Summary of "Jak robić prospecting, research i ofertowanie wielokrotnie szybciej z AI – AI_Sales LIVE"
Core thesis (what the video teaches)
AI-enabled prospecting/research/offering becomes materially faster and more effective only when you combine:
- a sales method (playbook),
- consistent company context, and
- buyer/customer context—without leaking confidential customer data into public LLMs.
The presenters argue most firms chase more inquiries rather than improving conversion effectiveness from existing leads.
Key frameworks / playbooks (explicitly referenced or operationalized)
3-context framework for good AI outputs in sales
To produce useful, sales-ready outputs, they emphasize combining:
- Method context: a consistent salesperson workflow (how AI should help inside a repeatable sales process)
- Company context: who the vendor is, how it delivers value, and how customers buy
- Market context: what’s happening in the prospect’s market/competitive landscape
Then they add a crucial missing piece:
- Customer data context: a prepared value proposition + buyer persona + purchasing process for the specific buyer
- kept confidential
Buyer Persona / “Bayer Persona” model (their terminology)
- Not just “a person,” but a segment made of multiple internal roles in the buying committee.
- Deal cards auto-build a purchasing committee (e.g., estimator/PM/director roles) and map each role’s concerns.
Value proposition creation & iterative improvement
- Value propositions are refined via AI-guided questioning and experiments.
- Refinement shows up over multiple weeks:
- Week 1: context design
- Later weeks: iterative refinement
Hallucination control process
- Cross-check outputs across at least two models and/or by reading sources.
- They stress hallucination risk doesn’t remove the need to prepare for meetings—AI mainly reduces prep time.
Meeting/Call preparation playbook
Structure:
- opening
- discovery
- problem & cost
- why it happens
- stakeholder/role context
- positioning
- objections & responses
- next steps
AI-generated artifacts:
- deal/prospect cards
- pre-call briefs
- discovery question sets
- negotiation plans (including an example focused on salesperson negotiation consistency)
AI Sales workflow structure (operationalized via the “Seller” app)
- Assistants use consistent stages.
- Outputs are attached to a single deal card so later assistants can reuse the full history.
Concrete examples / case studies mentioned
Example 1: Preparing an “InPost meeting” from minimal vs. full context
They demonstrate that a “lazy prompt” (tasking AI to prepare a meeting) still yields some useful facts, such as:
- ~1.4B parcels delivered
- strategic challenges: acquisition and ownership transformation (including mention of “Advent in 2026”)
- market tensions (e.g., “war with Allegro”)
- international expansion via acquisitions
- entry into e-commerce/marketplaces
With full method + company + market context + their internal methods, the model produces a much more tailored meeting scenario, including:
- persona/segment fit differences (large scale, public company, formal decision process)
- specific implications for Sellwise’s value
Example 2: Deal-card-driven prospecting + CRM-linked notes (InPost)
- The app (“Seller”) generates a deal card immediately when prospecting starts.
- It stores “meeting/research history” on the card so different assistants share the same context.
- It integrates with CRM for at least:
- HubSpot
- Pipedrive
- It generates role-specific outreach:
- Sales Director message ≠ CEO message (different problems/goals)
Example 3: “Szympol” (wood/sawmills segment) with weak public data
They show AI can still build a prospecting plan even when:
- the company has no website / limited public info
- compliance/verification risks exist
Included red flags:
- lack of verifiable public signals
- decision-maker unknown → “check before contacting”
- inbound absence → reliance on traders/activity
Example 4: Mock conversation quality review (effectiveness focus)
A simulated discovery conversation was evaluated as weak by their “Seller” feedback, including issues like:
- no discovery questions asked
- generalized guesses instead of probing
- inconsistent next step vs customer signal
- missing details: priorities, must-have vs nice-to-have, budget, purchasing process, problem impact
They treat this as a system to generate actionable coaching feedback from meeting context.
