Sit next to an SDR for an afternoon and count the minutes they spend actually talking to somebody. I’ve done this more than once. It’s usually under two hours. The rest goes into moving between four browser tabs: a contact database, LinkedIn, the prospect’s careers page, and the spreadsheet where all three eventually get glued together by hand.
We’ve built an entire software category on the assumption that finding people is the easy part and selling is the hard part. Anyone who’s done the job knows it runs the other way. Tools like Lessie AI come at it from the opposite end: you describe who you’re after in a normal sentence, and the software handles the finding, the ranking and the first email as one job rather than three. Whether that particular product fits your team matters less than understanding why the older approach ran out of road.
Filters solved a shortage that doesn’t exist anymore
For about a decade the playbook was fixed. License a database, build a query out of dropdowns, export, start emailing.
That made sense when contact data was genuinely scarce. It isn’t anymore. Most of your competitors license from the same handful of providers, which means the list you pull on Monday is close enough to the list they pulled on Friday. Access stopped being the advantage. Precision became the advantage, and precision is the thing dropdowns are worst at.
A filter only sees what a profile field says. It will happily tell you someone’s title is VP of Operations at a 300-person logistics company. It can’t tell you they spoke at a supply chain conference in March, that their firm just signed a lease on a second warehouse, or that they posted last week complaining about the exact problem you sell against. Those facts live in conference agendas, regional business press, funding announcements and job ads. No dropdown reaches them.
So teams patch the gap with people. A rep pulls 400 accounts, then burns three weeks qualifying them one at a time. The database was never the bottleneck. The manual reading stacked on top of it was.
What changes when the software understands the request
The pitch behind agent-based tools is that the reading can be automated, since interpreting a messy request is roughly what language models are for.
In practice it collapses four things that used to be separate.
You start by writing the request the way you’d say it to a colleague: find heads of engineering at Series B fintech companies in Western Europe who are actively hiring backend developers. No boolean, no hunting for which dropdown holds “recently funded.”
From there the tool searches wide rather than deep. Lessie’s published figure is 100+ sources, covering professional networks, social platforms, podcasts, company sites and industry databases. The number matters less than the principle behind it. The facts that qualify a lead are scattered, so the search has to be scattered too.
Results come back ranked instead of dumped. Teams underrate this part. Two thousand rows isn’t an asset, it’s a chore somebody handed you. What a rep needs is the forty accounts worth calling this week and a reason for the order.
Then outreach gets drafted from whatever the search turned up, sent, followed up, tracked. Vendors in this space quote something like triple the reply rate against generic manual sending, which is about what good personalization has always produced. The novelty isn’t the personalization. It’s that the personalization survives being done at volume.
The quiet advantage in all of this is that nothing falls out between steps. In a normal stack, the research a rep did on Tuesday is gone by the time they write the email on Thursday. It lived in their head, or in a notes column nobody opens. When one system runs the whole sequence, it still knows why it surfaced a given contact when it writes to them.
Timing beats fit, and filters can’t see timing
Here’s the more fundamental issue with the old model: attributes are static and buying isn’t.
Industry, headcount, revenue band, tech stack. All of it tells you whether a company could plausibly buy from you. None of it tells you whether this month is different from last month. In B2B, timing is most of the deal. A company that just raised, just opened a second office, just posted six roles on the same team, just replaced its VP of Sales, or just churned off a competitor is a different prospect than it was in January. Identical filters. Completely different conversation.
Signal-based selling isn’t a new idea and nobody argues against it. The reason it doesn’t happen is labor. Somebody has to watch careers pages, funding news, product launches and executive moves across a few hundred accounts, and that somebody has never been hired.
Which is where breadth of sources stops being a spec-sheet number. A careers page is a source. A funding announcement is a source. A podcast where a VP walks through next year’s roadmap is a source. Search all of them in one pass and a request like “operations leaders at mid-market logistics companies that added capacity last quarter” becomes something you can actually run. Not because the contact data got better. Because the trigger and the contact came back together.
You can see the difference in the first line of the email. Attribute-based outreach opens with what the company is. Signal-based outreach opens with what just happened to it. Prospects clock that difference in about a second.
The part that isn’t really about technology
There’s a pricing story here that gets less attention than it deserves.
Doing outbound properly has always been something larger companies did, because it needed either a real SDR team or an enterprise data platform, and usually both. Sales intelligence suites tend to start north of $100 per seat per month before the add-ons that make them worth having. That puts disciplined prospecting out of reach for exactly the companies that need pipeline most, the ones where the founder is still the best rep on the team.
Agent-style tools starting around $35 a month don’t just undercut that. They break the assumption sitting underneath it. Per-seat pricing exists because the software assumes one human driving one pipeline. If discovery, scoring and first touch run as a single automated pass, that assumption stops holding.
Worth watching as a market shift rather than a feature comparison. When running a competent outbound motion gets an order of magnitude cheaper, a small company’s constraint stops being budget and becomes how clearly it can describe its own best customer. That’s a much better problem to have.
Questions worth asking any vendor in this category
A few things separate an actual agent from a database with a chat box stapled to the front.
Give it a compound request, three unrelated conditions in one sentence. Filter tools fall over. Agents at least degrade gracefully.
Ask how many sources it really searches, and which ones. Anything sitting on a single platform inherits that platform’s blind spots no matter how good the interface looks.
Check whether context survives to the outreach step. If you have to re-explain the prospect to the message generator, you still have a fragmented workflow, just with a nicer front end on it.
Ask about verified contact rates specifically, not accuracy in general. Discovery is worth nothing if the email bounces.
None of this takes judgment out of the job. A rep still decides which accounts deserve real effort, still runs the call, still reads the room on pricing. What it takes away is the three weeks of tab-switching sitting between you and the conversation you were trying to have.
