zero follow-up six months ago
Not bad for our newest agent, deployed just over a month ago at Dreamforce.
The Gap We Needed to Fill
By the time we got to 2025 SaaStr Annual + AI Summit this past May, we had already deployed our first AI SDR tool for outbound and were starting to experiment with other tools. But we had a massive gap.
After 2025 SaaStr Annual + AI Summit, we discovered we had about 1,000 leads our sales team never followed up with. These were people who literally filled out a form saying they wanted information about SaaStr Annual. I routed it to a rep. And then… nothing. The rep did nothing.
This wasn’t inbound in the way our AI Agent Qualified handles it (instant, on-site conversations). And it wasn’t outbound in the way our AI Agent Artisan mostly operates (cold, highly personalized sequences). It was this middle ground – people who raised their hand, were already in our CRM (Salesforce), but got ghosted.
We needed AI that could solve for this specific use case. And we needed it to leverage all the data we already had in Salesforce.
Why Agentforce Made Sense
Here’s what I think most people miss about Agentforce: it knows everything your Salesforce knows.
That’s the beauty of it. It has all your Salesforce data if you click a button and want it to – and then you can use that to contextualize the emails it sends. Other AI Agents can hook into your Salesforce instance, and they do it well. But Agentforce is native.
For us, that was the magic. These 1,000 ghosted leads were already in our system and already in Salesforce. We had context on them. We knew what they were interested in. We just needed an AI agent that could:
- Access all that existing Salesforce data
- Send highly contextualized follow-ups
- Do it at scale without us manually managing each contact
The Setup: Easier Than You Think
We’re not Salesforce admins. Our Chief AI Officer isn’t Salesforce certified. In our ranking of tools in our stack, we were way better at Marketo than Salesforce – we pushed a lot of things from Marketo to Salesforce just because that’s what our Chief AI Officer was trained on earlier in her career.
But we learned a lot in this process. And here’s the critical thing: Salesforce helped us set it up. Just like Artisan helped us with Artisan, and Qualified helped us with Qualified. FDEs (Forward Deployed Engineers) are the magic behind the magic in complex and more enterprise AI agents.
This misconception that you need a third-party agency or a dedicated Salesforce admin to deploy Agentforce? Wrong. We figured this out with an FDE’s help. Now we need less help. Now we get how it works.
The truth is, once you learn one AI tool deeply, the others get easier. The concepts are surprisingly universal – even if the UX looks different. We actually copied all the instructions we put into one AI agent and just rewrote them for Agentforce’s specific use case. And it worked.
The Results: 1 Month In From Go Live
Remember, we only launched this at Dreamforce. This is our newest agent. And we’ve deliberately kept it capped while we learn.
Here’s what we’re seeing so far:
Total outbound sent: Still working through that initial 1,000-person list, sending roughly 3,000 emails/month
Open rate: 72%
Response rate: Highest rate of all our platforms (partly a function of who are reaching out to, but still impressive)
Let me say that again: 72% open rate.
You can’t get that in generic email marketing. You can’t get that with a human who never even sends the email (which equals zero). Even though this is early days, we’re pretty bullish on these numbers.
Why The Results Work
The magic is in the context. Because Agentforce lives in Salesforce, it can pull from:
- Past interactions with SaaStr, in many cases over many years
- Company information we already have
- Previous event attendance
- Any notes or data in their record
Here’s a real example. We sent a follow-up to someone we had ghosted. The agent referenced their past interaction with us and what we knew about their company. Kyle responded literally right away saying “no thanks but I’ll attend again.”
That would never have happened six months ago. Our sales team had decided not to follow up with these thousand people – nothing against them, they just had other priorities. The revenue was zero. Now it’s not. $600 (ticket) or $60,000 (Gold Sponsor), either way is a lot more than zero.
The Surprising Learning: Low-Hanging Fruit Is Actually Strategic
Some people might say we started with an easy use case – people who already raised their hand. Sure. But why wouldn’t you?
Salesforce itself did the same. They found they had 100,000,000 leads they never followed up on.
We already had so many other use cases running on our other AI agents. We wanted one that was near and dear to us because if we inbounded to somebody and they never followed up, we’d be salty.
Plus, these people were already in our Salesforce with context about them. The AI could send a personalized note acknowledging the relationship: “Hey, we dropped the ball, but here’s what we can help you with based on what we know about your company.”
