Autopilot
// Industry cases

What AC delivers specifically in your world.

From the GM perspective. Four industries, four real setups, four measurable outcomes. So you recognise what AC means for your type of mandate. No glossy cases — anonymised case profiles with numbers, sign-off block and an honest transferability question at the end.

// Small Mittelstand · machining · 18 employees

CNC contract manufacturer East Westphalia

Owner-GM · 18 employees · €3.2M revenue · 70% repeat customers

Repeat customer asks about AI quoting. AC: DXF auto-quote with human final review. Quote turnaround from 22 to 8 minutes — without a consulting sprint.

Setting

Owner-led CNC contract manufacturer, 3 anchor customers make up 70% of revenue. 14 quotes per week, manual calculation from DXF attachment. Q4 bottleneck — owner still 2 days/week on the machine, wife half-time in the office. Repeat customer asked last week: „Do you have AI-supported quoting?"

Pain

  • Concrete repeat-customer question on AI quoting — declining costs the contract
  • Own AI vendor pitch would overwhelm (owner + wife = entire management)
  • 2023 „digitalisation consulting" cost €8,000 and only delivered „cloud backup" — consultant trauma runs deep
  • No IT staff, no in-house tool-building possible

AC response

Standard retainer mandate focused on one use case: quote calculation from DXF with human final review. Make-or-buy first: don’t build, buy an existing EU tool (vendor from AC pool) and integrate into the workshop workflow.

  • Tool research: 5 vendors evaluated, 3 EU options identified, recommendation „Tool X" (EU cloud, €280/mo, 4-day setup)
  • Integration into existing workflow: incoming email with DXF → tool generates quote suggestion → owner checks 2 spot checks (material, manufacturing time) → signs off
  • Human final review as SLA: AC account lead reviews a sample of quotes per quarter, sign-off block documented
  • Repeat-customer interface: answer to the customer is „yes, AI-supported, with professional signature" — contract secured

Outcome

Quote turnaround 22 → 8 min
Repeat-customer requirement met
Year-1 cost €2,000/mo AC + €280/mo Tool X
Tool-selection time saved ~80h owner time
Sign-off block of quarterly review
Reviewed by
A. Müller (Account Lead AC, machinery)
Modell
Tool X (EU cloud Frankfurt, DPA per § 28 GDPR)
Risk-Class
minimal · EU-AI-Act
Anmerkung
Sample of 20 quotes Q3 — 18 standard OK, 2 corrections (complex geometry). Repeat-customer feedback documented.

What this means for your mandate

If you’re small Mittelstand with 8–50 employees and have one clear use case (quoting, offers, service knowledge) — no strategy sprint needed. Single retainer focused on the one use case, make-or-buy-first reduces investment risk. Trial 7 days without a card as entry point useful when the trigger isn’t acute.

// B2B distribution · wholesale · 240 employees

Technical wholesale Ruhr region

GM, 3rd generation · 240 employees · €78M revenue · 32,000 SKUs

2025 hallucination bot shut down. AC: human final review per product advisor output with source link. Amazon Business pressure addressed.

Setting

Family-owned B2B distributor for hydraulics, pneumatics, drive technology. 4,500 active B2B customers, 32,000 SKUs, quote hit rate 28%. Margin under pressure since 2023 — Amazon Business and specialised online distributors eat B2C-style into B2B. Own 2025 product-advisor-bot project shut down after 8 months due to hallucinations on detail questions.

Pain

  • 2025 hallucination trauma — association reputation sensitive, embarrassing incident spreads
  • Amazon Business pressure on the B2B business — margin erosion
  • 65 inside/outside-sales staff groan under backoffice load on quote intake
  • B2B commerce AI vendor pitch (€120k/year) declined — too much „Amazon-style B2B"

AC response

Standard retainer mandate with two prioritised use cases: productive product-advisor bot with mandatory human final review (no more hallucination risk) and outside-sales backoffice relief (auto-classification of incoming requests).

