← All playbooks

The AI-Native Sales Engine

AI doesn't fix weak sales, it just accelerates what's already there. Here's what changed for SDRs, the three stacks that actually work, and what you should never hand over to an AI.

What this is

Founders today face a confusing question: hire SDRs, plug in AI agents, rebuild outbound, or replace the sales team entirely? The premise of the question is already wrong.

Sales has already changed. Most of what an SDR used to do, sourcing, enrichment, drafting, follow-up, CRM hygiene, is already done better by tools. The work didn't disappear. It moved.

This playbook covers what shifted, the three AI stacks that actually work today, and the boundary that decides whether AI helps a sales engine, or just speeds up its chaos.

What changed (and what didn't)

An SDR in 2022 spent most of the week on operations. Today most of that operations layer can be done, or done better, by a stack of tools talking to each other. Here's the shift, task by task.

AI doesn't fix weak sales. It accelerates what's already there. Bad ICP plus AI = wrong leads, faster. Empty CRM plus AI = an empty CRM that updates itself.

Founder diagnosis: what's actually broken?

Most founders ask the wrong question first: "Should we add AI? Hire SDRs? Switch tools?" The right question is what is actually broken in the motion, because tools and headcount can't fix a broken layer underneath.

The 3 working stacks

Three combinations of tools that actually work right now. Pick the one that matches your stage. Don't try to run all three at once.

1. The outbound stack. For early-stage B2B with founder-led sales, building the first systematic outbound motion. Finds the ICP, enriches contacts, drafts personalised emails, runs sequences, captures replies and triggers follow-up.
2. The call-intelligence stack. For teams already running calls but losing the insight: notes scattered, follow-ups inconsistent, deal risks invisible. Writes summaries, extracts objections, captures next steps, flags deal risk, surfaces where reps are weak.
3. The founder copilot. For founders still closing deals themselves: losing context across deals, no second pair of eyes. Briefs before calls, breaks down accounts, drafts follow-ups, keeps deal notes, flags where the founder avoids the next step.

Important: not autopilot. Copilot. The human still drives the deal.

What to give AI, and what to keep

Give to AI:

• Research
• Email drafts
• Enrichment
• Call summaries
• CRM hygiene
• Reminders
• First-pass follow-ups
• Objection analysis
• ICP segment compare

Keep human:

• Understanding the customer
• Offer strategy
• Hard negotiations
• Closing enterprise deals
• Trust
• Pricing
• The moment to push, or step back

This is where founders fail most often. They automate the human part of sales, the listening, the negotiation, the close, and wonder why the conversion rate dies. AI runs the wheel. People drive the deal.

Hire, or fix the system first?

Hiring is the most expensive way to discover the system isn't ready. A bad SDR hire burns 30 to 60 thousand euros. A wrong Head of Sales burns 100 to 200 thousand and 6 to 9 months. An agency on a broken offer burns budget plus reputation in the inbox.

For most post-PMF founders, the right move is fix the system first, then hire. Hiring scales a working motion. It does not invent one.

A concrete end-to-end workflow

Ten steps, from a list of zero accounts to a CRM updated automatically after the call.

1. Pull the initial list of 200 to 500 ICP-fit companies. Filters: industry, headcount, funding stage, geography. No enrichment yet: validate the filter logic first.
2. Add firmographic and signal data: tech stack, headcount changes, funding rounds, hiring posts, leadership moves, public news.
3. Flag compound triggers, e.g. "raised seed in last 90 days AND hiring Head of Sales", that move accounts to the top of the list.
4. Score each account 1 to 100 on ICP fit. Anything below 60: drop. Don't waste a sequence slot on it.
5. Generate 2 to 3 angle variants per top account based on the trigger that fired. Founder reviews and picks one. AI drafts, human chooses.
6. 5-touch sequence: connect, DM, email, LinkedIn follow-up, email breakup. Inbox rotation and warmup handled automatically.
7. Auto-categorize replies into interested / referral / not now / wrong person. Founder only sees interested and referral.
8. Research dossier per booked call: company facts, recent news, likely objections, three discovery questions tailored to the trigger.
9. Personalized post-call follow-up using transcript plus brief. Drafted in 30 seconds, edited by a human in 90. Sent same day.
10. Call transcript, AI summary, auto-update CRM fields: next step, deal stage, decision criteria, blockers, stakeholders mentioned.

One operator running this pipeline produces what a 3-SDR team produced in 2022. The bottleneck moves from operations to judgment: who to talk to, what to say, when to push.

7-day implementation plan

1. Define one narrow segment, the buying trigger, top objection, and one-sentence pain. 200-word output max: if it's longer, the ICP isn't narrow enough.
2. Pull the list with filters that match the ICP one-pager. No enrichment yet: validate the filter logic against the ICP.
3. Enrich: funding rounds, hiring activity, tech stack, recent news. Every account gets 3 to 5 enrichment fields and at least one trigger.
4. Score 1 to 100 on ICP fit. Drop the bottom 60%. Keep about 80 high-fit accounts for the test.
5. Generate 2 to 3 angle variants per account. Founder picks one per account, in batches of 20.
6. Split the top 80 accounts into two groups of 40 and run two different angles. One is the control, the other is the test.
7. Read every reply by hand. Note which angle pulled, which triggers correlated with replies. Plan adjustments for week two.

Read every reply manually for the first four weeks. The founder has to feel the rhythm of replies, what hooks and what dies, before delegating triage to a model. Skipping this step is how you end up with a "working system" that ships generic messages forever.

The bottom line

AI doesn't replace sales. It replaces the weak manual ops around sales. If you don't have an ICP, an offer, a process, and someone who actually understands commerce, tools just speed up the chaos.

The winners aren't the ones who added AI. They're the ones who rebuilt the sales engine around AI.

The new model isn't AI instead of people, and it isn't people without AI. It's the right stitching: product, ICP, offer, copy, tools, agents, CRM, follow-up, human close. That stitching is what most teams are missing. Not tools. Not AI. The architecture.