Starting with this “middle ground” audience made perfect sense. They weren’t ice cold. They weren’t actively in conversation. They were just… forgotten.
How We Actually Trigger Campaigns
Since people asked: we can trigger Agentforce at the lead level, contact level, or campaign level.
In the early days, we did it very manually to make sure it wasn’t spamming 1,000 people at once. But Salesforce has since made this way easier. Now I can:
- Put contacts into a Salesforce campaign
- Have the entire campaign go live on the agent
- Let it intelligently parse things out – checking time zones, email send times, etc.
The agent queues everything up smartly. It’s not hitting everyone at 9am Pacific. It’s looking at where people are located and when they’re most likely to engage.
What We’re Not Doing (Yet)
Even one month in, we haven’t empowered Agentforce to book meetings itself. That’s deliberate.
For now, it’s asking for meetings rather than showing a direct booking link. Why? Because we’re still learning how it handles complex, multi-threaded questions. With another AI agents we use, for example, people sometimes ask about speaking AND sponsorships in one email. That requires human nuance.
I’ll get there with Agentforce. But right now,we’re using early success to build trust and understanding before we give it more autonomy. We do this with all our AI Agents. You should, too.
The Technical Reality: It Still Requires Time
Let me be clear about something: this isn’t set-it-and-forget-it. And it’s not for any remotely similar, complex AI Agent. Not today, and probably not in 24 months.
We still review draft responses before they go out for certain contact segments. We’re still iterating on the prompt builder instructions. We’re still thinking through which Salesforce campaigns we want it to exclude or target.
But here’s what makes it worth it: the agent’s output quality directly correlates to the time we invest in it. Just like with all our other AI Agents.
When we have more time to spend with the agent, it just plain performs better. When we’re busy (like right now with SaaStr AI London coming up), it still works on autopilot, but we see a drop-off in response rates.
That’s not a bug – that’s the nature of these AI agent tools right now. They clone your best processes, but they need you to keep refining those processes.
Costs & Budgeting Real Talk
People always ask about pricing. Here’s what we’ll say:
Most of these enterprise AI tools – whether Agentforce, Artisan, Qualified, or others – are in that $50K-$100K annual range to start when you factor in licenses, training, and onboarding. Larger enterprise will of course pay more.
Some are right below it. Some are right above it. Pricing fluctuates based on use cases.
But here’s the key: we reallocated headcount budget.
Around the time of our May SaaStr Annual and AI Summit, a few people naturally left the team. Instead of replacing that headcount (which was already budgeted), we replaced them with AI agents. We’re not saying fire someone to hire AI. But when people leave through natural attrition, that’s your budget. And a natural time to start looking, and implementing, a new AI Agent.
The other thing to budget: time. The direct tool cost matters less than the time investment from your best people to make it work.
Our Advice If You’re Considering Agentforce
Do:
- Start with a clear, specific use case (like our ghosted leads)
- Get help from Salesforce FDEs during setup – that’s what they’re there for
- Copy successful prompts from other tools you’ve used
- Be patient through the learning curve
- Demand to talk to technical experts, not just sales reps
Don’t:
- Assume you need to be a Salesforce admin
- Expect it to magically fix broken processes
- Try to deploy everything at once
- Give up if you have a bad experience with one rep – ask for someone technical
The Bottom Line
It’s only been 30+ days on Agentforce. We’ve sent but a fraction of what we plan to ultimately send through Agentforce in the run up to SaaStr Annual + AI Summit 2026 (and more). We’re still learning.
But a 72% open rate on a segment that previously got zero follow-up? That’s not incremental improvement. That’s transformational.
The people who dismissed Agentforce as “too technical” or “only for enterprises with dedicated admins” are missing the point. Like every AI tool worth using, it requires investment – in time, in training, in iteration.
But if you’re willing to put in that work? The middle ground between cold outbound and real-time inbound is massive. And Agentforce is helping us finally fill it.
We’ll have more data to share in Q1. But if you’re sitting on a bunch of leads that never got follow-up, or contacts that fell through the cracks, or people who raised their hand but never heard back?
This might be exactly what you need.
Want to see all our agents in action? Check out saastr.ai/agents for the full stack. And if you’re at SaaStr AI London next month, come find me – I’ll be diving deeper into all of this, including marketing and CS use cases we haven’t covered yet.