  • Human final review per product-advisor output: AC reviewer checks all detail-question answers before sending to customer — sign-off block with source link visible
  • Quarterly review with hallucination statistics: 2 hallucinations caught in Q1, both in new SKU category — mitigation built in
  • Outside-sales request classification: incoming emails auto-prioritised + pre-answered, outside sales saved 4-5h/week
  • Make-or-buy first: no in-house build, but integration of two specialised B2B tools (EU-based, DPA-compliant)

Outcome

Hallucination incidents (Q1) 0 → customer
Quote hit rate 28% → 33%
Outside-sales backoffice −4.5h/week per employee
CFO view (CapEx) 0 — all OpEx
Sign-off block of a product-advisor answer
Reviewed by
M. Zielinski (Account Lead) + S. Brandt (Specialist B2B distribution)
Modell
Claude Opus 4.7 + RAG over SKU master data (EU cloud)
Risk-Class
limited · EU-AI-Act
Anmerkung
Hydraulic cylinder specification in answer verified against supplier datasheet v3.2. No hallucination risk in this SKU category after Q1 mitigation.

What this means for your mandate

If you’re B2B distributor / wholesale and have your own AI tool trauma — AC is the counter-force: no new build, but existing tools with a human review layer. If you’re also under margin pressure from Amazon Business and online specialists, outside-sales backoffice relief is the direct lever. Contract points with output IP + cancellable monthly + single-tier fit the CFO view in the Mittelstand.

// Automotive supplier · TIER-1.5 · 620 employees

Electronics supplier Black Forest

Co-CEO family-owned · 620 employees · €145M revenue · TIER-1 customers VW/BMW/Daimler

TIER-1 audit asks for EU-AI-Act compliance. AC: Annex-III classification per use case, sign-off as audit material. No €90k strategy sprint needed.

Setting

Family-owned electronics supplier for automotive, three plants (DE/CZ). TISAX-certified, IATF 16949. TIER-1 audit questionnaires since 2024 demand „AI strategy and AI compliance" — answered with „in build-up" so far, that’s wearing thin. Own Q-data pattern recognition 2024 as successful PoC, productivisation planned for 2026. €90k consulting sprint by a top-3 consulting firm 2025 produced a 60-page maturity report, half of which was unrealistic.

Pain

  • TIER-1 audit Q3 2026 mandates binding EU-AI-Act compliance documentation
  • Own Q-data analysis with ML likely falls under Annex-III (high risk)
  • €90k consultant trauma sits in the supervisory board — no second strategy discussion possible
  • IG-Metall-active works council (11 members) — stumbling block on every AI rollout
  • 18 unfilled engineering positions — operational pressure, no slack for strategy

AC response

Standard retainer mandate focused on EU-AI-Act classification documentation per use case. Productivisation of existing Q-data pattern recognition with documented high-risk compliance. Works council consultation as mandatory step from day 1.

  • EU-AI-Act specialist in AC pool: classification of the three active AI applications — Q-data analysis as high-risk Annex III, service RAG as limited, marketing text as low-risk
  • Risk management system per high-risk application: risk analysis, mitigation measures, quarterly re-check dates — documented in Mandant-Memory
  • Conformity assessment high-risk (Annex III): documentation set for internal + external auditors, sign-off block per productive output
  • Structured works council consultation: WCA template as starting point, works council meeting with AC GM in Q3, anonymised adoption metrics instead of person tracking
  • Hosting class per use case: Q-data analysis on-prem (plant server), RAG EU cloud (Hetzner), marketing Anthropic with zero retention

Outcome

TIER-1 audit Q3 EU-AI-Act doc available
Supervisory board report clean, no reputation risk
Works council conflicts avoided WCA signed before go-live
Cost vs. top-3 consulting €24k/yr instead of €90k one-off
Sign-off block of a high-risk classification
Reviewed by
Dr. R. Engel (EU-AI-Act specialist AC) + M. Zielinski (Account Lead)
Modell
Internal Q-data model (on-prem, no external LLM)
Risk-Class
high risk · Annex III no. 4 (AI systems affecting employment relationship)
Anmerkung
Classification per EU-AI-Act Art. 6 + Annex III no. 4. Risk management doc as of 2026-05, re-check Q4 2026. Works council consultation 2026-04-15 documented.

