All systems
System 03AI Sales Systems

An operating rhythm for the deals you're already working.

Account research, discovery prep, stakeholder mapping, follow-ups and MEDDICC notes — chained into one repeatable workflow you can run on every opportunity.

~6h saved / week
Intermediate
ChatGPT · Claude · Notion · Any CRM
What this system solves

Pipeline leaks where the system breaks.

Most reps don't lose deals to better products. They lose them to context that wasn't captured, follow-ups that didn't go out, and notes nobody could find on Friday.

  • Shallow account research
    You walk into discovery with the website tab open and a vague sense of who the buyer is.
  • Generic discovery
    The same five questions on every call — none of them tailored to this company or this role.
  • Slow, templated follow-ups
    The recap goes out the next day, sounds like every other rep, and lands in spam-tier attention.
  • Stakeholder blind spots
    You know your champion. You don't know who quietly kills the deal in week six.
  • MEDDICC as theatre
    The fields get filled in for the forecast meeting, not because they help you sell.
  • Sales admin tax
    Hours per week on CRM hygiene that the system should be drafting for you.
Who it's for

Built around real workflows.

Account ExecutivesFounders selling their own productSales leaders standardising the teamConsultants who sell as well as deliverRevOps building playbooks

If you're working a real pipeline — not just sending cold outreach — this system gives that pipeline a backbone.

Workflow overview

Six steps. One repeatable account rhythm.

Run them in order on a single opportunity first. Once the rhythm clicks, layer it over the rest of your pipeline.

01
Account research

Build a usable picture of the business before the first call.

How to run it: Prompt the model to summarise the company: what they sell, recent signals, strategic direction, and what would matter to them in your category.

02
Stakeholder map

Know everyone who touches the decision — not just your champion.

How to run it: Generate short profiles for each known stakeholder: role, likely priorities, and the specific angle that lands with each one.

03
Discovery prep

Walk in with sharp, tailored questions instead of a script.

How to run it: Ask the model to draft 5 discovery questions specific to this company, this role, and the outcome you want from the call.

04
Follow-up draft

Send a personal, specific recap within an hour — not a template the next day.

How to run it: Paste the call notes back and generate a 140-word follow-up: one specific callback, two key points, one clear next step.

05
MEDDICC capture

Update the deal record in a way that actually informs the next move.

How to run it: Have the model extract Metrics, Economic Buyer, Decision Criteria, Decision Process, Identify Pain, Champion, Competition from the notes — flag what's missing.

06
Next-step planner

End every cycle with one concrete next action you can put on the calendar.

How to run it: Ask the model to propose the single next step most likely to advance the deal, plus the email needed to set it up.

Example prompts

The prompts that run the system.

Tested across GPT-5, Claude Sonnet, and Gemini 2.5. Replace anything in {curly braces} before sending.

Account research
Act as my sales research assistant. Produce a one-page brief on {COMPANY}.

Sections:
1. What they do, in plain English (3 lines).
2. Buyer personas most relevant to {OUR_CATEGORY}.
3. Recent signals from the last 6 months: funding, hires, product launches, public posts.
4. Strategic priorities — what they're betting on.
5. Three angles for a first conversation about {OUR_CATEGORY}.

Tone: calm, factual. No hype. Cite sources where possible.
Discovery questions
I have a discovery call with {ATTENDEES} from {COMPANY} about {TOPIC}.

Draft 5 discovery questions. They should:
- Surface real constraints, not surface-level answers
- Be specific to this company and these roles
- Include at least one about current process, one about success criteria, and one about what's already been tried
- Avoid leading or pitch-style framing

Format: numbered list, no preamble.
MEDDICC extractor
Below are my notes from a call with {COMPANY}.

Notes:
"""
{NOTES}
"""

Extract MEDDICC from the notes:
- Metrics they care about
- Economic Buyer (named or inferred)
- Decision Criteria
- Decision Process and timeline
- Identify Pain — in their words
- Champion — and how strong
- Competition — incumbents or alternatives

For anything missing, flag it as "Open" and suggest the question to ask next call. No filler.
Example output

What the system actually produces.

Acme Logistics · Discovery prep · 30 min callAccount brief
Company
Acme Logistics — mid-market 3PL focused on cross-border e-commerce. ~$120M ARR, ~600 staff, HQ Rotterdam.

Recent signals
- Series C closed Q3 (€80M, lead: Index)
- Hired new CRO from Maersk in October
- Public push into US fulfilment — 2 new warehouses

Buyer personas
- VP Operations — owns SLA and capacity planning, cares about predictability
- Head of RevOps — newly hired, cares about pipeline visibility before the board
- CFO — focused on unit economics post-Series C

Three angles for the first call
1. RevOps tooling under the new CRO — likely a re-evaluation in flight
2. US expansion = new pipeline, new forecasting needs
3. Series C → board reporting cadence increases — visibility becomes urgent

Open questions to ask
- Who currently owns the forecast number day-to-day?
- What changed in how you sell after the CRO hire?
- What are you tracking weekly that wasn't tracked six months ago?
Common mistakes

Where this system tends to break.

Treating AI research as ground truth
Models hallucinate funding rounds and headcount. Verify anything before you quote it on a call.
Generic MEDDICC
Filling fields with one-word answers. The value is in the specifics — exact metric, named buyer, exact next step.
Over-personalising the follow-up
If the email reads like you spent 20 minutes on it, it sounds like you're selling. Keep it short and specific.
Skipping the stakeholder map
Most lost deals weren't killed by your champion. Map the room, including the people who aren't in it yet.
The human layer

AI runs the system. You run the deal.

Use the system to compress research, structure, and admin. Keep the parts that actually move pipeline — judgement on timing, the relationship with your champion, the call you make when things go quiet.

AI is for
  • Research and synthesis
  • Drafting follow-ups
  • Structuring CRM notes
You are for
  • Reading the room
  • Building trust
  • Calling the next move
Issue 001Shipping weekly

Weekly practical AI systems for modern work.

Useful workflows, operational systems, and practical AI implementation strategies for modern professionals. No hype, no trend-chasing — just systems that work.

Next dispatch · Monday 06:00 CET

No spam. Unsubscribe anytime.