Use case — Functional area

AI in sales

Gut-feel sales becomes a structured process: ICP, data sources, scoring, briefings — all with an honest call on GDPR, German UWG and the thin line between good personalisation and politely packaged mass spam.

Sales starts with one question: who are realistic customers? In many SMBs this question is answered with Google, LinkedIn and gut feel — the result is an Excel list with two hundred names, of whom one hundred and eighty will never reply. The problem isn't the list, it's that there was never a clean target-group definition.

in sales shifts two levers: research speed and personalisation effort. An address list becomes a lead database with clear scores, trigger signals and concrete anchor points. Hours of preparation per first contact become a briefing readable in minutes.

What doesn't change: B2B outreach in Germany is legally tight. Cold mails without consent or a clear legitimate interest aren't allowed — UWG, GDPR and relevant rulings set narrow limits that no , however clever, gets around. Anyone cutting corners here buys cease-and-desists and a ruined domain reputation. That's the most important trade-off question and gets clarified first in every setup.

Three setup tiers

Which tier fits depends on three factors: team size, maturity of the ICP and data-protection ambition of your own industry.

Tier 1

AI assistant for research and outreach

Tool mix

Frontier LLM (Claude, GPT, Gemini) as a research assistant — directly in the browser or as an API for structured tasks
Sales tool with a database (e.g. Apollo, Lusha, Cognism, LinkedIn Sales Navigator) — firmographics, contacts, trigger signals
Manual ICP list in a spreadsheet tool or Notion — who's the target customer, which industries, what size
AI-supported mail drafts: briefing per lead into the LLM, draft out, human edits and sends
CRM (HubSpot Free, Pipedrive, Folk) — pipeline tracking, notes, follow-up dates

Fit

Sales teams of 1–3 people needing a research lever, without ambition for a fully automated pipeline. Fastest path to productive usage in sales.

Effort & cost

Setup 2–5 days. Running cost approx. €50–200/month (sales tool + + CRM). Scales with team size.

Trade-off

A lot stays manual — the lever is research speed, not volume. Anyone thinking “then I'll just mail 200 leads a day” quickly learns the GDPR, UWG and spam-filter reality.

Tier 2

Own lead database with scoring

Tool mix

Tier 1 in full scope
Own lead database (PostgreSQL) with firmographics, tech-stack signals and score — fed from sales tools, public web and possibly permissible scraping
n8n workflows for enrichment, score update, deduplication, CRM sync
Trigger-based sales signals: new funding round, job postings for key roles, tool changes, press releases
AI briefing per lead: a one-pager with a company snapshot, identified trigger signals and an anchor — the human decides what becomes of it
Weekly pipeline report: which leads were added, which score shifts happened, which topic clusters are emerging

Fit

SMBs with a defined ICP, multiple salespeople and a need for a structured pipeline. If the ICP question is still open, it's better to start with tier 1.

Effort & cost

Setup 8–15 days. Running cost approx. €100–400/month (database hosting, , sales tools, hosting). Scales with data and volume.

Trade-off

A lead database lives on upkeep. Stale data is worse than no data, because it creates false confidence. Without a clear owner for data quality, the system decays within a few months.

Tier 3

Self-hosted lead platform with knowledge graph

Tool mix

Tier 2 in full scope, data and AI workflows fully on your own infrastructure
Knowledge graph (e.g. Neo4j, Memgraph or a property graph in PostgreSQL): relations between companies, people, parent/subsidiary entities, tech-stack changes, customer referrals
Own embeddings + vector store for semantic lead search: “show me all leads that show similar triggers as customer X did before closing”
Optional: local language model (Llama, Qwen, Mistral) for lead briefings and outreach drafts — when lead data shouldn't go to frontier APIs
Audit trail of all data in and out: when did a lead enter the database, from where, on what legal basis — documented for GDPR access and deletion duties

Fit

Companies with high data-protection ambition (e.g. own industry is itself in B2B, lead data contains sensitive areas) or a need for structured relationship analysis beyond plain lists.

Effort & cost

Setup 15–30 days, plus running hosting cost from approx. €150/month. A local model on top only makes sense if cloud are deliberately excluded.

Trade-off

and semantic search are powerful but need attention. Without a clear for them (besides “sounds advanced”), tier 2 is better — complexity is cost, not status.

What your team should understand

in sales only works if the professional base is solid. Six competency areas that have to be anchored in every setup:

ICP — Ideal Customer Profile

Which companies are realistically customers and which aren't. Which industries, sizes, regions, trigger signals. Without a clean ICP, every lead pipeline becomes a list no one calls.

Understand data sources

Sales tools (Apollo, Lusha, Cognism) offer different data quality per region. LinkedIn Sales Navigator has its own usage rules. Own web scraping is legally tighter than many think — rulings in DE/EU narrow the room noticeably.