What this means for your mandate

If you’re an automotive / industrial supplier with TIER-1 or corporate customers and feel audit pressure (IATF, TISAX, EU-AI-Act) on the supplier side — AC delivers audit material, not strategy decks. Single-tier at €2,000/mo is the base; with your complexity (Annex III high-risk, multiple plants) increased effort arises. In the first call we clarify where crew capacity hits limits — extension models possible, but transparent.

// Structural engineering · engineering firm · 180 employees

Engineering firm Northern Germany

GM + Co-GM · 180 employees · €22M fee revenue · 4 locations

Hamburg competitor advertises AI tender scout. AC: tender filtering + norm RAG with human final review. Research from 25h to 8h/week.

Setting

Northern German engineering firm for structural design, 4 locations (3 mid-sized cities + Berlin). 180 employees, €22M fee revenue. Three young project engineers spend a combined 25-30h/week on Eurofin tender research, hit rate 1:15. Hamburg competitor posts 2025 AI tender setup on LinkedIn — pressure in the VBI network. RFEM vendor announces 2026 AI module at premium price.

Pain

  • Competitor posting annoyed the GM for three days — research bottleneck becomes strategic
  • 25-30h junior time/week burned on tender search, hit rate 1:15 frustrating
  • Norm search (Eurocode, DIN) decentralised per project lead — duplicate work, hallucination risk on junior ChatGPT attempts
  • Half-day VDI workshop „AI in construction" Q4 2025 = 6h PowerPoint, no concrete output
  • Junior engineers actively seek career differentiation — tool frustration is a recruiting risk

AC response

Standard retainer mandate with two use cases: tender scout (filtering across EU procurement platforms with human review) and norm RAG (Eurocode research with clickable DIN reference). Both with sign-off block, liability stays with the engineer.

  • Make-or-buy: tender scout tool bought from the market (EU cloud, IONOS hosting, DPA) — no in-house development
  • Norm RAG with DIN-Verlag licence: answer contains clickable DIN reference (table, section, formula) — junior engineer verifies themselves, no hallucination
  • Sign-off with specialist reviewer: civil engineer with 22 years of practice reviews all high-risk answers (critical structural calculations)
  • Tech-champion programme: two junior engineers trained in adoption phase, take on multiplier role in the office
  • Hosting class: client data EU cloud, city specifications on-prem (Co-GM requirement)

Outcome

Tender research 25h → 8h/week
Norm search per project 5h → 40min
Tender hit rate 1:15 → 1:9
Junior frustration noticeably reduced (HR survey Q3)
Sign-off block of a norm RAG answer
Reviewed by
Dr.-Ing. T. Hartmann (specialist civil engineer, 22 yrs practice)
Modell
Claude Opus 4.7 + Eurocode RAG (DIN-Verlag licence)
Risk-Class
limited · EU-AI-Act
Anmerkung
DIN EN 1991-1-3:2010-12 Table NA.1 verified. Junior engineer checked source link + transferred to RFEM. Professional liability remains with professional (signed in the plan).

What this means for your mandate

If you’re an engineering or architecture firm with 50-500 employees and professional liability sensitivity (structural, building permit, structural safety) — AC delivers human final review as SLA, AI stays assistive. Norm RAG is the direct lever here (Eurocode, DIN, Beuth search), tender scout the second. Trial 7 days without a card only useful if a junior tech champion exists — otherwise direct first call with demo.

// Your mandate in detail

Don’t recognise yourself?

Four cases can’t cover every type of mandate. If your setup looks different — holding, group subsidiary, family AG, other industry — write to us. We say honestly whether it fits or not.

Note: These cases are anonymised case profiles based on real mandate constellations and industry data. Numbers are not fictional, but rounded to industry-typical values. We don’t do named-client cases — confidentiality is a precondition for the collaboration.