Key metrics / KPIs and numbers mentioned (business-relevant)
Sales efficiency / effectiveness claims (inputs, time, and business value)
- Preparation time reduction:
- manually collecting similar meeting material: “2–3 hours”
- with their system: “much less time” (not precisely quantified)
- Small business leakage estimate from chaos:
- small service companies lose about 100k–300k PLN annually (range repeated)
- sometimes ~200k–300k PLN per year plus marketing/expenses
- ROI-style problem sizing examples:
- problems worth ~3M PLN/year
- 8M PLN/year
- and a large case of ~48M PLN/year
- (used to justify pricing by tying solution value to quantified problem impact)
Subscription / pricing timeline (program)
- “AI Sales” described as being on pre-sale until March 4th (price increases after).
- Pre-sale price:
- PLN 1,000 net cheaper per person
- Team discount condition:
- at least 3 people for better pricing
- Access timing:
- “Seller” access in training mentioned as starting only since April
- (after the buyer persona/value proposition context is prepared earlier)
- Training duration/workload:
- 6 weeks
- 6–10 hours/week, asynchronously possible
- Refund policy:
- after second week (if prospecting module/context isn’t right): full refund, no questions asked
Infrastructure / token usage (cost drivers)
- Token cost depends on usage frequency:
- example test spend: ~10 dollars during a full day of testing
- heuristic “normal use” budget: PLN 100–200
- Public model fallback costs:
- “from $20/month” for GPT/Gemini/Claude-style usage (not tied to Seller)
Actionable recommendations (how to apply what they teach)
-
Do not start with “a prompt.” Start with the workflow
- Define the method your salespeople follow.
- Build company and market context aligned to that method.
- Only then query AI.
-
Use AI to create meeting-ready artifacts
- Create a deal card containing:
- target segment + roles (buying committee)
- signals and hypotheses
- research summaries with citations (where available)
- Generate:
- pre-call brief
- contextual discovery questions
- objections + responses
- negotiation plan
- Create a deal card containing:
-
Guard against hallucinations
- Cross-check across models or verify sources.
- Treat AI output as input to preparation, not guaranteed truth.
-
Handle confidential customer data via API-based systems
- Don’t paste transcripts/emails/contract info into public chat tools.
- Their stance: enterprises use API querying due to privacy/security policies.
- They criticize “just uncheck training data” approaches as insufficient for enterprise compliance.
-
Tie AI output to CRM and a single source of truth
- Store assistant outputs and meeting history in one place (deal card + CRM).
- This prevents repetition and context loss across team members and time.
-
Make outreach role-specific
- Sales Director vs CEO messaging should reflect different incentives and decision processes.
“Seller” app: what it is and why it matters (operational strategy)
- Purpose: enable AI sales while keeping customer data confidential and maintaining full context continuity.
- Core capabilities:
- value proposition, buyer personas, purchasing process
- prospecting: generate deal cards on the spot
- research: deeper client research before outreach
- CRM: notes/updates to HubSpot and Pipedrive initially
- meeting prep: pre-call briefs + discovery questions + objection handling
- call/discovery feedback: evaluate conversations and extract coaching insights
- Pricing/access (as described):
- “Seller” is free for participants of the AI Sales course for a year
- Seller access/training access timelines depend on when persona/value proposition context is prepared
Presenters / sources (mentioned at end of transcript)
- Szymon (main presenter; host/coach)
- Dominik (demonstration request)
- Daniel (asks about CRM integration)
- Karol (asks about usage/approach)
- Grzegorz (previous webinar reference)
- Bartek Puck (mentioned for a separate AI implementation event)
- Dawid (asks about token cost/usage)
- Simon (references appear across the discussion)
- Sellwise / AI Sales (organization brand; product “Seller” created by the team)
Mentioned entities: InPost, Allegro, FedEx, HubSpot, Pipedrive, OpenAI, Gemini, Claude, GPT Cloud/Chat, Brave Education, AI Managers.
Category
Business
Share this summary
Is the summary off?
If you think the summary is inaccurate, you can reprocess it with the latest model.
Preparing reprocess...