An AI SDR is an AI employee that runs top-of-funnel prospecting: it discovers accounts matching your ideal customer profile, enriches prospects, researches companies, scores leads, and drafts outreach for a human to review and send. In Turtle’s Sales Engine pack, that AI SDR is Sasha — you define your ICP once, and the pipeline fills on a schedule instead of whenever someone finds a spare afternoon.
Every small sales team knows this trade. The person best at closing deals is also the person doing the prospecting — and one of those activities always loses. This post walks through what that costs, how an AI SDR runs the outbound grind, and exactly where the human stays in the loop, because “AI for outbound sales” without a human on the send button is just a spam cannon with better grammar.
A day in founder-led outbound
Suppose you’re a founder or an AE at a 10-person company doing your own outbound. Tuesday morning, you block two hours for prospecting. Here’s where they go.
Forty minutes on list building: searching for companies that fit your ICP — an ideal customer profile, your written definition of who actually buys — cross-checking size and industry, skimming websites to rule out the obvious misses. Thirty minutes on enrichment: hunting for the right contact, their title, their email, what the company just announced. Twenty minutes researching the two accounts that look genuinely promising. And then, with the last half hour, you draft three cold emails — the actual output of the morning.
Three drafts, two hours. Then a customer call runs long, Wednesday’s block gets eaten by a demo, and outbound goes quiet for nine days. That’s not a discipline problem. It’s arithmetic: prospecting is a volume game being played in the leftover minutes of people hired for judgment.
What does stalled outbound actually cost?
Two costs, and the second one is worse.
The first is the price of the hours. A dedicated SDR runs $70,000–$90,000 fully loaded in most US markets, which is why small teams don’t hire one — so the work lands on a founder or AE whose time is worth more. At a fully loaded $75/hour, an AE doing six hours of prospecting a week is spending roughly $1,800 a month on list building and first-touch drafting. Work a junior could do with a good brief. We walked through this cost-per-task math in the ROI of an AI employee.
The second is pipeline decay. Outbound is a pipeline with a lag: outreach sent this month becomes meetings next month and revenue the quarter after. When prospecting stops for two weeks — because a deal heated up, because someone was on holiday — nothing visible breaks. Then six weeks later the calendar is empty, and the team scrambles back to top-of-funnel work exactly when they should be closing. Feast, famine, repeat. The cruelest part: outbound stops precisely when things are going well, which guarantees a trough after every peak.
Consistency, not brilliance, is what the top of the funnel rewards. And consistency is what humans with other jobs cannot supply.
How does an AI SDR fill the pipeline?
The Sales Engine pack installs in about 20 minutes and centers on Sasha, an AI SDR. The setup is deliberately front-loaded into two assets you build once:
- ICP Profiles table — who you sell to: industry, size, geography, the signals that make an account worth pursuing. Your active ICP drives everything downstream.
- Product Knowledge Base — what you sell: your product details, positioning, the problems you solve. Every draft Sasha writes is grounded in this, not in generic sales-speak.
With those in place, Sasha runs the sequence your Tuesday mornings used to approximate:
- Discovers and vets target accounts against your active ICP — including an optional daily discovery run, so fresh accounts arrive every morning whether or not anyone remembered to prospect.
- Enriches prospects — and this part is automated on arrival: when a new prospect lands in the table, enrichment kicks off without anyone asking.
- Researches companies, so outreach can reference something real about the account instead of “I came across your website.”
- Scores leads, so your attention goes to the accounts most worth it rather than the ones at the top of the list.
- Drafts outreach — grounded in your product knowledge base and the account research.
Inbound leads and your deal pipeline live in tables the founder can watch — the whole funnel in one place, not scattered across a scraper, a spreadsheet, and someone’s drafts folder.
Sasha does the two-hour grind continuously. The half hour of judgment — is this worth sending, and is it right? — stays yours.
Will an AI SDR spam my prospects?
No, and this is worth saying plainly: Sasha drafts outreach; a human reviews and sends it. There is no spam cannon here. Nothing goes to a prospect without a person approving it.
That’s a deliberate design choice, not a limitation. Cold outreach carries your company’s name into inboxes of people you want as customers. A wrong-fit email at volume doesn’t just get ignored — it burns addresses, poisons domains, and teaches your market to delete you on sight. The economics of AI-drafted outreach only work if quality holds, and quality holds because a human signs every send. Reviewing a well-researched draft takes a minute or two; writing it from scratch took twenty. That’s the trade, and it’s the same draft-and-approve model Turtle uses everywhere.
One honest setup note: the Sales Engine needs an LLM provider plus the Apollo, Serper, Exa, and Google Search connectors — that’s where discovery and enrichment data comes from. And the pack ships with sample knowledge base content and an example ICP row you’ll replace with your own. Budget a focused hour to write your real ICP well; it’s the single highest-leverage input in the system.
What about the rest of the funnel?
Discovery-to-draft is the Sales Engine. Sales Engine Pro extends it across the full lifecycle on one shared data layer — seven tables covering ICP profiles, target accounts, prospects, website leads, CRM contacts, opportunities, and customer accounts.
On the outbound side, Sasha’s range widens: scouting buying signals, and drafting cold email, LinkedIn, and nurture outreach. Downstream, Maya, an AI Account Manager, picks up where the SDR motion hands off: qualifying opportunities, drafting and reviewing proposals, flagging pipeline risk, forecasting renewals and expansion, and keeping the CRM clean — the hygiene work every sales team swears it will do on Fridays. And for inbound, the companion Website Sales Concierge team routes visitor questions to product, pricing, and objection specialists and captures leads into the Website Leads table, so the people your outbound attracts don’t hit a dead end on your site. Proposals, like outreach, are drafted for a human to review and send.
Frequently asked questions
Will an AI SDR spam my prospects?
No. Sasha drafts every piece of outreach for a human to review and send — nothing reaches a prospect automatically. The AI does the research, scoring, and drafting; a person approves each send. That keeps volume honest and protects your domain reputation and your market’s goodwill.
What data does an AI SDR need to work?
Two things you define once: an ideal customer profile in the ICP Profiles table (industry, size, geography, buying signals) and your product details in the Product Knowledge Base. Every discovered account, score, and draft is grounded in those. You’ll also connect Apollo, Serper, Exa, and Google Search, plus an LLM provider.
Does this replace my SDR — or my own founder-led sales time?
It replaces the grind portion of either: list building, enrichment, research, scoring, and first drafts. Humans keep the judgment — approving sends, taking calls, running demos, closing. If you have an SDR, they work a bigger, better-researched pipeline; if you’re founder-led, your selling hours move from prospecting to conversations.
How is this different from a cold email tool?
Cold email tools send sequences to lists you build elsewhere. An AI SDR builds the list — discovering and vetting accounts against your ICP, enriching and scoring prospects — then drafts outreach grounded in your actual product knowledge, with a human approving every send. The work happens upstream of the send button, which is where most of the hours go.
If your pipeline depends on whoever has a spare afternoon, install the Sales Engine pack: define your ICP, paste in your product details, connect the data sources, and let Sasha run discovery for two weeks while you keep the send button. If it fills the top of your funnel, Sales Engine Pro is waiting when the deals start stacking up.