Lead scoring

Which signals raise the closing probability (company size, industry, tech stack, trigger events), which are pure noise. How scoring gets calibrated so it actually supports pipeline decisions — not just creates activity.

Compliance — GDPR, UWG, TTDSG

B2B email outreach in Germany is tight: consent or a concrete legitimate interest plus mandatory notices. B2B phone outreach is possible but not unlimited. Web scraping touches copyright, T&Cs and possibly the database directive. What goes wrong here rarely stays without a lawyer's letter.

CRM integration and pipeline discipline

How leads get into the CRM, how stages are defined, when a lead is disqualified. Which fields are mandatory, which add value. Discipline comes from clear definitions, not from more fields.

Outreach quality

Where good personalisation ends and mass spam with pretty variables begins. Which anchor points actually produce replies. Why three really relevant mails per day often beat a hundred with “{first_name}, I saw your website”.

What gets automated

Eight routine steps the pipeline takes over in running operations — at different depths depending on the setup tier:

Daily lead research

By ICP criteria, new hits are searched daily — new company entries, job postings, funding rounds, tool-change signals. Hits land in the database with a score, not in the inbox.

Enrichment with firmographics

Per lead, headcount, industry, tech stack and location are filled in from available sources — where licensing and data protection allow.

Score update

When trigger signals change (new funding round, relevant job posted), the lead score is adjusted and an alert may fire.

Duplicate detection

Same company under different spellings, subsidiary vs. parent, personnel change — automatic consolidation prevents the CRM from drowning in five versions of the same lead.

CRM sync

Qualified leads are automatically transferred into the CRM, with all research notes and score history — the salesperson sees at first glance why this lead is in front of them.

AI briefing per lead

A one-pager with a company snapshot, trigger signals and a concrete anchor — as preparation for the human first contact, not as a finished mail.

Trigger-based sales signals

Funding, personnel changes in key positions, press releases, tool changes — signals where a first contact now makes sense trigger an alert with a short rationale.

Weekly pipeline report

Which leads were added, where scores shift, which topic clusters are visible, what share of first contacts gets replies — narrative, not just numbers.

What stays MANUAL on purpose

Sales is relationship work, and is owner responsibility. These six points don't belong in a :

ICP definition and upkeep

Who is a target customer in which phase is a sales decision, not an algorithm question. ICP gets maintained as soon as the offer or the market shifts — otherwise the pipeline runs past current realities.

Compliance and legal-basis decision

Which sources get used, which deliberately not, what legal basis for first contact — this decision sits with the owner together with the data protection officer, not the .

Do the first contact

The first mail or call belongs in human hands. Preparation can be automated, the contact itself can't — writing style is now noticeable and drags the reply rate down.

Qualification call

Need, budget, decision structures, timeline — gets clarified in conversation, not read off a score. The score gets people on the phone, the conversation decides whether it goes further.

Offer and negotiation

Price talk, contract details, relationship work in the sales cycle — even good write sentences here that cost a deal. Manual, with clear responsibility.

Pipeline and relationship upkeep

Which leads get followed up, which frozen, which disqualified — upkeep decides whether the pipeline is a working tool or a graveyard of old score points.

How the build runs

From the ICP workshop to full self-operation usually 8–14 weeks, depending on tier, ICP maturity and data depth:

ICP workshop

Inventory of existing customers, identification of the traits they share — industry, size, trigger, decision structure. Out of it comes a sharp ICP that makes lead research sensible in the first place.

Data source and compliance audit

Which sales tools, which public-web sources, which own research are realistic — and what of that is legally clean. Clear split of what may come in and what deliberately doesn't.

Choose the setup tier

Research assistant, own lead database or platform — depending on team size, ICP clarity, data-protection ambition and volume. Reasoned recommendation, you decide.

Build the first lead list

A pilot list with 100–300 real leads per ICP, with scoring calibrated against known existing customers. Those scoring well there should also rank high in the list.

CRM integration and workflow build

Define CRM stages, mandatory fields clean, configure workflows for enrichment, scoring, trigger alerts and CRM sync.

Training & hands-on handover

4–6-hour workshop with the sales team: read ICP, interpret lead briefings, edit drafts, CRM discipline, basics.

Guided pilot month

Four weeks with weekly sparring: which leads came in, how did first contacts go, where does scoring need recalibrating, which data source brings what.

Self-operation with data-upkeep discipline

From then the system is yours. Optional: quarterly refresher on market changes (new tool in the stack, ICP shift, rulings) or a one-off database review.

Effort and investment depend on the chosen tier and the existing ICP and data state — a concrete estimate comes after the workshop and as part of the pricing overview.

Ready for the next step?

Free intro call, no strings attached. In 30 minutes you'll know whether and how AI can help your business.

Book a call Check eligibility